wangdai commited on
Commit
80f3eb9
·
verified ·
1 Parent(s): 8414d99

Upload sensor_failure_benchmark dataset (1.4GB, 30770 files)

Browse files

RGB-D 失效对照数据集:24 episodes, MuJoCo backend, 140 frames/episode

This view is limited to 50 files because it contains too many changes.   See raw diff
Files changed (50) hide show
  1. commercial/acceptance.json +207 -0
  2. commercial/buyer_brief.md +19 -0
  3. commercial/coverage.csv +25 -0
  4. commercial/dataset_card.md +33 -0
  5. commercial/evaluation_template.csv +25 -0
  6. commercial/pair_comparison.csv +13 -0
  7. commercial/quote_assumptions.json +21 -0
  8. commercial/sample_index.csv +0 -0
  9. config.json +34 -0
  10. dataset/README.md +7 -0
  11. dataset/data/chunk-000/file-000.parquet +3 -0
  12. dataset/meta/accepted_pairs.json +24 -0
  13. dataset/meta/episodes/chunk-000/file-000.parquet +3 -0
  14. dataset/meta/info.json +187 -0
  15. dataset/meta/input_contract.json +19 -0
  16. dataset/meta/normalization.json +16 -0
  17. dataset/meta/quality.json +32 -0
  18. dataset/meta/scenarios.json +1562 -0
  19. dataset/meta/splits.json +108 -0
  20. dataset/meta/stats.json +443 -0
  21. dataset/meta/tasks.parquet +3 -0
  22. dataset/videos/observation.images.front_observed/chunk-000/file-000.mp4 +3 -0
  23. episodes/episode_000000/depth/front/000000.npy +3 -0
  24. episodes/episode_000000/depth/front/000000.png +3 -0
  25. episodes/episode_000000/depth/front/000001.npy +3 -0
  26. episodes/episode_000000/depth/front/000001.png +3 -0
  27. episodes/episode_000000/depth/front/000002.npy +3 -0
  28. episodes/episode_000000/depth/front/000002.png +3 -0
  29. episodes/episode_000000/depth/front/000003.npy +3 -0
  30. episodes/episode_000000/depth/front/000003.png +3 -0
  31. episodes/episode_000000/depth/front/000004.npy +3 -0
  32. episodes/episode_000000/depth/front/000004.png +3 -0
  33. episodes/episode_000000/depth/front/000005.npy +3 -0
  34. episodes/episode_000000/depth/front/000005.png +3 -0
  35. episodes/episode_000000/depth/front/000006.npy +3 -0
  36. episodes/episode_000000/depth/front/000006.png +3 -0
  37. episodes/episode_000000/depth/front/000007.npy +3 -0
  38. episodes/episode_000000/depth/front/000007.png +3 -0
  39. episodes/episode_000000/depth/front/000008.npy +3 -0
  40. episodes/episode_000000/depth/front/000008.png +3 -0
  41. episodes/episode_000000/depth/front/000009.npy +3 -0
  42. episodes/episode_000000/depth/front/000009.png +3 -0
  43. episodes/episode_000000/depth/front/000010.npy +3 -0
  44. episodes/episode_000000/depth/front/000010.png +3 -0
  45. episodes/episode_000000/depth/front/000011.npy +3 -0
  46. episodes/episode_000000/depth/front/000011.png +3 -0
  47. episodes/episode_000000/depth/front/000012.npy +3 -0
  48. episodes/episode_000000/depth/front/000012.png +3 -0
  49. episodes/episode_000000/depth/front/000013.npy +3 -0
  50. episodes/episode_000000/depth/front/000013.png +3 -0
commercial/acceptance.json ADDED
@@ -0,0 +1,207 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "summary": {
3
+ "product_id": "sensor_failure_benchmark",
4
+ "coverage_split": "matched_conditions",
5
+ "independent_scene_pairs": 12,
6
+ "status": "technical_pass",
7
+ "passed_checks": 7,
8
+ "total_checks": 7,
9
+ "coverage": {
10
+ "expected_cases": 24,
11
+ "observed_cases": 24,
12
+ "complete": true
13
+ },
14
+ "split_counts": {
15
+ "train": 8,
16
+ "validation": 8,
17
+ "test": 8
18
+ },
19
+ "customer_validation": "not_run",
20
+ "review_reasons": [],
21
+ "accepted_recovery_pairs": 22,
22
+ "negative_recoveries": 2,
23
+ "rgb_images": 3360,
24
+ "invalid_label_samples": 0,
25
+ "fully_occluded_or_out_of_view_images": 0,
26
+ "generation_seconds": 182.029
27
+ },
28
+ "checks": [
29
+ {
30
+ "id": "format",
31
+ "label": "数据结构与视频同步",
32
+ "passed": true
33
+ },
34
+ {
35
+ "id": "coverage",
36
+ "label": "实际配置与覆盖计划一致",
37
+ "passed": true
38
+ },
39
+ {
40
+ "id": "split_isolation",
41
+ "label": "完整场景对与种子隔离",
42
+ "passed": true,
43
+ "detail": ""
44
+ },
45
+ {
46
+ "id": "labels",
47
+ "label": "图像及几何标签可读且一致",
48
+ "passed": true
49
+ },
50
+ {
51
+ "id": "sensor_observations",
52
+ "label": "观测与真值分离、文件一致性及观测条件对核验",
53
+ "passed": true
54
+ },
55
+ {
56
+ "id": "observed_faults",
57
+ "label": "指定故障已实际观察到",
58
+ "passed": true
59
+ },
60
+ {
61
+ "id": "verified_recovery",
62
+ "label": "至少包含一条物理验证的恢复正例",
63
+ "passed": true
64
+ }
65
+ ],
66
+ "quality": {
67
+ "episodes": 24,
68
+ "frames": 3360,
69
+ "valid_pairs": 22,
70
+ "recovery_success_rate": 0.9166666666666666,
71
+ "failure_coverage": {
72
+ "slip": 6,
73
+ "drop": 6,
74
+ "pose_offset": 6,
75
+ "miss_grasp": 6
76
+ },
77
+ "scenario_coverage": {
78
+ "machine_tending": 12,
79
+ "conveyor_pick": 6,
80
+ "pallet_cell_placement": 6
81
+ },
82
+ "object_shape_coverage": {
83
+ "box": 18,
84
+ "cylinder": 6
85
+ },
86
+ "rejected_pairs": 2,
87
+ "structurally_valid": true,
88
+ "structural_errors": [],
89
+ "physics_validated": true,
90
+ "sensor_model_calibrated": false,
91
+ "real_robot_transfer_tested": false,
92
+ "grasp_candidates": 0,
93
+ "validated_successful_grasps": 0,
94
+ "grasp_physics_validated": false,
95
+ "label_semantics": "depth_gt is geometric camera-z in meters; not measured sensor depth",
96
+ "acceptance": "valid pairs require observed failure, executed recovery, final-state success and physics validation"
97
+ },
98
+ "splits": {
99
+ "method": "paired_scene_group_holdout",
100
+ "coverage_split": "matched_conditions",
101
+ "condition_coverage": {
102
+ "train": [
103
+ "camera_interrupt.challenge",
104
+ "camera_interrupt.reference",
105
+ "compound_shift.challenge",
106
+ "compound_shift.reference",
107
+ "range_noise.challenge",
108
+ "range_noise.reference",
109
+ "target_missing.challenge",
110
+ "target_missing.reference"
111
+ ],
112
+ "validation": [
113
+ "camera_interrupt.challenge",
114
+ "camera_interrupt.reference",
115
+ "compound_shift.challenge",
116
+ "compound_shift.reference",
117
+ "range_noise.challenge",
118
+ "range_noise.reference",
119
+ "target_missing.challenge",
120
+ "target_missing.reference"
121
+ ],
122
+ "test": [
123
+ "camera_interrupt.challenge",
124
+ "camera_interrupt.reference",
125
+ "compound_shift.challenge",
126
+ "compound_shift.reference",
127
+ "range_noise.challenge",
128
+ "range_noise.reference",
129
+ "target_missing.challenge",
130
+ "target_missing.reference"
131
+ ]
132
+ },
133
+ "independent_scene_pairs": {
134
+ "train": 4,
135
+ "validation": 4,
136
+ "test": 4
137
+ },
138
+ "episode_indices": {
139
+ "train": [
140
+ 0,
141
+ 1,
142
+ 2,
143
+ 3,
144
+ 4,
145
+ 5,
146
+ 6,
147
+ 7
148
+ ],
149
+ "validation": [
150
+ 8,
151
+ 9,
152
+ 10,
153
+ 11,
154
+ 12,
155
+ 13,
156
+ 14,
157
+ 15
158
+ ],
159
+ "test": [
160
+ 16,
161
+ 17,
162
+ 18,
163
+ 19,
164
+ 20,
165
+ 21,
166
+ 22,
167
+ 23
168
+ ]
169
+ },
170
+ "ranges": {
171
+ "train": "0:8",
172
+ "validation": "8:16",
173
+ "test": "16:24"
174
+ },
175
+ "pair_assignments": {
176
+ "sensor_failure_benchmark.range_noise.train.r000": "train",
177
+ "sensor_failure_benchmark.target_missing.train.r000": "train",
178
+ "sensor_failure_benchmark.camera_interrupt.train.r000": "train",
179
+ "sensor_failure_benchmark.compound_shift.train.r000": "train",
180
+ "sensor_failure_benchmark.range_noise.validation.r000": "validation",
181
+ "sensor_failure_benchmark.target_missing.validation.r000": "validation",
182
+ "sensor_failure_benchmark.camera_interrupt.validation.r000": "validation",
183
+ "sensor_failure_benchmark.compound_shift.validation.r000": "validation",
184
+ "sensor_failure_benchmark.range_noise.test.r000": "test",
185
+ "sensor_failure_benchmark.target_missing.test.r000": "test",
186
+ "sensor_failure_benchmark.camera_interrupt.test.r000": "test",
187
+ "sensor_failure_benchmark.compound_shift.test.r000": "test"
188
+ },
189
+ "seed_assignments": {
190
+ "20261024": "train",
191
+ "20262024": "train",
192
+ "20263024": "train",
193
+ "20264024": "train",
194
+ "20261124": "validation",
195
+ "20262124": "validation",
196
+ "20263124": "validation",
197
+ "20264124": "validation",
198
+ "20261224": "test",
199
+ "20262224": "test",
200
+ "20263224": "test",
201
+ "20264224": "test"
202
+ },
203
+ "group_isolation_verified": true,
204
+ "scope": "complete-episode split; paired scenes and scene seeds remain isolated when a holdout is used"
205
+ },
206
+ "scope": "technical evidence only; customer model performance and real deployment are not evaluated"
207
+ }
commercial/buyer_brief.md ADDED
@@ -0,0 +1,19 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # 机器人 RGB-D 失效对照包 · 客户评估简报
2
+
3
+ 适用团队:机器人视觉团队、深度补全团队与工位验收负责人
4
+
5
+ 要解决的任务:现场深度缺测和短时相机中断难以重复,模型升级缺少一致的压力评测。
6
+
7
+ 预期价值(待客户验证):保留相同状态、动作与几何真值,交付配对观测、有效像素及实际退化统计。
8
+
9
+ ## 客户输入
10
+
11
+ > 流程验证小样:四种条件各自在训练、验证、测试中具备独立种子的场景对;用于分开检查同条件泛化与未知条件压力评测。非真实客户数据。
12
+
13
+ ## 试点路径
14
+
15
+ 先用本样品确认字段、加载和任务匹配,再用固定真实测试集做基线/增量评估;双方确认条件覆盖、验收阈值和用途后报价,只有获得试点评估信号才扩量。
16
+
17
+ 建议客户补充:相机内参/分辨率、目标物尺寸与材质、实际失败案例、机器人动作定义、模型输入格式、真实测试集和可接受指标。
18
+
19
+ 当前技术检查7/7通过,客户效果验证尚未开始。报价假设见quote_assumptions.json;该文件不表示已成交价格或收益承诺。
commercial/coverage.csv ADDED
@@ -0,0 +1,25 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ case_id,pair_id,split,role,scenario_id,object_shape,material,lighting,background,changed_parameter,seed,expected_episodes,observed_episodes,accepted_recovery_pairs,failure_observed,mean_target_image_fraction,matches_plan,issues
2
+ sensor_failure_benchmark.range_noise.train.r000.reference,sensor_failure_benchmark.range_noise.train.r000,train,reference,machine_tending,box,opaque,studio,studio,sensor_profile,20261024,1,1,1,1,0.0042567,True,
3
+ sensor_failure_benchmark.range_noise.train.r000.challenge,sensor_failure_benchmark.range_noise.train.r000,train,challenge,machine_tending,box,opaque,studio,studio,sensor_profile,20261024,1,1,1,1,0.0042567,True,
4
+ sensor_failure_benchmark.target_missing.train.r000.reference,sensor_failure_benchmark.target_missing.train.r000,train,reference,machine_tending,box,opaque,studio,studio,sensor_profile,20262024,1,1,1,1,0.00427083,True,
5
+ sensor_failure_benchmark.target_missing.train.r000.challenge,sensor_failure_benchmark.target_missing.train.r000,train,challenge,machine_tending,box,opaque,studio,studio,sensor_profile,20262024,1,1,1,1,0.00427083,True,
6
+ sensor_failure_benchmark.camera_interrupt.train.r000.reference,sensor_failure_benchmark.camera_interrupt.train.r000,train,reference,conveyor_pick,cylinder,opaque,studio,studio,sensor_profile,20263024,1,1,1,1,0.00470499,True,
7
+ sensor_failure_benchmark.camera_interrupt.train.r000.challenge,sensor_failure_benchmark.camera_interrupt.train.r000,train,challenge,conveyor_pick,cylinder,opaque,studio,studio,sensor_profile,20263024,1,1,1,1,0.00470499,True,
8
+ sensor_failure_benchmark.compound_shift.train.r000.reference,sensor_failure_benchmark.compound_shift.train.r000,train,reference,pallet_cell_placement,box,opaque,studio,studio,sensor_profile,20264024,1,1,1,1,0.00417039,True,
9
+ sensor_failure_benchmark.compound_shift.train.r000.challenge,sensor_failure_benchmark.compound_shift.train.r000,train,challenge,pallet_cell_placement,box,opaque,studio,studio,sensor_profile,20264024,1,1,1,1,0.00417039,True,
10
+ sensor_failure_benchmark.range_noise.validation.r000.reference,sensor_failure_benchmark.range_noise.validation.r000,validation,reference,machine_tending,box,opaque,studio,studio,sensor_profile,20261124,1,1,1,1,0.0041875,True,
11
+ sensor_failure_benchmark.range_noise.validation.r000.challenge,sensor_failure_benchmark.range_noise.validation.r000,validation,challenge,machine_tending,box,opaque,studio,studio,sensor_profile,20261124,1,1,1,1,0.0041875,True,
12
+ sensor_failure_benchmark.target_missing.validation.r000.reference,sensor_failure_benchmark.target_missing.validation.r000,validation,reference,machine_tending,box,opaque,studio,studio,sensor_profile,20262124,1,1,1,1,0.00429315,True,
13
+ sensor_failure_benchmark.target_missing.validation.r000.challenge,sensor_failure_benchmark.target_missing.validation.r000,validation,challenge,machine_tending,box,opaque,studio,studio,sensor_profile,20262124,1,1,1,1,0.00429315,True,
14
+ sensor_failure_benchmark.camera_interrupt.validation.r000.reference,sensor_failure_benchmark.camera_interrupt.validation.r000,validation,reference,conveyor_pick,cylinder,opaque,studio,studio,sensor_profile,20263124,1,1,1,1,0.00466964,True,
15
+ sensor_failure_benchmark.camera_interrupt.validation.r000.challenge,sensor_failure_benchmark.camera_interrupt.validation.r000,validation,challenge,conveyor_pick,cylinder,opaque,studio,studio,sensor_profile,20263124,1,1,1,1,0.00466964,True,
16
+ sensor_failure_benchmark.compound_shift.validation.r000.reference,sensor_failure_benchmark.compound_shift.validation.r000,validation,reference,pallet_cell_placement,box,opaque,studio,studio,sensor_profile,20264124,1,1,0,1,0.00379687,True,
17
+ sensor_failure_benchmark.compound_shift.validation.r000.challenge,sensor_failure_benchmark.compound_shift.validation.r000,validation,challenge,pallet_cell_placement,box,opaque,studio,studio,sensor_profile,20264124,1,1,0,1,0.00379687,True,
18
+ sensor_failure_benchmark.range_noise.test.r000.reference,sensor_failure_benchmark.range_noise.test.r000,test,reference,machine_tending,box,opaque,studio,studio,sensor_profile,20261224,1,1,1,1,0.00429353,True,
19
+ sensor_failure_benchmark.range_noise.test.r000.challenge,sensor_failure_benchmark.range_noise.test.r000,test,challenge,machine_tending,box,opaque,studio,studio,sensor_profile,20261224,1,1,1,1,0.00429353,True,
20
+ sensor_failure_benchmark.target_missing.test.r000.reference,sensor_failure_benchmark.target_missing.test.r000,test,reference,machine_tending,box,opaque,studio,studio,sensor_profile,20262224,1,1,1,1,0.00421949,True,
21
+ sensor_failure_benchmark.target_missing.test.r000.challenge,sensor_failure_benchmark.target_missing.test.r000,test,challenge,machine_tending,box,opaque,studio,studio,sensor_profile,20262224,1,1,1,1,0.00421949,True,
22
+ sensor_failure_benchmark.camera_interrupt.test.r000.reference,sensor_failure_benchmark.camera_interrupt.test.r000,test,reference,conveyor_pick,cylinder,opaque,studio,studio,sensor_profile,20263224,1,1,1,1,0.00475186,True,
23
+ sensor_failure_benchmark.camera_interrupt.test.r000.challenge,sensor_failure_benchmark.camera_interrupt.test.r000,test,challenge,conveyor_pick,cylinder,opaque,studio,studio,sensor_profile,20263224,1,1,1,1,0.00475186,True,
24
+ sensor_failure_benchmark.compound_shift.test.r000.reference,sensor_failure_benchmark.compound_shift.test.r000,test,reference,pallet_cell_placement,box,opaque,studio,studio,sensor_profile,20264224,1,1,1,1,0.00424033,True,
25
+ sensor_failure_benchmark.compound_shift.test.r000.challenge,sensor_failure_benchmark.compound_shift.test.r000,test,challenge,pallet_cell_placement,box,opaque,studio,studio,sensor_profile,20264224,1,1,1,1,0.00424033,True,
commercial/dataset_card.md ADDED
@@ -0,0 +1,33 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # 机器人 RGB-D 失效对照包 · 数据说明书
2
+
3
+ 技术验收:technical_pass。客户实际模型效果与真机迁移:**未验证**。
4
+
5
+ 本包包含24条完整episode、3360个时间步、3360张RGB;覆盖24个计划条件,物理验证恢复正例22条。
6
+
7
+ ## 使用与评估
8
+
9
+ 1. 根据coverage.csv核对条件与实际交付,根据sample_index.csv定位图像与标签。
10
+ 2. 使用dataset/meta/splits.json选择完整episode。场景对和种子不跨集合;同类程序化形状可能跨集合。
11
+ 3. 仅用train拟合,validation调参,锁定test评估;另用客户真实数据测迁移。
12
+ 导出的归一化统计仅由train计算,记录见dataset/meta/normalization.json。
13
+ 4. 在evaluation_template.csv填写基线与加入本包后的结果,记录模型版本、真实样本数与验收阈值。
14
+
15
+ 参考/挑战样本是可控场景对,不代表已完成因果实验。默认小样用于接口与覆盖评估,不能据此宣称训练规模或泛化效果。
16
+
17
+ 三个划分包含相同的条件目录,每个划分使用独立场景种子,完整对照/挑战对保持同组。这用于同条件的种子留出评估,不证明未知物体、工位或退化类型泛化。
18
+
19
+ ## 标签与适用边界
20
+
21
+ 深度为米制首表面几何真值,非特定传感器实测;法线坐标系见episode。目标掩码像素占整幅图的比例不等于物体可见面比例;完全遮挡/出视野样本作为显式负例保留。感知线action为占位,恢复线为简化三轴夹爪动作。
22
+
23
+ v0.5的sensors/index.jsonl逐条区分受损观测输入与干净真值标签;sensors/report.json报告缺测和误差,并核对观测挑战对的物理轨迹一致性。观测退化没有驱动策略重新执行,不能将原轨迹成功率用作受损观测下的闭环效果。dataset/meta/input_contract.json标明特权真值与建议训练输入。
24
+
25
+ - 默认8条是结构评估样品,不能代表训练规模或真实工况覆盖率。
26
+ - 匹配场景对用于条件比较,不宣称严格物理因果实验。
27
+ - 观测退化为未标定的压力模型,不等同某品牌深度相机。
28
+ - 控制器使用仿真真值;配对行为一致不代表受损观测下闭环策略仍然成功。
29
+ - 训练输入优先使用 observed 图像与 proprioception,物体真值仅作监督或诊断。
30
+
31
+ ## 来源和使用约定
32
+
33
+ 场景和物体由本项目程序化构建,来源见provenance.json;本说明书不授予独占权或替代双方约定的使用范围。交付前应明确用途、分发范围、客户输入资产来源与支持责任。
commercial/evaluation_template.csv ADDED
@@ -0,0 +1,25 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ case_id,split,metric,baseline_value,augmented_value,real_test_sample_count,customer_acceptance_threshold,notes
2
+ sensor_failure_benchmark.range_noise.train.r000.reference,train,real_task_success_rate,,,,,客户在固定真实测试集上填写;仅训练集可用于拟合
3
+ sensor_failure_benchmark.range_noise.train.r000.challenge,train,real_task_success_rate,,,,,客户在固定真实测试集上填写;仅训练集可用于拟合
4
+ sensor_failure_benchmark.target_missing.train.r000.reference,train,real_task_success_rate,,,,,客户在固定真实测试集上填写;仅训练集可用于拟合
5
+ sensor_failure_benchmark.target_missing.train.r000.challenge,train,real_task_success_rate,,,,,客户在固定真实测试集上填写;仅训练集可用于拟合
6
+ sensor_failure_benchmark.camera_interrupt.train.r000.reference,train,real_task_success_rate,,,,,客户在固定真实测试集上填写;仅训练集可用于拟合
7
+ sensor_failure_benchmark.camera_interrupt.train.r000.challenge,train,real_task_success_rate,,,,,客户在固定真实测试集上填写;仅训练集可用于拟合
8
+ sensor_failure_benchmark.compound_shift.train.r000.reference,train,real_task_success_rate,,,,,客户在固定真实测试集上填写;仅训练集可用于拟合
9
+ sensor_failure_benchmark.compound_shift.train.r000.challenge,train,real_task_success_rate,,,,,客户在固定真实测试集上填写;仅训练集可用于拟合
10
+ sensor_failure_benchmark.range_noise.validation.r000.reference,validation,real_task_success_rate,,,,,客户在固定真实测试集上填写;仅训练集可用于拟合
11
+ sensor_failure_benchmark.range_noise.validation.r000.challenge,validation,real_task_success_rate,,,,,客户在固定真实测试集上填写;仅训练集可用于拟合
12
+ sensor_failure_benchmark.target_missing.validation.r000.reference,validation,real_task_success_rate,,,,,客户在固定真实测试集上填写;仅训练集可用于拟合
13
+ sensor_failure_benchmark.target_missing.validation.r000.challenge,validation,real_task_success_rate,,,,,客户在固定真实测试集上填写;仅训练集可用于拟合
14
+ sensor_failure_benchmark.camera_interrupt.validation.r000.reference,validation,real_task_success_rate,,,,,客户在固定真实测试集上填写;仅训练集可用于拟合
15
+ sensor_failure_benchmark.camera_interrupt.validation.r000.challenge,validation,real_task_success_rate,,,,,客户在固定真实测试集上填写;仅训练集可用于拟合
16
+ sensor_failure_benchmark.compound_shift.validation.r000.reference,validation,real_task_success_rate,,,,,客户在固定真实测试集上填写;仅训练集可用于拟合
17
+ sensor_failure_benchmark.compound_shift.validation.r000.challenge,validation,real_task_success_rate,,,,,客户在固定真实测试集上填写;仅训练集可用于拟合
18
+ sensor_failure_benchmark.range_noise.test.r000.reference,test,real_task_success_rate,,,,,客户在固定真实测试集上填写;仅训练集可用于拟合
19
+ sensor_failure_benchmark.range_noise.test.r000.challenge,test,real_task_success_rate,,,,,客户在固定真实测试集上填写;仅训练集可用于拟合
20
+ sensor_failure_benchmark.target_missing.test.r000.reference,test,real_task_success_rate,,,,,客户在固定真实测试集上填写;仅训练集可用于拟合
21
+ sensor_failure_benchmark.target_missing.test.r000.challenge,test,real_task_success_rate,,,,,客户在固定真实测试集上填写;仅训练集可用于拟合
22
+ sensor_failure_benchmark.camera_interrupt.test.r000.reference,test,real_task_success_rate,,,,,客户在固定真实测试集上填写;仅训练集可用于拟合
23
+ sensor_failure_benchmark.camera_interrupt.test.r000.challenge,test,real_task_success_rate,,,,,客户在固定真实测试集上填写;仅训练集可用于拟合
24
+ sensor_failure_benchmark.compound_shift.test.r000.reference,test,real_task_success_rate,,,,,客户在固定真实测试集上填写;仅训练集可用于拟合
25
+ sensor_failure_benchmark.compound_shift.test.r000.challenge,test,real_task_success_rate,,,,,客户在固定真实测试集上填写;仅训练集可用于拟合
commercial/pair_comparison.csv ADDED
@@ -0,0 +1,13 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ pair_id,split,changed_parameter,reference_value,reference_accepted_recoveries,reference_target_image_fraction,challenge_value,challenge_accepted_recoveries,challenge_target_image_fraction,customer_model_comparison
2
+ sensor_failure_benchmark.range_noise.train.r000,train,sensor_profile,clean,1,0.0042567,depth_noise,1,0.0042567,not_run
3
+ sensor_failure_benchmark.target_missing.train.r000,train,sensor_profile,clean,1,0.00427083,target_dropout,1,0.00427083,not_run
4
+ sensor_failure_benchmark.camera_interrupt.train.r000,train,sensor_profile,clean,1,0.00470499,camera_outage,1,0.00470499,not_run
5
+ sensor_failure_benchmark.compound_shift.train.r000,train,sensor_profile,clean,1,0.00417039,combined,1,0.00417039,not_run
6
+ sensor_failure_benchmark.range_noise.validation.r000,validation,sensor_profile,clean,1,0.0041875,depth_noise,1,0.0041875,not_run
7
+ sensor_failure_benchmark.target_missing.validation.r000,validation,sensor_profile,clean,1,0.00429315,target_dropout,1,0.00429315,not_run
8
+ sensor_failure_benchmark.camera_interrupt.validation.r000,validation,sensor_profile,clean,1,0.00466964,camera_outage,1,0.00466964,not_run
9
+ sensor_failure_benchmark.compound_shift.validation.r000,validation,sensor_profile,clean,0,0.00379687,combined,0,0.00379687,not_run
10
+ sensor_failure_benchmark.range_noise.test.r000,test,sensor_profile,clean,1,0.00429353,depth_noise,1,0.00429353,not_run
11
+ sensor_failure_benchmark.target_missing.test.r000,test,sensor_profile,clean,1,0.00421949,target_dropout,1,0.00421949,not_run
12
+ sensor_failure_benchmark.camera_interrupt.test.r000,test,sensor_profile,clean,1,0.00475186,camera_outage,1,0.00475186,not_run
13
+ sensor_failure_benchmark.compound_shift.test.r000,test,sensor_profile,clean,1,0.00424033,combined,1,0.00424033,not_run
commercial/quote_assumptions.json ADDED
@@ -0,0 +1,21 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "status": "unvalidated_hypothesis",
3
+ "currency": "CNY",
4
+ "product_price_hypothesis": {
5
+ "currency": "CNY",
6
+ "min": 1500,
7
+ "max": 4500,
8
+ "unit": "范围确认后的付费试点",
9
+ "status": "unvalidated",
10
+ "note": "仅用于客户访谈的报价假设,不是市场成交价、自动报价或当前小样售价。"
11
+ },
12
+ "measured_generation_seconds": 182.029,
13
+ "delivered_episodes": 24,
14
+ "delivered_rgb_images": 3360,
15
+ "compute_cost_per_hour": null,
16
+ "human_review_hours": null,
17
+ "customer_integration_hours": null,
18
+ "commercial_acceptance_agreed": false,
19
+ "pricing_basis": "范围、覆盖条件、可用交付和适配工时;原始帧数不等同独立样本数",
20
+ "next_step": "确认客户输入规格和真实评估集后,填写成本、报价与验收条款"
21
+ }
commercial/sample_index.csv ADDED
The diff for this file is too large to render. See raw diff
 
