File size: 4,861 Bytes
a79330b
4f220a4
a79330b
 
 
 
 
 
 
 
 
 
 
 
 
4f220a4
7e38625
 
a79330b
4f220a4
a79330b
4f220a4
 
 
 
803c4cb
7e38625
a79330b
4f220a4
7e38625
4f220a4
 
a79330b
7e38625
 
 
 
 
4f220a4
a79330b
4f220a4
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
803c4cb
 
 
 
 
 
 
 
 
a79330b
 
7e38625
 
 
 
 
 
 
 
 
 
 
 
 
 
 
4f220a4
803c4cb
4f220a4
 
a79330b
4f220a4
a79330b
4f220a4
 
a79330b
803c4cb
4f220a4
 
 
 
 
 
 
a79330b
4f220a4
a79330b
 
803c4cb
4f220a4
a79330b
 
4f220a4
803c4cb
 
a79330b
4f220a4
a79330b
4f220a4
 
 
 
a79330b
4f220a4
 
a79330b
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
---
pretty_name: CROSS navigation debug cases
language:
  - en
tags:
  - robotics
  - navigation
  - vggt
  - rgb-d
  - odometry
  - debugging
---

# CROSS navigation debug

Recorded Stretch navigation cases using CROSS, its VGGT navigation heads and a local robot adapter.
Supervised runs were recorded on **