config.json ADDED
@@ -0,0 +1,34 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "kind": "recovery",
3
+ "backend": "mujoco",
4
+ "name": "RGB-D 失效对照 · 同条件独立种子留出",
5
+ "product_id": "sensor_failure_benchmark",
6
+ "coverage_repeats": 1,
7
+ "coverage_split": "matched_conditions",
8
+ "customer_brief": "流程验证小样:四种条件各自在训练、验证、测试中具备独立种子的场景对;用于分开检查同条件泛化与未知条件压力评测。非真实客户数据。",
9
+ "scenario_id": "machine_tending",
10
+ "object_shape": "box",
11
+ "material": "opaque",
12
+ "background": "studio",
13
+ "lighting": "studio",
14
+ "difficulty": "standard",
15
+ "object_scale": 1.0,
16
+ "object_mass_kg": 0.1,
17
+ "surface_friction": 1.1,
18
+ "camera_jitter": 0.0,
19
+ "sensor_profile": "clean",
20
+ "sensor_severity": 0.5,
21
+ "empirical_sensor_profile": null,
22
+ "episodes": 24,
23
+ "split_strategy": "single_train",
24
+ "frames_per_episode": 140,
25
+ "fps": 20,
26
+ "width": 160,
27
+ "height": 120,
28
+ "seed": 20261024,
29
+ "camera_count": 1,
30
+ "failure_types": [
31
+ "slip"
32
+ ],
33
+ "blender_samples": 16
34
+ }
dataset/README.md ADDED
@@ -0,0 +1,7 @@
 
 
 
 
 
 
 
 
1
+ # SimForge generated corpus
2
+
3
+ LeRobotDataset v3 Parquet/MP4 layout. Lossless RGB and geometric labels remain in the source episodes.
4
+
5
+ Failed recoveries are labelled negatives, not successful demonstrations. Consult meta/accepted_pairs.json and meta/quality.json.
6
+
7
+ Sensor response and real robot transfer are not validated. RGB video is compressed; use source PNG/NPY for precision labels.
dataset/data/chunk-000/file-000.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dd5067122b429fb7977ae324d41843728942b439858073f519eb4a146029a0dc
3
+ size 342373
dataset/meta/accepted_pairs.json ADDED
@@ -0,0 +1,24 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ [
2
+ 0,
3
+ 1,
4
+ 2,
5
+ 3,
6
+ 4,
7
+ 5,
8
+ 6,
9
+ 7,
10
+ 8,
11
+ 9,
12
+ 10,
13
+ 11,
14
+ 12,
15
+ 13,
16
+ 16,
17
+ 17,
18
+ 18,
19
+ 19,
20
+ 20,
21
+ 21,
22
+ 22,
23
+ 23
24
+ ]
dataset/meta/episodes/chunk-000/file-000.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:116882a6c8eda0f84c96cdb379881f80fb0085617c11049915a5e4910159f196
3
+ size 5258
dataset/meta/info.json ADDED
@@ -0,0 +1,187 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "codebase_version": "v3.0",
3
+ "robot_type": "cartesian_parallel_gripper",
4
+ "total_episodes": 24,
5
+ "total_frames": 3360,
6
+ "total_tasks": 3,
7
+ "total_videos": 1,
8
+ "total_chunks": 1,
9
+ "chunks_size": 1000,
10
+ "fps": 20,
11
+ "splits": {
12
+ "train": "0:8",
13
+ "validation": "8:16",
14
+ "test": "16:24"
15
+ },
16
+ "data_path": "data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet",
17
+ "video_path": "videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4",
18
+ "features": {
19
+ "observation.state": {
20
+ "dtype": "float32",
21
+ "shape": [
22
+ 10
23
+ ],
24
+ "names": null
25
+ },
26
+ "action": {
27
+ "dtype": "float32",
28
+ "shape": [
29
+ 5
30
+ ],
31
+ "names": null
32
+ },
33
+ "supervision.object_pose": {
34
+ "dtype": "float32",
35
+ "shape": [
36
+ 7
37
+ ],
38
+ "names": null
39
+ },
40
+ "timestamp": {
41
+ "dtype": "float32",
42
+ "shape": [
43
+ 1
44
+ ],
45
+ "names": null
46
+ },
47
+ "frame_index": {
48
+ "dtype": "int64",
49
+ "shape": [
50
+ 1
51
+ ],
52
+ "names": null
53
+ },
54
+ "episode_index": {
55
+ "dtype": "int64",
56
+ "shape": [
57
+ 1
58
+ ],
59
+ "names": null
60
+ },
61
+ "index": {
62
+ "dtype": "int64",
63
+ "shape": [
64
+ 1
65
+ ],
66
+ "names": null
67
+ },
68
+ "task_index": {
69
+ "dtype": "int64",
70
+ "shape": [
71
+ 1
72
+ ],
73
+ "names": null
74
+ },
75
+ "annotation.failure": {
76
+ "dtype": "bool",
77
+ "shape": [
78
+ 1
79
+ ],
80
+ "names": null
81
+ },
82
+ "annotation.success": {
83
+ "dtype": "bool",
84
+ "shape": [
85
+ 1
86
+ ],
87
+ "names": null
88
+ },
89
+ "annotation.phase": {
90
+ "dtype": "string",
91
+ "shape": [
92
+ 1
93
+ ],
94
+ "names": null
95
+ },
96
+ "annotation.failure_type": {
97
+ "dtype": "string",
98
+ "shape": [
99
+ 1
100
+ ],
101
+ "names": null
102
+ },
103
+ "annotation.scenario_id": {
104
+ "dtype": "string",
105
+ "shape": [
106
+ 1
107
+ ],
108
+ "names": null
109
+ },
110
+ "annotation.object_shape": {
111
+ "dtype": "string",
112
+ "shape": [
113
+ 1
114
+ ],
115
+ "names": null
116
+ },
117
+ "annotation.material": {
118
+ "dtype": "string",
119
+ "shape": [
120
+ 1
121
+ ],
122
+ "names": null
123
+ },
124
+ "annotation.valid_recovery_pair": {
125
+ "dtype": "bool",
126
+ "shape": [
127
+ 1
128
+ ],
129
+ "names": null
130
+ },
131
+ "annotation.split": {
132
+ "dtype": "string",
133
+ "shape": [
134
+ 1
135
+ ],
136
+ "names": null
137
+ },
138
+ "annotation.case_id": {
139
+ "dtype": "string",
140
+ "shape": [
141
+ 1
142
+ ],
143
+ "names": null
144
+ },
145
+ "annotation.pair_id": {
146
+ "dtype": "string",
147
+ "shape": [
148
+ 1
149
+ ],
150
+ "names": null
151
+ },
152
+ "annotation.variant_role": {
153
+ "dtype": "string",
154
+ "shape": [
155
+ 1
156
+ ],
157
+ "names": null
158
+ },
159
+ "privileged.simulator_state": {
160
+ "dtype": "float32",
161
+ "shape": [
162
+ 23
163
+ ],
164
+ "names": null
165
+ },
166
+ "observation.images.front_observed": {
167
+ "dtype": "video",
168
+ "shape": [
169
+ 120,
170
+ 160,
171
+ 3
172
+ ],
173
+ "names": [
174
+ "height",
175
+ "width",
176
+ "channels"
177
+ ],
178
+ "info": {
179
+ "video.fps": 20,
180
+ "video.codec": "h264",
181
+ "video.pix_fmt": "yuv420p",
182
+ "video.is_depth_map": false,
183
+ "has_audio": false
184
+ }
185
+ }
186
+ }
187
+ }
dataset/meta/input_contract.json ADDED
@@ -0,0 +1,19 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "recommended_policy_inputs": [
3
+ "observation.state",
4
+ "observation.images.front_observed"
5
+ ],
6
+ "privileged_simulator_fields": [
7
+ "privileged.simulator_state",
8
+ "supervision.object_pose"
9
+ ],
10
+ "label_only_fields": [
11
+ "annotation.*"
12
+ ],
13
+ "proprioception_layout": "qpos[0:5] + qvel[0:5]; Cartesian stages and fingers only",
14
+ "path_base": "job_root",
15
+ "observed_depth_index": "sensors/index.jsonl",
16
+ "sensor_calibrated": false,
17
+ "closed_loop_policy_evaluated": false,
18
+ "note": "Policy state contains only Cartesian gripper proprioception. Clean RGB remains outside auto-discovered video features. Commands are generated by an oracle controller; observation degradation is offline. Legacy v0.4 observation.state contained object truth; regenerate for safe defaults."
19
+ }
dataset/meta/normalization.json ADDED
@@ -0,0 +1,16 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "fit_split": "train",
3
+ "episode_indices": [
4
+ 0,
5
+ 1,
6
+ 2,
7
+ 3,
8
+ 4,
9
+ 5,
10
+ 6,
11
+ 7
12
+ ],
13
+ "frame_count": 1120,
14
+ "rgb_source": "lossless source PNG",
15
+ "held_out_observations_used": false
16
+ }
dataset/meta/quality.json ADDED
@@ -0,0 +1,32 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "episodes": 24,
3
+ "frames": 3360,
4
+ "valid_pairs": 22,
5
+ "recovery_success_rate": 0.9166666666666666,
6
+ "failure_coverage": {
7
+ "slip": 6,
8
+ "drop": 6,
9
+ "pose_offset": 6,
10
+ "miss_grasp": 6
11
+ },
12
+ "scenario_coverage": {
13
+ "machine_tending": 12,
14
+ "conveyor_pick": 6,
15
+ "pallet_cell_placement": 6
16
+ },
17
+ "object_shape_coverage": {
18
+ "box": 18,
19
+ "cylinder": 6
20
+ },
21
+ "rejected_pairs": 2,
22
+ "structurally_valid": true,
23
+ "structural_errors": [],
24
+ "physics_validated": true,
25
+ "sensor_model_calibrated": false,
26
+ "real_robot_transfer_tested": false,
27
+ "grasp_candidates": 0,
28
+ "validated_successful_grasps": 0,
29
+ "grasp_physics_validated": false,
30
+ "label_semantics": "depth_gt is geometric camera-z in meters; not measured sensor depth",
31
+ "acceptance": "valid pairs require observed failure, executed recovery, final-state success and physics validation"
32
+ }
dataset/meta/scenarios.json ADDED
@@ -0,0 +1,1562 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ [
2
+ {
3
+ "episode_index": 0,
4
+ "scenario_id": "machine_tending",
5
+ "object_shape": "box",
6
+ "geometry": {
7
+ "shape": "box",
8
+ "half_extents_m": [
9
+ 0.022,
10
+ 0.022,
11
+ 0.035
12
+ ],
13
+ "radius_m": null,
14
+ "height_m": 0.07,
15
+ "mass_kg": 0.1,
16
+ "collision_shape": "box",
17
+ "contact_dimension": 4,
18
+ "rolling_friction": null,
19
+ "capsule_long_axis": null
20
+ },
21
+ "scene_parameters": null,
22
+ "commercial_case": {
23
+ "episode_index": 0,
24
+ "case_id": "sensor_failure_benchmark.range_noise.train.r000.reference",
25
+ "condition_id": "range_noise.reference",
26
+ "condition_group_id": "range_noise",
27
+ "coverage_split": "matched_conditions",
28
+ "pair_id": "sensor_failure_benchmark.range_noise.train.r000",
29
+ "split": "train",
30
+ "role": "reference",
31
+ "product_id": "sensor_failure_benchmark",
32
+ "changed_parameter": "sensor_profile"
33
+ },
34
+ "resolved_config": {
35
+ "kind": "recovery",
36
+ "backend": "mujoco",
37
+ "name": "RGB-D 失效对照 · 同条件独立种子留出",
38
+ "coverage_split": "condition_holdout",
39
+ "scenario_id": "machine_tending",
40
+ "object_shape": "box",
41
+ "material": "opaque",
42
+ "background": "studio",
43
+ "lighting": "studio",
44
+ "difficulty": "standard",
45
+ "object_scale": 1.0,
46
+ "object_mass_kg": 0.1,
47
+ "surface_friction": 1.1,
48
+ "camera_jitter": 0.0,
49
+ "sensor_profile": "clean",
50
+ "sensor_severity": 0.5,
51
+ "empirical_sensor_profile": null,
52
+ "episodes": 1,
53
+ "split_strategy": "single_train",
54
+ "frames_per_episode": 140,
55
+ "fps": 20,
56
+ "width": 160,
57
+ "height": 120,
58
+ "seed": 20261024,
59
+ "camera_count": 1,
60
+ "failure_types": [
61
+ "slip"
62
+ ],
63
+ "blender_samples": 16
64
+ },
65
+ "task_instruction": "Recover the box and place it into the machine fixture nest."
66
+ },
67
+ {
68
+ "episode_index": 1,
69
+ "scenario_id": "machine_tending",
70
+ "object_shape": "box",
71
+ "geometry": {
72
+ "shape": "box",
73
+ "half_extents_m": [
74
+ 0.022,
75
+ 0.022,
76
+ 0.035
77
+ ],
78
+ "radius_m": null,
79
+ "height_m": 0.07,
80
+ "mass_kg": 0.1,
81
+ "collision_shape": "box",
82
+ "contact_dimension": 4,
83
+ "rolling_friction": null,
84
+ "capsule_long_axis": null
85
+ },
86
+ "scene_parameters": null,
87
+ "commercial_case": {
88
+ "episode_index": 1,
89
+ "case_id": "sensor_failure_benchmark.range_noise.train.r000.challenge",
90
+ "condition_id": "range_noise.challenge",
91
+ "condition_group_id": "range_noise",
92
+ "coverage_split": "matched_conditions",
93
+ "pair_id": "sensor_failure_benchmark.range_noise.train.r000",
94
+ "split": "train",
95
+ "role": "challenge",
96
+ "product_id": "sensor_failure_benchmark",
97
+ "changed_parameter": "sensor_profile"
98
+ },
99
+ "resolved_config": {
100
+ "kind": "recovery",
101
+ "backend": "mujoco",
102
+ "name": "RGB-D 失效对照 · 同条件独立种子留出",
103
+ "coverage_split": "condition_holdout",
104
+ "scenario_id": "machine_tending",
105
+ "object_shape": "box",
106
+ "material": "opaque",
107
+ "background": "studio",
108
+ "lighting": "studio",
109
+ "difficulty": "standard",
110
+ "object_scale": 1.0,
111
+ "object_mass_kg": 0.1,
112
+ "surface_friction": 1.1,
113
+ "camera_jitter": 0.0,
114
+ "sensor_profile": "depth_noise",
115
+ "sensor_severity": 0.5,
116
+ "empirical_sensor_profile": null,
117
+ "episodes": 1,
118
+ "split_strategy": "single_train",
119
+ "frames_per_episode": 140,
120
+ "fps": 20,
121
+ "width": 160,
122
+ "height": 120,
123
+ "seed": 20261024,
124
+ "camera_count": 1,
125
+ "failure_types": [
126
+ "slip"
127
+ ],
128
+ "blender_samples": 16
129
+ },
130
+ "task_instruction": "Recover the box and place it into the machine fixture nest."
131
+ },
132
+ {
133
+ "episode_index": 2,
134
+ "scenario_id": "machine_tending",
135
+ "object_shape": "box",
136
+ "geometry": {
137
+ "shape": "box",
138
+ "half_extents_m": [
139
+ 0.022,
140
+ 0.022,
141
+ 0.035
142
+ ],
143
+ "radius_m": null,
144
+ "height_m": 0.07,
145
+ "mass_kg": 0.1,
146
+ "collision_shape": "box",
147
+ "contact_dimension": 4,
148
+ "rolling_friction": null,
149
+ "capsule_long_axis": null
150
+ },
151
+ "scene_parameters": null,
152
+ "commercial_case": {
153
+ "episode_index": 2,
154
+ "case_id": "sensor_failure_benchmark.target_missing.train.r000.reference",
155
+ "condition_id": "target_missing.reference",
156
+ "condition_group_id": "target_missing",
157
+ "coverage_split": "matched_conditions",
158
+ "pair_id": "sensor_failure_benchmark.target_missing.train.r000",
159
+ "split": "train",
160
+ "role": "reference",
161
+ "product_id": "sensor_failure_benchmark",
162
+ "changed_parameter": "sensor_profile"
163
+ },
164
+ "resolved_config": {
165
+ "kind": "recovery",
166
+ "backend": "mujoco",
167
+ "name": "RGB-D 失效对照 · 同条件独立种子留出",
168
+ "coverage_split": "condition_holdout",
169
+ "scenario_id": "machine_tending",
170
+ "object_shape": "box",
171
+ "material": "opaque",
172
+ "background": "studio",
173
+ "lighting": "studio",
174
+ "difficulty": "standard",
175
+ "object_scale": 1.0,
176
+ "object_mass_kg": 0.1,
177
+ "surface_friction": 1.1,
178
+ "camera_jitter": 0.0,
179
+ "sensor_profile": "clean",
180
+ "sensor_severity": 0.5,
181
+ "empirical_sensor_profile": null,
182
+ "episodes": 1,
183
+ "split_strategy": "single_train",
184
+ "frames_per_episode": 140,
185
+ "fps": 20,
186
+ "width": 160,
187
+ "height": 120,
188
+ "seed": 20262024,
189
+ "camera_count": 1,
190
+ "failure_types": [
191
+ "drop"
192
+ ],
193
+ "blender_samples": 16
194
+ },
195
+ "task_instruction": "Recover the box and place it into the machine fixture nest."
196
+ },
197
+ {
198
+ "episode_index": 3,
199
+ "scenario_id": "machine_tending",
200
+ "object_shape": "box",
201
+ "geometry": {
202
+ "shape": "box",
203
+ "half_extents_m": [
204
+ 0.022,
205
+ 0.022,
206
+ 0.035
207
+ ],
208
+ "radius_m": null,
209
+ "height_m": 0.07,
210
+ "mass_kg": 0.1,
211
+ "collision_shape": "box",
212
+ "contact_dimension": 4,
213
+ "rolling_friction": null,
214
+ "capsule_long_axis": null
215
+ },
216
+ "scene_parameters": null,
217
+ "commercial_case": {
218
+ "episode_index": 3,
219
+ "case_id": "sensor_failure_benchmark.target_missing.train.r000.challenge",
220
+ "condition_id": "target_missing.challenge",
221
+ "condition_group_id": "target_missing",
222
+ "coverage_split": "matched_conditions",
223
+ "pair_id": "sensor_failure_benchmark.target_missing.train.r000",
224
+ "split": "train",
225
+ "role": "challenge",
226
+ "product_id": "sensor_failure_benchmark",
227
+ "changed_parameter": "sensor_profile"
228
+ },
229
+ "resolved_config": {
230
+ "kind": "recovery",
231
+ "backend": "mujoco",
232
+ "name": "RGB-D 失效对照 · 同条件独立种子留出",
233
+ "coverage_split": "condition_holdout",
234
+ "scenario_id": "machine_tending",
235
+ "object_shape": "box",
236
+ "material": "opaque",
237
+ "background": "studio",
238
+ "lighting": "studio",
239
+ "difficulty": "standard",
240
+ "object_scale": 1.0,
241
+ "object_mass_kg": 0.1,
242
+ "surface_friction": 1.1,
243
+ "camera_jitter": 0.0,
244
+ "sensor_profile": "target_dropout",
245
+ "sensor_severity": 0.5,
246
+ "empirical_sensor_profile": null,
247
+ "episodes": 1,
248
+ "split_strategy": "single_train",
249
+ "frames_per_episode": 140,
250
+ "fps": 20,
251
+ "width": 160,
252
+ "height": 120,
253
+ "seed": 20262024,
254
+ "camera_count": 1,
255
+ "failure_types": [
256
+ "drop"
257
+ ],
258
+ "blender_samples": 16
259
+ },
260
+ "task_instruction": "Recover the box and place it into the machine fixture nest."
261
+ },
262
+ {
263
+ "episode_index": 4,
264
+ "scenario_id": "conveyor_pick",
265
+ "object_shape": "cylinder",
266
+ "geometry": {
267
+ "shape": "cylinder",
268
+ "half_extents_m": [
269
+ 0.022,
270
+ 0.022,
271
+ 0.035
272
+ ],
273
+ "radius_m": 0.022,
274
+ "height_m": 0.07,
275
+ "mass_kg": 0.1,
276
+ "collision_shape": "cylinder",
277
+ "contact_dimension": 4,
278
+ "rolling_friction": null,
279
+ "capsule_long_axis": null
280
+ },
281
+ "scene_parameters": null,
282
+ "commercial_case": {
283
+ "episode_index": 4,
284
+ "case_id": "sensor_failure_benchmark.camera_interrupt.train.r000.reference",
285
+ "condition_id": "camera_interrupt.reference",
286
+ "condition_group_id": "camera_interrupt",
287
+ "coverage_split": "matched_conditions",
288
+ "pair_id": "sensor_failure_benchmark.camera_interrupt.train.r000",
289
+ "split": "train",
290
+ "role": "reference",
291
+ "product_id": "sensor_failure_benchmark",
292
+ "changed_parameter": "sensor_profile"
293
+ },
294
+ "resolved_config": {
295
+ "kind": "recovery",
296
+ "backend": "mujoco",
297
+ "name": "RGB-D 失效对照 · 同条件独立种子留出",
298
+ "coverage_split": "condition_holdout",
299
+ "scenario_id": "conveyor_pick",
300
+ "object_shape": "cylinder",
301
+ "material": "opaque",
302
+ "background": "studio",
303
+ "lighting": "studio",
304
+ "difficulty": "standard",
305
+ "object_scale": 1.0,
306
+ "object_mass_kg": 0.1,
307
+ "surface_friction": 1.1,
308
+ "camera_jitter": 0.0,
309
+ "sensor_profile": "clean",
310
+ "sensor_severity": 0.5,
311
+ "empirical_sensor_profile": null,
312
+ "episodes": 1,
313
+ "split_strategy": "single_train",
314
+ "frames_per_episode": 140,
315
+ "fps": 20,
316
+ "width": 160,
317
+ "height": 120,
318
+ "seed": 20263024,
319
+ "camera_count": 1,
320
+ "failure_types": [
321
+ "pose_offset"
322
+ ],
323
+ "blender_samples": 16
324
+ },
325
+ "task_instruction": "Recover the cylinder from the stopped conveyor and place it in the output tray."
326
+ },
327
+ {
328
+ "episode_index": 5,
329
+ "scenario_id": "conveyor_pick",
330
+ "object_shape": "cylinder",
331
+ "geometry": {
332
+ "shape": "cylinder",
333
+ "half_extents_m": [
334
+ 0.022,
335
+ 0.022,
336
+ 0.035
337
+ ],
338
+ "radius_m": 0.022,
339
+ "height_m": 0.07,
340
+ "mass_kg": 0.1,
341
+ "collision_shape": "cylinder",
342
+ "contact_dimension": 4,
343
+ "rolling_friction": null,
344
+ "capsule_long_axis": null
345
+ },
346
+ "scene_parameters": null,
347
+ "commercial_case": {
348
+ "episode_index": 5,
349
+ "case_id": "sensor_failure_benchmark.camera_interrupt.train.r000.challenge",
350
+ "condition_id": "camera_interrupt.challenge",
351
+ "condition_group_id": "camera_interrupt",
352
+ "coverage_split": "matched_conditions",
353
+ "pair_id": "sensor_failure_benchmark.camera_interrupt.train.r000",
354
+ "split": "train",
355
+ "role": "challenge",
356
+ "product_id": "sensor_failure_benchmark",
357
+ "changed_parameter": "sensor_profile"
358
+ },
359
+ "resolved_config": {
360
+ "kind": "recovery",
361
+ "backend": "mujoco",
362
+ "name": "RGB-D 失效对照 · 同条件独立种子留出",
363
+ "coverage_split": "condition_holdout",
364
+ "scenario_id": "conveyor_pick",
365
+ "object_shape": "cylinder",
366
+ "material": "opaque",
367
+ "background": "studio",
368
+ "lighting": "studio",
369
+ "difficulty": "standard",
370
+ "object_scale": 1.0,
371
+ "object_mass_kg": 0.1,
372
+ "surface_friction": 1.1,
373
+ "camera_jitter": 0.0,
374
+ "sensor_profile": "camera_outage",
375
+ "sensor_severity": 0.5,
376
+ "empirical_sensor_profile": null,
377
+ "episodes": 1,
378
+ "split_strategy": "single_train",
379
+ "frames_per_episode": 140,
380
+ "fps": 20,
381
+ "width": 160,
382
+ "height": 120,
383
+ "seed": 20263024,
384
+ "camera_count": 1,
385
+ "failure_types": [
386
+ "pose_offset"
387
+ ],
388
+ "blender_samples": 16
389
+ },
390
+ "task_instruction": "Recover the cylinder from the stopped conveyor and place it in the output tray."
391
+ },
392
+ {
393
+ "episode_index": 6,
394
+ "scenario_id": "pallet_cell_placement",
395
+ "object_shape": "box",
396
+ "geometry": {
397
+ "shape": "box",
398
+ "half_extents_m": [
399
+ 0.022,
400
+ 0.022,
401
+ 0.035
402
+ ],
403
+ "radius_m": null,
404
+ "height_m": 0.07,
405
+ "mass_kg": 0.1,
406
+ "collision_shape": "box",
407
+ "contact_dimension": 4,
408
+ "rolling_friction": null,
409
+ "capsule_long_axis": null
410
+ },
411
+ "scene_parameters": null,
412
+ "commercial_case": {
413
+ "episode_index": 6,
414
+ "case_id": "sensor_failure_benchmark.compound_shift.train.r000.reference",
415
+ "condition_id": "compound_shift.reference",
416
+ "condition_group_id": "compound_shift",
417
+ "coverage_split": "matched_conditions",
418
+ "pair_id": "sensor_failure_benchmark.compound_shift.train.r000",
419
+ "split": "train",
420
+ "role": "reference",
421
+ "product_id": "sensor_failure_benchmark",
422
+ "changed_parameter": "sensor_profile"
423
+ },
424
+ "resolved_config": {
425
+ "kind": "recovery",
426
+ "backend": "mujoco",
427
+ "name": "RGB-D 失效对照 · 同条件独立种子留出",
428
+ "coverage_split": "condition_holdout",
429
+ "scenario_id": "pallet_cell_placement",
430
+ "object_shape": "box",
431
+ "material": "opaque",
432
+ "background": "studio",
433
+ "lighting": "studio",
434
+ "difficulty": "standard",
435
+ "object_scale": 1.0,
436
+ "object_mass_kg": 0.1,
437
+ "surface_friction": 1.1,
438
+ "camera_jitter": 0.0,
439
+ "sensor_profile": "clean",
440
+ "sensor_severity": 0.5,
441
+ "empirical_sensor_profile": null,
442
+ "episodes": 1,
443
+ "split_strategy": "single_train",
444
+ "frames_per_episode": 140,
445
+ "fps": 20,
446
+ "width": 160,
447
+ "height": 120,
448
+ "seed": 20264024,
449
+ "camera_count": 1,
450
+ "failure_types": [
451
+ "miss_grasp"
452
+ ],
453
+ "blender_samples": 16
454
+ },
455
+ "task_instruction": "Recover the box and place it into one pallet cell."
456
+ },
457
+ {
458
+ "episode_index": 7,
459
+ "scenario_id": "pallet_cell_placement",
460
+ "object_shape": "box",
461
+ "geometry": {
462
+ "shape": "box",
463
+ "half_extents_m": [
464
+ 0.022,
465
+ 0.022,
466
+ 0.035
467
+ ],
468
+ "radius_m": null,
469
+ "height_m": 0.07,
470
+ "mass_kg": 0.1,
471
+ "collision_shape": "box",
472
+ "contact_dimension": 4,
473
+ "rolling_friction": null,
474
+ "capsule_long_axis": null
475
+ },
476
+ "scene_parameters": null,
477
+ "commercial_case": {
478
+ "episode_index": 7,
479
+ "case_id": "sensor_failure_benchmark.compound_shift.train.r000.challenge",
480
+ "condition_id": "compound_shift.challenge",
481
+ "condition_group_id": "compound_shift",
482
+ "coverage_split": "matched_conditions",
483
+ "pair_id": "sensor_failure_benchmark.compound_shift.train.r000",
484
+ "split": "train",
485
+ "role": "challenge",
486
+ "product_id": "sensor_failure_benchmark",
487
+ "changed_parameter": "sensor_profile"
488
+ },
489
+ "resolved_config": {
490
+ "kind": "recovery",
491
+ "backend": "mujoco",
492
+ "name": "RGB-D 失效对照 · 同条件独立种子留出",
493
+ "coverage_split": "condition_holdout",
494
+ "scenario_id": "pallet_cell_placement",
495
+ "object_shape": "box",
496
+ "material": "opaque",
497
+ "background": "studio",
498
+ "lighting": "studio",
499
+ "difficulty": "standard",
500
+ "object_scale": 1.0,
501
+ "object_mass_kg": 0.1,
502
+ "surface_friction": 1.1,
503
+ "camera_jitter": 0.0,
504
+ "sensor_profile": "combined",
505
+ "sensor_severity": 0.5,
506
+ "empirical_sensor_profile": null,
507
+ "episodes": 1,
508
+ "split_strategy": "single_train",
509
+ "frames_per_episode": 140,
510
+ "fps": 20,
511
+ "width": 160,
512
+ "height": 120,
513
+ "seed": 20264024,
514
+ "camera_count": 1,
515
+ "failure_types": [
516
+ "miss_grasp"
517
+ ],
518
+ "blender_samples": 16
519
+ },
520
+ "task_instruction": "Recover the box and place it into one pallet cell."
521
+ },
522
+ {
523
+ "episode_index": 8,
524
+ "scenario_id": "machine_tending",
525
+ "object_shape": "box",
526
+ "geometry": {
527
+ "shape": "box",
528
+ "half_extents_m": [
529
+ 0.022,
530
+ 0.022,
531
+ 0.035
532
+ ],
533
+ "radius_m": null,
534
+ "height_m": 0.07,
535
+ "mass_kg": 0.1,
536
+ "collision_shape": "box",
537
+ "contact_dimension": 4,
538
+ "rolling_friction": null,
539
+ "capsule_long_axis": null
540
+ },
541
+ "scene_parameters": null,
542
+ "commercial_case": {
543
+ "episode_index": 8,
544
+ "case_id": "sensor_failure_benchmark.range_noise.validation.r000.reference",
545
+ "condition_id": "range_noise.reference",
546
+ "condition_group_id": "range_noise",
547
+ "coverage_split": "matched_conditions",
548
+ "pair_id": "sensor_failure_benchmark.range_noise.validation.r000",
549
+ "split": "validation",
550
+ "role": "reference",
551
+ "product_id": "sensor_failure_benchmark",
552
+ "changed_parameter": "sensor_profile"
553
+ },
554
+ "resolved_config": {
555
+ "kind": "recovery",
556
+ "backend": "mujoco",
557
+ "name": "RGB-D 失效对照 · 同条件独立种子留出",
558
+ "coverage_split": "condition_holdout",
559
+ "scenario_id": "machine_tending",
560
+ "object_shape": "box",
561
+ "material": "opaque",
562
+ "background": "studio",
563
+ "lighting": "studio",
564
+ "difficulty": "standard",
565
+ "object_scale": 1.0,
566
+ "object_mass_kg": 0.1,
567
+ "surface_friction": 1.1,
568
+ "camera_jitter": 0.0,
569
+ "sensor_profile": "clean",
570
+ "sensor_severity": 0.5,
571
+ "empirical_sensor_profile": null,
572
+ "episodes": 1,
573
+ "split_strategy": "single_train",
574
+ "frames_per_episode": 140,
575
+ "fps": 20,
576
+ "width": 160,
577
+ "height": 120,
578
+ "seed": 20261124,
579
+ "camera_count": 1,
580
+ "failure_types": [
581
+ "slip"
582
+ ],
583
+ "blender_samples": 16
584
+ },
585
+ "task_instruction": "Recover the box and place it into the machine fixture nest."
586
+ },
587
+ {
588
+ "episode_index": 9,
589
+ "scenario_id": "machine_tending",
590
+ "object_shape": "box",
591
+ "geometry": {
592
+ "shape": "box",
593
+ "half_extents_m": [
594
+ 0.022,
595
+ 0.022,
596
+ 0.035
597
+ ],
598
+ "radius_m": null,
599
+ "height_m": 0.07,
600
+ "mass_kg": 0.1,
601
+ "collision_shape": "box",
602
+ "contact_dimension": 4,
603
+ "rolling_friction": null,
604
+ "capsule_long_axis": null
605
+ },
606
+ "scene_parameters": null,
607
+ "commercial_case": {
608
+ "episode_index": 9,
609
+ "case_id": "sensor_failure_benchmark.range_noise.validation.r000.challenge",
610
+ "condition_id": "range_noise.challenge",
611
+ "condition_group_id": "range_noise",
612
+ "coverage_split": "matched_conditions",
613
+ "pair_id": "sensor_failure_benchmark.range_noise.validation.r000",
614
+ "split": "validation",
615
+ "role": "challenge",
616
+ "product_id": "sensor_failure_benchmark",
617
+ "changed_parameter": "sensor_profile"
618
+ },
619
+ "resolved_config": {
620
+ "kind": "recovery",
621
+ "backend": "mujoco",
622
+ "name": "RGB-D 失效对照 · 同条件独立种子留出",
623
+ "coverage_split": "condition_holdout",
624
+ "scenario_id": "machine_tending",
625
+ "object_shape": "box",
626
+ "material": "opaque",
627
+ "background": "studio",
628
+ "lighting": "studio",
629
+ "difficulty": "standard",
630
+ "object_scale": 1.0,
631
+ "object_mass_kg": 0.1,
632
+ "surface_friction": 1.1,
633
+ "camera_jitter": 0.0,
634
+ "sensor_profile": "depth_noise",
635
+ "sensor_severity": 0.5,
636
+ "empirical_sensor_profile": null,
637
+ "episodes": 1,
638
+ "split_strategy": "single_train",
639
+ "frames_per_episode": 140,
640
+ "fps": 20,
641
+ "width": 160,
642
+ "height": 120,
643
+ "seed": 20261124,
644
+ "camera_count": 1,
645
+ "failure_types": [
646
+ "slip"
647
+ ],
648
+ "blender_samples": 16
649
+ },
650
+ "task_instruction": "Recover the box and place it into the machine fixture nest."
651
+ },
652
+ {
653
+ "episode_index": 10,
654
+ "scenario_id": "machine_tending",
655
+ "object_shape": "box",
656
+ "geometry": {
657
+ "shape": "box",
658
+ "half_extents_m": [
659
+ 0.022,
660
+ 0.022,
661
+ 0.035
662
+ ],
663
+ "radius_m": null,
664
+ "height_m": 0.07,
665
+ "mass_kg": 0.1,
666
+ "collision_shape": "box",
667
+ "contact_dimension": 4,
668
+ "rolling_friction": null,
669
+ "capsule_long_axis": null
670
+ },
671
+ "scene_parameters": null,
672
+ "commercial_case": {
673
+ "episode_index": 10,
674
+ "case_id": "sensor_failure_benchmark.target_missing.validation.r000.reference",
675
+ "condition_id": "target_missing.reference",
676
+ "condition_group_id": "target_missing",
677
+ "coverage_split": "matched_conditions",
678
+ "pair_id": "sensor_failure_benchmark.target_missing.validation.r000",
679
+ "split": "validation",
680
+ "role": "reference",
681
+ "product_id": "sensor_failure_benchmark",
682
+ "changed_parameter": "sensor_profile"
683
+ },
684
+ "resolved_config": {
685
+ "kind": "recovery",
686
+ "backend": "mujoco",
687
+ "name": "RGB-D 失效对照 · 同条件独立种子留出",
688
+ "coverage_split": "condition_holdout",
689
+ "scenario_id": "machine_tending",
690
+ "object_shape": "box",
691
+ "material": "opaque",
692
+ "background": "studio",
693
+ "lighting": "studio",
694
+ "difficulty": "standard",
695
+ "object_scale": 1.0,
696
+ "object_mass_kg": 0.1,
697
+ "surface_friction": 1.1,
698
+ "camera_jitter": 0.0,
699
+ "sensor_profile": "clean",
700
+ "sensor_severity": 0.5,
701
+ "empirical_sensor_profile": null,
702
+ "episodes": 1,
703
+ "split_strategy": "single_train",
704
+ "frames_per_episode": 140,
705
+ "fps": 20,
706
+ "width": 160,
707
+ "height": 120,
708
+ "seed": 20262124,
709
+ "camera_count": 1,
710
+ "failure_types": [
711
+ "drop"
712
+ ],
713
+ "blender_samples": 16
714
+ },
715
+ "task_instruction": "Recover the box and place it into the machine fixture nest."
716
+ },
717
+ {
718
+ "episode_index": 11,
719
+ "scenario_id": "machine_tending",
720
+ "object_shape": "box",
721
+ "geometry": {
722
+ "shape": "box",
723
+ "half_extents_m": [
724
+ 0.022,
725
+ 0.022,
726
+ 0.035
727
+ ],
728
+ "radius_m": null,
729
+ "height_m": 0.07,
730
+ "mass_kg": 0.1,
731
+ "collision_shape": "box",
732
+ "contact_dimension": 4,
733
+ "rolling_friction": null,
734
+ "capsule_long_axis": null
735
+ },
736
+ "scene_parameters": null,
737
+ "commercial_case": {
738
+ "episode_index": 11,
739
+ "case_id": "sensor_failure_benchmark.target_missing.validation.r000.challenge",
740
+ "condition_id": "target_missing.challenge",
741
+ "condition_group_id": "target_missing",
742
+ "coverage_split": "matched_conditions",
743
+ "pair_id": "sensor_failure_benchmark.target_missing.validation.r000",
744
+ "split": "validation",
745
+ "role": "challenge",
746
+ "product_id": "sensor_failure_benchmark",
747
+ "changed_parameter": "sensor_profile"
748
+ },
749
+ "resolved_config": {
750
+ "kind": "recovery",
751
+ "backend": "mujoco",
752
+ "name": "RGB-D 失效对照 · 同条件独立种子留出",
753
+ "coverage_split": "condition_holdout",
754
+ "scenario_id": "machine_tending",
755
+ "object_shape": "box",
756
+ "material": "opaque",
757
+ "background": "studio",
758
+ "lighting": "studio",
759
+ "difficulty": "standard",
760
+ "object_scale": 1.0,
761
+ "object_mass_kg": 0.1,
762
+ "surface_friction": 1.1,
763
+ "camera_jitter": 0.0,
764
+ "sensor_profile": "target_dropout",
765
+ "sensor_severity": 0.5,
766
+ "empirical_sensor_profile": null,
767
+ "episodes": 1,
768
+ "split_strategy": "single_train",
769
+ "frames_per_episode": 140,
770
+ "fps": 20,
771
+ "width": 160,
772
+ "height": 120,
773
+ "seed": 20262124,
774
+ "camera_count": 1,
775
+ "failure_types": [
776
+ "drop"
777
+ ],
778
+ "blender_samples": 16
779
+ },
780
+ "task_instruction": "Recover the box and place it into the machine fixture nest."
781
+ },
782
+ {
783
+ "episode_index": 12,
784
+ "scenario_id": "conveyor_pick",
785
+ "object_shape": "cylinder",
786
+ "geometry": {
787
+ "shape": "cylinder",
788
+ "half_extents_m": [
789
+ 0.022,
790
+ 0.022,
791
+ 0.035
792
+ ],
793
+ "radius_m": 0.022,
794
+ "height_m": 0.07,
795
+ "mass_kg": 0.1,
796
+ "collision_shape": "cylinder",
797
+ "contact_dimension": 4,
798
+ "rolling_friction": null,
799
+ "capsule_long_axis": null
800
+ },
801
+ "scene_parameters": null,
802
+ "commercial_case": {
803
+ "episode_index": 12,
804
+ "case_id": "sensor_failure_benchmark.camera_interrupt.validation.r000.reference",
805
+ "condition_id": "camera_interrupt.reference",
806
+ "condition_group_id": "camera_interrupt",
807
+ "coverage_split": "matched_conditions",
808
+ "pair_id": "sensor_failure_benchmark.camera_interrupt.validation.r000",
809
+ "split": "validation",
810
+ "role": "reference",
811
+ "product_id": "sensor_failure_benchmark",
812
+ "changed_parameter": "sensor_profile"
813
+ },
814
+ "resolved_config": {
815
+ "kind": "recovery",
816
+ "backend": "mujoco",
817
+ "name": "RGB-D 失效对照 · 同条件独立种子留出",
818
+ "coverage_split": "condition_holdout",
819
+ "scenario_id": "conveyor_pick",
820
+ "object_shape": "cylinder",
821
+ "material": "opaque",
822
+ "background": "studio",
823
+ "lighting": "studio",
824
+ "difficulty": "standard",
825
+ "object_scale": 1.0,
826
+ "object_mass_kg": 0.1,
827
+ "surface_friction": 1.1,
828
+ "camera_jitter": 0.0,
829
+ "sensor_profile": "clean",
830
+ "sensor_severity": 0.5,
831
+ "empirical_sensor_profile": null,
832
+ "episodes": 1,
833
+ "split_strategy": "single_train",
834
+ "frames_per_episode": 140,
835
+ "fps": 20,
836
+ "width": 160,
837
+ "height": 120,
838
+ "seed": 20263124,
839
+ "camera_count": 1,
840
+ "failure_types": [
841
+ "pose_offset"
842
+ ],
843
+ "blender_samples": 16
844
+ },
845
+ "task_instruction": "Recover the cylinder from the stopped conveyor and place it in the output tray."
846
+ },
847
+ {
848
+ "episode_index": 13,
849
+ "scenario_id": "conveyor_pick",
850
+ "object_shape": "cylinder",
851
+ "geometry": {
852
+ "shape": "cylinder",
853
+ "half_extents_m": [
854
+ 0.022,
855
+ 0.022,
856
+ 0.035
857
+ ],
858
+ "radius_m": 0.022,
859
+ "height_m": 0.07,
860
+ "mass_kg": 0.1,
861
+ "collision_shape": "cylinder",
862
+ "contact_dimension": 4,
863
+ "rolling_friction": null,
864
+ "capsule_long_axis": null
865
+ },
866
+ "scene_parameters": null,
867
+ "commercial_case": {
868
+ "episode_index": 13,
869
+ "case_id": "sensor_failure_benchmark.camera_interrupt.validation.r000.challenge",
870
+ "condition_id": "camera_interrupt.challenge",
871
+ "condition_group_id": "camera_interrupt",
872
+ "coverage_split": "matched_conditions",
873
+ "pair_id": "sensor_failure_benchmark.camera_interrupt.validation.r000",
874
+ "split": "validation",
875
+ "role": "challenge",
876
+ "product_id": "sensor_failure_benchmark",
877
+ "changed_parameter": "sensor_profile"
878
+ },
879
+ "resolved_config": {
880
+ "kind": "recovery",
881
+ "backend": "mujoco",
882
+ "name": "RGB-D 失效对照 · 同条件独立种子留出",
883
+ "coverage_split": "condition_holdout",
884
+ "scenario_id": "conveyor_pick",
885
+ "object_shape": "cylinder",
886
+ "material": "opaque",
887
+ "background": "studio",
888
+ "lighting": "studio",
889
+ "difficulty": "standard",
890
+ "object_scale": 1.0,
891
+ "object_mass_kg": 0.1,
892
+ "surface_friction": 1.1,
893
+ "camera_jitter": 0.0,
894
+ "sensor_profile": "camera_outage",
895
+ "sensor_severity": 0.5,
896
+ "empirical_sensor_profile": null,
897
+ "episodes": 1,
898
+ "split_strategy": "single_train",
899
+ "frames_per_episode": 140,
900
+ "fps": 20,
901
+ "width": 160,
902
+ "height": 120,
903
+ "seed": 20263124,
904
+ "camera_count": 1,
905
+ "failure_types": [
906
+ "pose_offset"
907
+ ],
908
+ "blender_samples": 16
909
+ },
910
+ "task_instruction": "Recover the cylinder from the stopped conveyor and place it in the output tray."
911
+ },
912
+ {
913
+ "episode_index": 14,
914
+ "scenario_id": "pallet_cell_placement",
915
+ "object_shape": "box",
916
+ "geometry": {
917
+ "shape": "box",
918
+ "half_extents_m": [
919
+ 0.022,
920
+ 0.022,
921
+ 0.035
922
+ ],
923
+ "radius_m": null,
924
+ "height_m": 0.07,
925
+ "mass_kg": 0.1,
926
+ "collision_shape": "box",
927
+ "contact_dimension": 4,
928
+ "rolling_friction": null,
929
+ "capsule_long_axis": null
930
+ },
931
+ "scene_parameters": null,
932
+ "commercial_case": {
933
+ "episode_index": 14,
934
+ "case_id": "sensor_failure_benchmark.compound_shift.validation.r000.reference",
935
+ "condition_id": "compound_shift.reference",
936
+ "condition_group_id": "compound_shift",
937
+ "coverage_split": "matched_conditions",
938
+ "pair_id": "sensor_failure_benchmark.compound_shift.validation.r000",
939
+ "split": "validation",
940
+ "role": "reference",
941
+ "product_id": "sensor_failure_benchmark",
942
+ "changed_parameter": "sensor_profile"
943
+ },
944
+ "resolved_config": {
945
+ "kind": "recovery",
946
+ "backend": "mujoco",
947
+ "name": "RGB-D 失效对照 · 同条件独立种子留出",
948
+ "coverage_split": "condition_holdout",
949
+ "scenario_id": "pallet_cell_placement",
950
+ "object_shape": "box",
951
+ "material": "opaque",
952
+ "background": "studio",
953
+ "lighting": "studio",
954
+ "difficulty": "standard",
955
+ "object_scale": 1.0,
956
+ "object_mass_kg": 0.1,
957
+ "surface_friction": 1.1,
958
+ "camera_jitter": 0.0,
959
+ "sensor_profile": "clean",
960
+ "sensor_severity": 0.5,
961
+ "empirical_sensor_profile": null,
962
+ "episodes": 1,
963
+ "split_strategy": "single_train",
964
+ "frames_per_episode": 140,
965
+ "fps": 20,
966
+ "width": 160,
967
+ "height": 120,
968
+ "seed": 20264124,
969
+ "camera_count": 1,
970
+ "failure_types": [
971
+ "miss_grasp"
972
+ ],
973
+ "blender_samples": 16
974
+ },
975
+ "task_instruction": "Recover the box and place it into one pallet cell."
976
+ },
977
+ {
978
+ "episode_index": 15,
979
+ "scenario_id": "pallet_cell_placement",
980
+ "object_shape": "box",
981
+ "geometry": {
982
+ "shape": "box",
983
+ "half_extents_m": [
984
+ 0.022,
985
+ 0.022,
986
+ 0.035
987
+ ],
988
+ "radius_m": null,
989
+ "height_m": 0.07,
990
+ "mass_kg": 0.1,
991
+ "collision_shape": "box",
992
+ "contact_dimension": 4,
993
+ "rolling_friction": null,
994
+ "capsule_long_axis": null
995
+ },
996
+ "scene_parameters": null,
997
+ "commercial_case": {
998
+ "episode_index": 15,
999
+ "case_id": "sensor_failure_benchmark.compound_shift.validation.r000.challenge",
1000
+ "condition_id": "compound_shift.challenge",
1001
+ "condition_group_id": "compound_shift",
1002
+ "coverage_split": "matched_conditions",
1003
+ "pair_id": "sensor_failure_benchmark.compound_shift.validation.r000",
1004
+ "split": "validation",
1005
+ "role": "challenge",
1006
+ "product_id": "sensor_failure_benchmark",
1007
+ "changed_parameter": "sensor_profile"
1008
+ },
1009
+ "resolved_config": {
1010
+ "kind": "recovery",
1011
+ "backend": "mujoco",
1012
+ "name": "RGB-D 失效对照 · 同条件独立种子留出",
1013
+ "coverage_split": "condition_holdout",
1014
+ "scenario_id": "pallet_cell_placement",
1015
+ "object_shape": "box",
1016
+ "material": "opaque",
1017
+ "background": "studio",
1018
+ "lighting": "studio",
1019
+ "difficulty": "standard",
1020
+ "object_scale": 1.0,
1021
+ "object_mass_kg": 0.1,
1022
+ "surface_friction": 1.1,
1023
+ "camera_jitter": 0.0,
1024
+ "sensor_profile": "combined",
1025
+ "sensor_severity": 0.5,
1026
+ "empirical_sensor_profile": null,
1027
+ "episodes": 1,
1028
+ "split_strategy": "single_train",
1029
+ "frames_per_episode": 140,
1030
+ "fps": 20,
1031
+ "width": 160,
1032
+ "height": 120,
1033
+ "seed": 20264124,
1034
+ "camera_count": 1,
1035
+ "failure_types": [
1036
+ "miss_grasp"
1037
+ ],
1038
+ "blender_samples": 16
1039
+ },
1040
+ "task_instruction": "Recover the box and place it into one pallet cell."
1041
+ },
1042
+ {
1043
+ "episode_index": 16,
1044
+ "scenario_id": "machine_tending",
1045
+ "object_shape": "box",
1046
+ "geometry": {
1047
+ "shape": "box",
1048
+ "half_extents_m": [
1049
+ 0.022,
1050
+ 0.022,
1051
+ 0.035
1052
+ ],
1053
+ "radius_m": null,
1054
+ "height_m": 0.07,
1055
+ "mass_kg": 0.1,
1056
+ "collision_shape": "box",
1057
+ "contact_dimension": 4,
1058
+ "rolling_friction": null,
1059
+ "capsule_long_axis": null
1060
+ },
1061
+ "scene_parameters": null,
1062
+ "commercial_case": {
1063
+ "episode_index": 16,
1064
+ "case_id": "sensor_failure_benchmark.range_noise.test.r000.reference",
1065
+ "condition_id": "range_noise.reference",
1066
+ "condition_group_id": "range_noise",
1067
+ "coverage_split": "matched_conditions",
1068
+ "pair_id": "sensor_failure_benchmark.range_noise.test.r000",
1069
+ "split": "test",
1070
+ "role": "reference",
1071
+ "product_id": "sensor_failure_benchmark",
1072
+ "changed_parameter": "sensor_profile"
1073
+ },
1074
+ "resolved_config": {
1075
+ "kind": "recovery",
1076
+ "backend": "mujoco",
1077
+ "name": "RGB-D 失效对照 · 同条件独立种子留出",
1078
+ "coverage_split": "condition_holdout",
1079
+ "scenario_id": "machine_tending",
1080
+ "object_shape": "box",
1081
+ "material": "opaque",
1082
+ "background": "studio",
1083
+ "lighting": "studio",
1084
+ "difficulty": "standard",
1085
+ "object_scale": 1.0,
1086
+ "object_mass_kg": 0.1,
1087
+ "surface_friction": 1.1,
1088
+ "camera_jitter": 0.0,
1089
+ "sensor_profile": "clean",
1090
+ "sensor_severity": 0.5,
1091
+ "empirical_sensor_profile": null,
1092
+ "episodes": 1,
1093
+ "split_strategy": "single_train",
1094
+ "frames_per_episode": 140,
1095
+ "fps": 20,
1096
+ "width": 160,
1097
+ "height": 120,
1098
+ "seed": 20261224,
1099
+ "camera_count": 1,
1100
+ "failure_types": [
1101
+ "slip"
1102
+ ],
1103
+ "blender_samples": 16
1104
+ },
1105
+ "task_instruction": "Recover the box and place it into the machine fixture nest."
1106
+ },
1107
+ {
1108
+ "episode_index": 17,
1109
+ "scenario_id": "machine_tending",
1110
+ "object_shape": "box",
1111
+ "geometry": {
1112
+ "shape": "box",
1113
+ "half_extents_m": [
1114
+ 0.022,
1115
+ 0.022,
1116
+ 0.035
1117
+ ],
1118
+ "radius_m": null,
1119
+ "height_m": 0.07,
1120
+ "mass_kg": 0.1,
1121
+ "collision_shape": "box",
1122
+ "contact_dimension": 4,
1123
+ "rolling_friction": null,
1124
+ "capsule_long_axis": null
1125
+ },
1126
+ "scene_parameters": null,
1127
+ "commercial_case": {
1128
+ "episode_index": 17,
1129
+ "case_id": "sensor_failure_benchmark.range_noise.test.r000.challenge",
1130
+ "condition_id": "range_noise.challenge",
1131
+ "condition_group_id": "range_noise",
1132
+ "coverage_split": "matched_conditions",
1133
+ "pair_id": "sensor_failure_benchmark.range_noise.test.r000",
1134
+ "split": "test",
1135
+ "role": "challenge",
1136
+ "product_id": "sensor_failure_benchmark",
1137
+ "changed_parameter": "sensor_profile"
1138
+ },
1139
+ "resolved_config": {
1140
+ "kind": "recovery",
1141
+ "backend": "mujoco",
1142
+ "name": "RGB-D 失效对照 · 同条件独立种子留出",
1143
+ "coverage_split": "condition_holdout",
1144
+ "scenario_id": "machine_tending",
1145
+ "object_shape": "box",
1146
+ "material": "opaque",
1147
+ "background": "studio",
1148
+ "lighting": "studio",
1149
+ "difficulty": "standard",
1150
+ "object_scale": 1.0,
1151
+ "object_mass_kg": 0.1,
1152
+ "surface_friction": 1.1,
1153
+ "camera_jitter": 0.0,
1154
+ "sensor_profile": "depth_noise",
1155
+ "sensor_severity": 0.5,
1156
+ "empirical_sensor_profile": null,
1157
+ "episodes": 1,
1158
+ "split_strategy": "single_train",
1159
+ "frames_per_episode": 140,
1160
+ "fps": 20,
1161
+ "width": 160,
1162
+ "height": 120,
1163
+ "seed": 20261224,
1164
+ "camera_count": 1,
1165
+ "failure_types": [
1166
+ "slip"
1167
+ ],
1168
+ "blender_samples": 16
1169
+ },
1170
+ "task_instruction": "Recover the box and place it into the machine fixture nest."
1171
+ },
1172
+ {
1173
+ "episode_index": 18,
1174
+ "scenario_id": "machine_tending",
1175
+ "object_shape": "box",
1176
+ "geometry": {
1177
+ "shape": "box",
1178
+ "half_extents_m": [
1179
+ 0.022,
1180
+ 0.022,
1181
+ 0.035
1182
+ ],
1183
+ "radius_m": null,
1184
+ "height_m": 0.07,
1185
+ "mass_kg": 0.1,
1186
+ "collision_shape": "box",
1187
+ "contact_dimension": 4,
1188
+ "rolling_friction": null,
1189
+ "capsule_long_axis": null
1190
+ },
1191
+ "scene_parameters": null,
1192
+ "commercial_case": {
1193
+ "episode_index": 18,
1194
+ "case_id": "sensor_failure_benchmark.target_missing.test.r000.reference",
1195
+ "condition_id": "target_missing.reference",
1196
+ "condition_group_id": "target_missing",
1197
+ "coverage_split": "matched_conditions",
1198
+ "pair_id": "sensor_failure_benchmark.target_missing.test.r000",
1199
+ "split": "test",
1200
+ "role": "reference",
1201
+ "product_id": "sensor_failure_benchmark",
1202
+ "changed_parameter": "sensor_profile"
1203
+ },
1204
+ "resolved_config": {
1205
+ "kind": "recovery",
1206
+ "backend": "mujoco",
1207
+ "name": "RGB-D 失效对照 · 同条件独立种子留出",
1208
+ "coverage_split": "condition_holdout",
1209
+ "scenario_id": "machine_tending",
1210
+ "object_shape": "box",
1211
+ "material": "opaque",
1212
+ "background": "studio",
1213
+ "lighting": "studio",
1214
+ "difficulty": "standard",
1215
+ "object_scale": 1.0,
1216
+ "object_mass_kg": 0.1,
1217
+ "surface_friction": 1.1,
1218
+ "camera_jitter": 0.0,
1219
+ "sensor_profile": "clean",
1220
+ "sensor_severity": 0.5,
1221
+ "empirical_sensor_profile": null,
1222
+ "episodes": 1,
1223
+ "split_strategy": "single_train",
1224
+ "frames_per_episode": 140,
1225
+ "fps": 20,
1226
+ "width": 160,
1227
+ "height": 120,
1228
+ "seed": 20262224,
1229
+ "camera_count": 1,
1230
+ "failure_types": [
1231
+ "drop"
1232
+ ],
1233
+ "blender_samples": 16
1234
+ },
1235
+ "task_instruction": "Recover the box and place it into the machine fixture nest."
1236
+ },
1237
+ {
1238
+ "episode_index": 19,
1239
+ "scenario_id": "machine_tending",
1240
+ "object_shape": "box",
1241
+ "geometry": {
1242
+ "shape": "box",
1243
+ "half_extents_m": [
1244
+ 0.022,
1245
+ 0.022,
1246
+ 0.035
1247
+ ],
1248
+ "radius_m": null,
1249
+ "height_m": 0.07,
1250
+ "mass_kg": 0.1,
1251
+ "collision_shape": "box",
1252
+ "contact_dimension": 4,
1253
+ "rolling_friction": null,
1254
+ "capsule_long_axis": null
1255
+ },
1256
+ "scene_parameters": null,
1257
+ "commercial_case": {
1258
+ "episode_index": 19,
1259
+ "case_id": "sensor_failure_benchmark.target_missing.test.r000.challenge",
1260
+ "condition_id": "target_missing.challenge",
1261
+ "condition_group_id": "target_missing",
1262
+ "coverage_split": "matched_conditions",
1263
+ "pair_id": "sensor_failure_benchmark.target_missing.test.r000",
1264
+ "split": "test",
1265
+ "role": "challenge",
1266
+ "product_id": "sensor_failure_benchmark",
1267
+ "changed_parameter": "sensor_profile"
1268
+ },
1269
+ "resolved_config": {
1270
+ "kind": "recovery",
1271
+ "backend": "mujoco",
1272
+ "name": "RGB-D 失效对照 · 同条件独立种子留出",
1273
+ "coverage_split": "condition_holdout",
1274
+ "scenario_id": "machine_tending",
1275
+ "object_shape": "box",
1276
+ "material": "opaque",
1277
+ "background": "studio",
1278
+ "lighting": "studio",
1279
+ "difficulty": "standard",
1280
+ "object_scale": 1.0,
1281
+ "object_mass_kg": 0.1,
1282
+ "surface_friction": 1.1,
1283
+ "camera_jitter": 0.0,
1284
+ "sensor_profile": "target_dropout",
1285
+ "sensor_severity": 0.5,
1286
+ "empirical_sensor_profile": null,
1287
+ "episodes": 1,
1288
+ "split_strategy": "single_train",
1289
+ "frames_per_episode": 140,
1290
+ "fps": 20,
1291
+ "width": 160,
1292
+ "height": 120,
1293
+ "seed": 20262224,
1294
+ "camera_count": 1,
1295
+ "failure_types": [
1296
+ "drop"
1297
+ ],
1298
+ "blender_samples": 16
1299
+ },
1300
+ "task_instruction": "Recover the box and place it into the machine fixture nest."
1301
+ },
1302
+ {
1303
+ "episode_index": 20,
1304
+ "scenario_id": "conveyor_pick",
1305
+ "object_shape": "cylinder",
1306
+ "geometry": {
1307
+ "shape": "cylinder",
1308
+ "half_extents_m": [
1309
+ 0.022,
1310
+ 0.022,
1311
+ 0.035
1312
+ ],
1313
+ "radius_m": 0.022,
1314
+ "height_m": 0.07,
1315
+ "mass_kg": 0.1,
1316
+ "collision_shape": "cylinder",
1317
+ "contact_dimension": 4,
1318
+ "rolling_friction": null,
1319
+ "capsule_long_axis": null
1320
+ },
1321
+ "scene_parameters": null,
1322
+ "commercial_case": {
1323
+ "episode_index": 20,
1324
+ "case_id": "sensor_failure_benchmark.camera_interrupt.test.r000.reference",
1325
+ "condition_id": "camera_interrupt.reference",
1326
+ "condition_group_id": "camera_interrupt",
1327
+ "coverage_split": "matched_conditions",
1328
+ "pair_id": "sensor_failure_benchmark.camera_interrupt.test.r000",
1329
+ "split": "test",
1330
+ "role": "reference",
1331
+ "product_id": "sensor_failure_benchmark",
1332
+ "changed_parameter": "sensor_profile"
1333
+ },
1334
+ "resolved_config": {
1335
+ "kind": "recovery",
1336
+ "backend": "mujoco",
1337
+ "name": "RGB-D 失效对照 · 同条件独立种子留出",
1338
+ "coverage_split": "condition_holdout",
1339
+ "scenario_id": "conveyor_pick",
1340
+ "object_shape": "cylinder",
1341
+ "material": "opaque",
1342
+ "background": "studio",
1343
+ "lighting": "studio",
1344
+ "difficulty": "standard",
1345
+ "object_scale": 1.0,
1346
+ "object_mass_kg": 0.1,
1347
+ "surface_friction": 1.1,
1348
+ "camera_jitter": 0.0,
1349
+ "sensor_profile": "clean",
1350
+ "sensor_severity": 0.5,
1351
+ "empirical_sensor_profile": null,
1352
+ "episodes": 1,
1353
+ "split_strategy": "single_train",
1354
+ "frames_per_episode": 140,
1355
+ "fps": 20,
1356
+ "width": 160,
1357
+ "height": 120,
1358
+ "seed": 20263224,
1359
+ "camera_count": 1,
1360
+ "failure_types": [
1361
+ "pose_offset"
1362
+ ],
1363
+ "blender_samples": 16
1364
+ },
1365
+ "task_instruction": "Recover the cylinder from the stopped conveyor and place it in the output tray."
1366
+ },
1367
+ {
1368
+ "episode_index": 21,
1369
+ "scenario_id": "conveyor_pick",
1370
+ "object_shape": "cylinder",
1371
+ "geometry": {
1372
+ "shape": "cylinder",
1373
+ "half_extents_m": [
1374
+ 0.022,
1375
+ 0.022,
1376
+ 0.035
1377
+ ],
1378
+ "radius_m": 0.022,
1379
+ "height_m": 0.07,
1380
+ "mass_kg": 0.1,
1381
+ "collision_shape": "cylinder",
1382
+ "contact_dimension": 4,
1383
+ "rolling_friction": null,
1384
+ "capsule_long_axis": null
1385
+ },
1386
+ "scene_parameters": null,
1387
+ "commercial_case": {
1388
+ "episode_index": 21,
1389
+ "case_id": "sensor_failure_benchmark.camera_interrupt.test.r000.challenge",
1390
+ "condition_id": "camera_interrupt.challenge",
1391
+ "condition_group_id": "camera_interrupt",
1392
+ "coverage_split": "matched_conditions",
1393
+ "pair_id": "sensor_failure_benchmark.camera_interrupt.test.r000",
1394
+ "split": "test",
1395
+ "role": "challenge",
1396
+ "product_id": "sensor_failure_benchmark",
1397
+ "changed_parameter": "sensor_profile"
1398
+ },
1399
+ "resolved_config": {
1400
+ "kind": "recovery",
1401
+ "backend": "mujoco",
1402
+ "name": "RGB-D 失效对照 · 同条件独立种子留出",
1403
+ "coverage_split": "condition_holdout",
1404
+ "scenario_id": "conveyor_pick",
1405
+ "object_shape": "cylinder",
1406
+ "material": "opaque",
1407
+ "background": "studio",
1408
+ "lighting": "studio",
1409
+ "difficulty": "standard",
1410
+ "object_scale": 1.0,
1411
+ "object_mass_kg": 0.1,
1412
+ "surface_friction": 1.1,
1413
+ "camera_jitter": 0.0,
1414
+ "sensor_profile": "camera_outage",
1415
+ "sensor_severity": 0.5,
1416
+ "empirical_sensor_profile": null,
1417
+ "episodes": 1,
1418
+ "split_strategy": "single_train",
1419
+ "frames_per_episode": 140,
1420
+ "fps": 20,
1421
+ "width": 160,
1422
+ "height": 120,
1423
+ "seed": 20263224,
1424
+ "camera_count": 1,
1425
+ "failure_types": [
1426
+ "pose_offset"
1427
+ ],
1428
+ "blender_samples": 16
1429
+ },
1430
+ "task_instruction": "Recover the cylinder from the stopped conveyor and place it in the output tray."
1431
+ },
1432
+ {
1433
+ "episode_index": 22,
1434
+ "scenario_id": "pallet_cell_placement",
1435
+ "object_shape": "box",
1436
+ "geometry": {
1437
+ "shape": "box",
1438
+ "half_extents_m": [
1439
+ 0.022,
1440
+ 0.022,
1441
+ 0.035
1442
+ ],
1443
+ "radius_m": null,
1444
+ "height_m": 0.07,
1445
+ "mass_kg": 0.1,
1446
+ "collision_shape": "box",
1447
+ "contact_dimension": 4,
1448
+ "rolling_friction": null,
1449
+ "capsule_long_axis": null
1450
+ },
1451
+ "scene_parameters": null,
1452
+ "commercial_case": {
1453
+ "episode_index": 22,
1454
+ "case_id": "sensor_failure_benchmark.compound_shift.test.r000.reference",
1455
+ "condition_id": "compound_shift.reference",
1456
+ "condition_group_id": "compound_shift",
1457
+ "coverage_split": "matched_conditions",
1458
+ "pair_id": "sensor_failure_benchmark.compound_shift.test.r000",
1459
+ "split": "test",
1460
+ "role": "reference",
1461
+ "product_id": "sensor_failure_benchmark",
1462
+ "changed_parameter": "sensor_profile"
1463
+ },
1464
+ "resolved_config": {
1465
+ "kind": "recovery",
1466
+ "backend": "mujoco",
1467
+ "name": "RGB-D 失效对照 · 同条件独立种子留出",
1468
+ "coverage_split": "condition_holdout",
1469
+ "scenario_id": "pallet_cell_placement",
1470
+ "object_shape": "box",
1471
+ "material": "opaque",
1472
+ "background": "studio",
1473
+ "lighting": "studio",
1474
+ "difficulty": "standard",
1475
+ "object_scale": 1.0,
1476
+ "object_mass_kg": 0.1,
1477
+ "surface_friction": 1.1,
1478
+ "camera_jitter": 0.0,
1479
+ "sensor_profile": "clean",
1480
+ "sensor_severity": 0.5,
1481
+ "empirical_sensor_profile": null,
1482
+ "episodes": 1,
1483
+ "split_strategy": "single_train",
1484
+ "frames_per_episode": 140,
1485
+ "fps": 20,
1486
+ "width": 160,
1487
+ "height": 120,
1488
+ "seed": 20264224,
1489
+ "camera_count": 1,
1490
+ "failure_types": [
1491
+ "miss_grasp"
1492
+ ],
1493
+ "blender_samples": 16
1494
+ },
1495
+ "task_instruction": "Recover the box and place it into one pallet cell."
1496
+ },
1497
+ {
1498
+ "episode_index": 23,
1499
+ "scenario_id": "pallet_cell_placement",
1500
+ "object_shape": "box",
1501
+ "geometry": {
1502
+ "shape": "box",
1503
+ "half_extents_m": [
1504
+ 0.022,
1505
+ 0.022,
1506
+ 0.035
1507
+ ],
1508
+ "radius_m": null,
1509
+ "height_m": 0.07,
1510
+ "mass_kg": 0.1,
1511
+ "collision_shape": "box",
1512
+ "contact_dimension": 4,
1513
+ "rolling_friction": null,
1514
+ "capsule_long_axis": null
1515
+ },
1516
+ "scene_parameters": null,
1517
+ "commercial_case": {
1518
+ "episode_index": 23,
1519
+ "case_id": "sensor_failure_benchmark.compound_shift.test.r000.challenge",
1520
+ "condition_id": "compound_shift.challenge",
1521
+ "condition_group_id": "compound_shift",
1522
+ "coverage_split": "matched_conditions",
1523
+ "pair_id": "sensor_failure_benchmark.compound_shift.test.r000",
1524
+ "split": "test",
1525
+ "role": "challenge",
1526
+ "product_id": "sensor_failure_benchmark",
1527
+ "changed_parameter": "sensor_profile"
1528
+ },
1529
+ "resolved_config": {
1530
+ "kind": "recovery",
1531
+ "backend": "mujoco",
1532
+ "name": "RGB-D 失效对照 · 同条件独立种子留出",
1533
+ "coverage_split": "condition_holdout",
1534
+ "scenario_id": "pallet_cell_placement",
1535
+ "object_shape": "box",
1536
+ "material": "opaque",
1537
+ "background": "studio",
1538
+ "lighting": "studio",
1539
+ "difficulty": "standard",
1540
+ "object_scale": 1.0,
1541
+ "object_mass_kg": 0.1,
1542
+ "surface_friction": 1.1,
1543
+ "camera_jitter": 0.0,
1544
+ "sensor_profile": "combined",
1545
+ "sensor_severity": 0.5,
1546
+ "empirical_sensor_profile": null,
1547
+ "episodes": 1,
1548
+ "split_strategy": "single_train",
1549
+ "frames_per_episode": 140,
1550
+ "fps": 20,
1551
+ "width": 160,
1552
+ "height": 120,
1553
+ "seed": 20264224,
1554
+ "camera_count": 1,
1555
+ "failure_types": [
1556
+ "miss_grasp"
1557
+ ],
1558
+ "blender_samples": 16
1559
+ },
1560
+ "task_instruction": "Recover the box and place it into one pallet cell."
1561
+ }
1562
+ ]
dataset/meta/splits.json ADDED
@@ -0,0 +1,108 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "method": "paired_scene_group_holdout",
3
+ "coverage_split": "matched_conditions",
4
+ "condition_coverage": {
5
+ "train": [
6
+ "camera_interrupt.challenge",
7
+ "camera_interrupt.reference",
8
+ "compound_shift.challenge",
9
+ "compound_shift.reference",
10
+ "range_noise.challenge",
11
+ "range_noise.reference",
12
+ "target_missing.challenge",
13
+ "target_missing.reference"
14
+ ],
15
+ "validation": [
16
+ "camera_interrupt.challenge",
17
+ "camera_interrupt.reference",
18
+ "compound_shift.challenge",
19
+ "compound_shift.reference",
20
+ "range_noise.challenge",
21
+ "range_noise.reference",
22
+ "target_missing.challenge",
23
+ "target_missing.reference"
24
+ ],
25
+ "test": [
26
+ "camera_interrupt.challenge",
27
+ "camera_interrupt.reference",
28
+ "compound_shift.challenge",
29
+ "compound_shift.reference",
30
+ "range_noise.challenge",
31
+ "range_noise.reference",
32
+ "target_missing.challenge",
33
+ "target_missing.reference"
34
+ ]
35
+ },
36
+ "independent_scene_pairs": {
37
+ "train": 4,
38
+ "validation": 4,
39
+ "test": 4
40
+ },
41
+ "episode_indices": {
42
+ "train": [
43
+ 0,
44
+ 1,
45
+ 2,
46
+ 3,
47
+ 4,
48
+ 5,
49
+ 6,
50
+ 7
51
+ ],
52
+ "validation": [
53
+ 8,
54
+ 9,
55
+ 10,
56
+ 11,
57
+ 12,
58
+ 13,
59
+ 14,
60
+ 15
61
+ ],
62
+ "test": [
63
+ 16,
64
+ 17,
65
+ 18,
66
+ 19,
67
+ 20,
68
+ 21,
69
+ 22,
70
+ 23
71
+ ]
72
+ },
73
+ "ranges": {
74
+ "train": "0:8",
75
+ "validation": "8:16",
76
+ "test": "16:24"
77
+ },
78
+ "pair_assignments": {
79
+ "sensor_failure_benchmark.range_noise.train.r000": "train",
80
+ "sensor_failure_benchmark.target_missing.train.r000": "train",
81
+ "sensor_failure_benchmark.camera_interrupt.train.r000": "train",
82
+ "sensor_failure_benchmark.compound_shift.train.r000": "train",
83
+ "sensor_failure_benchmark.range_noise.validation.r000": "validation",
84
+ "sensor_failure_benchmark.target_missing.validation.r000": "validation",
85
+ "sensor_failure_benchmark.camera_interrupt.validation.r000": "validation",
86
+ "sensor_failure_benchmark.compound_shift.validation.r000": "validation",
87
+ "sensor_failure_benchmark.range_noise.test.r000": "test",
88
+ "sensor_failure_benchmark.target_missing.test.r000": "test",
89
+ "sensor_failure_benchmark.camera_interrupt.test.r000": "test",
90
+ "sensor_failure_benchmark.compound_shift.test.r000": "test"
91
+ },
92
+ "seed_assignments": {
93
+ "20261024": "train",
94
+ "20262024": "train",
95
+ "20263024": "train",
96
+ "20264024": "train",
97
+ "20261124": "validation",
98
+ "20262124": "validation",
99
+ "20263124": "validation",
100
+ "20264124": "validation",
101
+ "20261224": "test",
102
+ "20262224": "test",
103
+ "20263224": "test",
104
+ "20264224": "test"
105
+ },
106
+ "group_isolation_verified": true,
107
+ "scope": "complete-episode split; paired scenes and scene seeds remain isolated when a holdout is used"
108
+ }
dataset/meta/stats.json ADDED
@@ -0,0 +1,443 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "observation.state": {
3
+ "min": [
4
+ -0.014122764021777623,
5
+ -0.1000000225028425,
6
+ -0.0002008240793322001,
7
+ -0.0017723264821356429,
8
+ -0.0014710312643219869,
9
+ -0.09362799765317938,
10
+ -0.2233885095369417,
11
+ -0.3456166325545538,
12
+ -0.05734092435609014,
13
+ -0.055497768418793066
14
+ ],
15
+ "max": [
16
+ 0.19999999999999982,
17
+ 0.09999999999999991,
18
+ 0.1493811324297467,
19
+ 0.04111885307612435,
20
+ 0.04107339680984956,
21
+ 0.35915696105202416,
22
+ 0.22226931286421447,
23
+ 0.34602697011587097,
24
+ 0.25920105647017433,
25
+ 0.2581810315060028
26
+ ],
27
+ "mean": [
28
+ 0.13069622870961353,
29
+ 0.012550514504649833,
30
+ 0.08184252925113868,
31
+ 0.016194231889651797,
32
+ 0.016217433140424682,
33
+ 0.017014304812024473,
34
+ 0.0026307121559819677,
35
+ 0.006947751533805151,
36
+ 0.003171810917318101,
37
+ 0.003905762686124316
38
+ ],
39
+ "std": [
40
+ 0.0787665187274976,
41
+ 0.06294078291544838,
42
+ 0.0394299427641389,
43
+ 0.019663202307695832,
44
+ 0.019685423812493552,
45
+ 0.06620288861031719,
46
+ 0.037415485652098034,
47
+ 0.11736775749429738,
48
+ 0.03152439092122224,
49
+ 0.03134431841053177
50
+ ],
51
+ "count": [
52
+ 1120
53
+ ]
54
+ },
55
+ "action": {
56
+ "min": [
57
+ -0.01412261510789007,
58
+ -0.1,
59
+ 0.0,
60
+ 0.0,
61
+ 0.0
62
+ ],
63
+ "max": [
64
+ 0.2,
65
+ 0.1,
66
+ 0.15,
67
+ 0.049999999999999996,
68
+ 0.049999999999999996
69
+ ],
70
+ "mean": [
71
+ 0.13267758704419044,
72
+ 0.012846030641390576,
73
+ 0.082819576669988,
74
+ 0.02026785714285731,
75
+ 0.02026785714285731
76
+ ],
77
+ "std": [
78
+ 0.07797911694586931,
79
+ 0.06348108309324098,
80
+ 0.03931571746563297,
81
+ 0.02454805132754095,
82
+ 0.02454805132754095
83
+ ],
84
+ "count": [
85
+ 1120
86
+ ]
87
+ },
88
+ "supervision.object_pose": {
89
+ "min": [
90
+ -0.014122820547901486,
91
+ -0.10000049289125487,
92
+ 0.034992227756181465,
93
+ 0.9999670474962795,
94
+ -0.0002752555201473424,
95
+ -6.818784791346758e-05,
96
+ -0.008117067846005913
97
+ ],
98
+ "max": [
99
+ 0.19996388437432028,
100
+ 0.0999999105858684,
101
+ 0.19670165071978118,
102
+ 1.0000000000000007,
103
+ 8.588243669944473e-06,
104
+ 0.0017382338128796209,
105
+ 4.585633436010687e-06
106
+ ],
107
+ "mean": [
108
+ 0.1297852103994433,
109
+ 0.012517312912129845,
110
+ 0.08346298002947035,
111
+ 0.9999954433437799,
112
+ 2.99078089925559e-08,
113
+ 0.0001255653891815535,
114
+ -0.0011098124579117807
115
+ ],
116
+ "std": [
117
+ 0.07936202254298606,
118
+ 0.06299715854546376,
119
+ 0.04908708526513718,
120
+ 1.1200407056436107e-05,
121
+ 1.1772460664899637e-05,
122
+ 0.0004066817577303082,
123
+ 0.002774921121205961
124
+ ],
125
+ "count": [
126
+ 1120
127
+ ]
128
+ },
129
+ "timestamp": {
130
+ "min": [
131
+ 0.0
132
+ ],
133
+ "max": [
134
+ 6.95
135
+ ],
136
+ "mean": [
137
+ 3.475
138
+ ],
139
+ "std": [
140
+ 2.0206743923749815
141
+ ],
142
+ "count": [
143
+ 1120
144
+ ]
145
+ },
146
+ "frame_index": {
147
+ "min": [
148
+ 0.0
149
+ ],
150
+ "max": [
151
+ 139.0
152
+ ],
153
+ "mean": [
154
+ 69.5
155
+ ],
156
+ "std": [
157
+ 40.413487847499624
158
+ ],
159
+ "count": [
160
+ 1120
161
+ ]
162
+ },
163
+ "episode_index": {
164
+ "min": [
165
+ 0.0
166
+ ],
167
+ "max": [
168
+ 7.0
169
+ ],
170
+ "mean": [
171
+ 3.5
172
+ ],
173
+ "std": [
174
+ 2.29128784747792
175
+ ],
176
+ "count": [
177
+ 1120
178
+ ]
179
+ },
180
+ "index": {
181
+ "min": [
182
+ 0.0
183
+ ],
184
+ "max": [
185
+ 1119.0
186
+ ],
187
+ "mean": [
188
+ 559.5
189
+ ],
190
+ "std": [
191
+ 323.3160218733368
192
+ ],
193
+ "count": [
194
+ 1120
195
+ ]
196
+ },
197
+ "task_index": {
198
+ "min": [
199
+ 0.0
200
+ ],
201
+ "max": [
202
+ 2.0
203
+ ],
204
+ "mean": [
205
+ 0.75
206
+ ],
207
+ "std": [
208
+ 0.82915619758885
209
+ ],
210
+ "count": [
211
+ 1120
212
+ ]
213
+ },
214
+ "annotation.failure": {
215
+ "min": [
216
+ 0.0
217
+ ],
218
+ "max": [
219
+ 1.0
220
+ ],
221
+ "mean": [
222
+ 0.026785714285714284
223
+ ],
224
+ "std": [
225
+ 0.16145661892879573
226
+ ],
227
+ "count": [
228
+ 1120
229
+ ]
230
+ },
231
+ "annotation.success": {
232
+ "min": [
233
+ 0.0
234
+ ],
235
+ "max": [
236
+ 1.0
237
+ ],
238
+ "mean": [
239
+ 0.48928571428571427
240
+ ],
241
+ "std": [
242
+ 0.49988519090050326
243
+ ],
244
+ "count": [
245
+ 1120
246
+ ]
247
+ },
248
+ "annotation.valid_recovery_pair": {
249
+ "min": [
250
+ 1.0
251
+ ],
252
+ "max": [
253
+ 1.0
254
+ ],
255
+ "mean": [
256
+ 1.0
257
+ ],
258
+ "std": [
259
+ 0.0
260
+ ],
261
+ "count": [
262
+ 1120
263
+ ]
264
+ },
265
+ "privileged.simulator_state": {
266
+ "min": [
267
+ -0.014122764021777623,
268
+ -0.1000000225028425,
269
+ -0.0002008240793322001,
270
+ -0.0017723264821356429,
271
+ -0.0014710312643219869,
272
+ -0.014122820547901486,
273
+ -0.10000049289125487,
274
+ 0.034992227756181465,
275
+ 0.9999670474962795,
276
+ -0.0002752555201473424,
277
+ -6.818784791346758e-05,
278
+ -0.008117067846005913,
279
+ -0.09362799765317938,
280
+ -0.2233885095369417,
281
+ -0.3456166325545538,
282
+ -0.05734092435609014,
283
+ -0.055497768418793066,
284
+ -0.0004169627255277605,
285
+ -0.25941308900882726,
286
+ -1.4725698314433182,
287
+ -0.012511614894423984,
288
+ -0.01304715267129018,
289
+ -0.0004697904868500916
290
+ ],
291
+ "max": [
292
+ 0.19999999999999982,
293
+ 0.09999999999999991,
294
+ 0.1493811324297467,
295
+ 0.04111885307612435,
296
+ 0.04107339680984956,
297
+ 0.19996388437432028,
298
+ 0.0999999105858684,
299
+ 0.19670165071978118,
300
+ 1.0000000000000007,
301
+ 8.588243669944473e-06,
302
+ 0.0017382338128796209,
303
+ 4.585633436010687e-06,
304
+ 0.35915696105202416,
305
+ 0.22226931286421447,
306
+ 0.34602697011587097,
307
+ 0.25920105647017433,
308
+ 0.2581810315060028,
309
+ 0.3514102324227401,
310
+ 0.2611663378332658,
311
+ 0.33753565970506433,
312
+ 0.001729835460969292,
313
+ 0.013249164338083569,
314
+ 0.0013111168364806443
315
+ ],
316
+ "mean": [
317
+ 0.13069622870961353,
318
+ 0.012550514504649833,
319
+ 0.08184252925113868,
320
+ 0.016194231889651797,
321
+ 0.016217433140424682,
322
+ 0.1297852103994433,
323
+ 0.012517312912129845,
324
+ 0.08346298002947035,
325
+ 0.9999954433437799,
326
+ 2.99078089925559e-08,
327
+ 0.0001255653891815535,
328
+ -0.0011098124579117807,
329
+ 0.017014304812024473,
330
+ 0.0026307121559819677,
331
+ 0.006947751533805151,
332
+ 0.003171810917318101,
333
+ 0.003905762686124316,
334
+ 0.016878971149232716,
335
+ 0.003433421170341602,
336
+ -0.010084862622635273,
337
+ -4.2124046187982724e-05,
338
+ -0.00016220849227443606,
339
+ 9.259036285258952e-06
340
+ ],
341
+ "std": [
342
+ 0.0787665187274976,
343
+ 0.06294078291544838,
344
+ 0.0394299427641389,
345
+ 0.019663202307695832,
346
+ 0.019685423812493552,
347
+ 0.07936202254298606,
348
+ 0.06299715854546376,
349
+ 0.04908708526513718,
350
+ 1.1200407056436107e-05,
351
+ 1.1772460664899637e-05,
352
+ 0.0004066817577303082,
353
+ 0.002774921121205961,
354
+ 0.06620288861031719,
355
+ 0.037415485652098034,
356
+ 0.11736775749429738,
357
+ 0.03152439092122224,
358
+ 0.03134431841053177,
359
+ 0.06454567696123781,
360
+ 0.04399064602885924,
361
+ 0.1801548186321963,
362
+ 0.0006579569781627707,
363
+ 0.0011615174305008684,
364
+ 0.00012450879113138792
365
+ ],
366
+ "count": [
367
+ 1120
368
+ ]
369
+ },
370
+ "observation.images.front_observed": {
371
+ "min": [
372
+ [
373
+ [
374
+ 0.0
375
+ ]
376
+ ],
377
+ [
378
+ [
379
+ 0.0
380
+ ]
381
+ ],
382
+ [
383
+ [
384
+ 0.0
385
+ ]
386
+ ]
387
+ ],
388
+ "max": [
389
+ [
390
+ [
391
+ 1.0
392
+ ]
393
+ ],
394
+ [
395
+ [
396
+ 1.0
397
+ ]
398
+ ],
399
+ [
400
+ [
401
+ 1.0
402
+ ]
403
+ ]
404
+ ],
405
+ "mean": [
406
+ [
407
+ [
408
+ 0.20017190071344537
409
+ ]
410
+ ],
411
+ [
412
+ [
413
+ 0.2325016381813311
414
+ ]
415
+ ],
416
+ [
417
+ [
418
+ 0.2669429979283086
419
+ ]
420
+ ]
421
+ ],
422
+ "std": [
423
+ [
424
+ [
425
+ 0.1472167387756444
426
+ ]
427
+ ],
428
+ [
429
+ [
430
+ 0.16675662183305479
431
+ ]
432
+ ],
433
+ [
434
+ [
435
+ 0.18073227383659632
436
+ ]
437
+ ]
438
+ ],
439
+ "count": [
440
+ 1120
441
+ ]
442
+ }
443
+ }
dataset/meta/tasks.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:96ba9f9df6e418ff9300fb43667eb81d9b1a9e00ab3347ad8982891100b3da4a
3
+ size 2402
dataset/videos/observation.images.front_observed/chunk-000/file-000.mp4 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:73738e005031fd35c3122fc637f58f2d75bc7ef77e553f384acfc33dc11e0428
3
+ size 4834892
episodes/episode_000000/depth/front/000000.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c79c84fde10f92ecc434a49a816b820d1db91a75c6a585acfd855846d885d558
3
+ size 76928
episodes/episode_000000/depth/front/000000.png ADDED

Git LFS Details

  • SHA256: b41d8d172e3b16929433b8fce5d3e1f9bc5ae7fd9fb91fef888c64a6e4c13e43
  • Pointer size: 129 Bytes
  • Size of remote file: 1.56 kB
episodes/episode_000000/depth/front/000001.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:036964b4f371a06d0c351aa391f23da931ed5074aef82952c0f7e5725012f28e
3
+ size 76928
episodes/episode_000000/depth/front/000001.png ADDED

Git LFS Details

  • SHA256: 9fb6abb05b709fe8ece8bc652031bcb70ab1746a588589bd458145b2f3d3722b
  • Pointer size: 129 Bytes
  • Size of remote file: 1.45 kB
episodes/episode_000000/depth/front/000002.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:043a7d21af2f1674ee817bded9b68fa6e6bcc8d8b2919ad52ecfa2affa1ff6a8
3
+ size 76928
episodes/episode_000000/depth/front/000002.png ADDED

Git LFS Details

  • SHA256: 132a7434dfdf2186e42a30d67a92818ff3538564d7e823f18f7938aa50ee77ce
  • Pointer size: 129 Bytes
  • Size of remote file: 1.46 kB
episodes/episode_000000/depth/front/000003.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ed2336ee46dfe7ef86800a22fcb0ecb288222c4acc73331e22ccf0049e8b8cd5
3
+ size 76928
episodes/episode_000000/depth/front/000003.png ADDED

Git LFS Details

  • SHA256: 132a7434dfdf2186e42a30d67a92818ff3538564d7e823f18f7938aa50ee77ce
  • Pointer size: 129 Bytes
  • Size of remote file: 1.46 kB
episodes/episode_000000/depth/front/000004.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:db7b3036280bae48f9566db7aaba9a7e48e50ccb9715ce36c1fea9569f4b21bf
3
+ size 76928
episodes/episode_000000/depth/front/000004.png ADDED

Git LFS Details

  • SHA256: 132a7434dfdf2186e42a30d67a92818ff3538564d7e823f18f7938aa50ee77ce
  • Pointer size: 129 Bytes
  • Size of remote file: 1.46 kB
episodes/episode_000000/depth/front/000005.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e03dc9f880e672d4520430c7444648bc7e98d1edb474aabdf414e828aca16a34
3
+ size 76928
episodes/episode_000000/depth/front/000005.png ADDED

Git LFS Details

  • SHA256: b753d2ca5753f1d92138b8aac6e9e4e5b3abe19e440bf8562a68f863dd3f70e0
  • Pointer size: 129 Bytes
  • Size of remote file: 1.46 kB
episodes/episode_000000/depth/front/000006.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9e6f563ccd1fe8c22b742cd7f38dbf08d368001e60402973ea5178ae26adb690
3
+ size 76928
episodes/episode_000000/depth/front/000006.png ADDED

Git LFS Details

  • SHA256: 94dcaafe9a590a21aa7dd27884a10cc315dd073de0d76e2cede101ece0b41602
  • Pointer size: 129 Bytes
  • Size of remote file: 1.46 kB
episodes/episode_000000/depth/front/000007.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4715d401a01b2ee89e170d8a0672208a03015ceed8d9a1da286957ecaa66c795
3
+ size 76928
episodes/episode_000000/depth/front/000007.png ADDED

Git LFS Details

  • SHA256: 9242ef445f02c403939d6a0c03acb6df116f73473d892761473f49ce7fd2caec
  • Pointer size: 129 Bytes
  • Size of remote file: 1.45 kB
episodes/episode_000000/depth/front/000008.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4cee9cd918448cc208cc1e6ec186d7d8f0fb8288fd7ab66a5b975f1aa0e25528
3
+ size 76928
episodes/episode_000000/depth/front/000008.png ADDED

Git LFS Details

  • SHA256: 2dedfaa985ec2d771aead73a8f9300afd72e1518576c2c13c33c7c94a82588af
  • Pointer size: 129 Bytes
  • Size of remote file: 1.45 kB
episodes/episode_000000/depth/front/000009.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3d4bdecaf93980d35efe0c5faa14ef10b373bf8ff09a29d7c1430f1af1205a3a
3
+ size 76928
episodes/episode_000000/depth/front/000009.png ADDED

Git LFS Details

  • SHA256: 0751b09bd5d579391610e94aad0aaa40a770b0bdc56e8961881432c8e4e123d6
  • Pointer size: 129 Bytes
  • Size of remote file: 1.44 kB
episodes/episode_000000/depth/front/000010.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2903aab757852f58c0b3c3c2292d5ef107e8158ad2cd3778a6a547e1920bf228
3
+ size 76928
episodes/episode_000000/depth/front/000010.png ADDED

Git LFS Details

  • SHA256: fb0054021fc2eb8d2204b039830a79054c84c0e3eca072c0e1e02e89bcf87ebc
  • Pointer size: 129 Bytes
  • Size of remote file: 1.47 kB
episodes/episode_000000/depth/front/000011.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ebb2f02393854558b46d986480e7c4362e8deee9bf2524e4452ab4deaf4cab41
3
+ size 76928
episodes/episode_000000/depth/front/000011.png ADDED

Git LFS Details

  • SHA256: aff90811526bea670c9c91affe3e848a2933e4f4b065260c4daf0ee2a7d45494
  • Pointer size: 129 Bytes
  • Size of remote file: 1.49 kB
episodes/episode_000000/depth/front/000012.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f5327d765f0f3ed6cfcb2b615c776e56459e765e80745b3afd0ba30e32007804
3
+ size 76928
episodes/episode_000000/depth/front/000012.png ADDED

Git LFS Details

  • SHA256: 48cc875dac064c626fa67927506bc3c97e770daecfce8053c6884a23d1ce628b
  • Pointer size: 129 Bytes
  • Size of remote file: 1.49 kB
episodes/episode_000000/depth/front/000013.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9d120d225194ef1da9240b2e0e2b4378231461e8cc3c33bf370ae451d89583f9
3
+ size 76928
episodes/episode_000000/depth/front/000013.png ADDED

Git LFS Details

  • SHA256: 005344116968229e73fe041e13666b717a62a6da84c301cdbebf9e1892943b52
  • Pointer size: 129 Bytes
  • Size of remote file: 1.49 kB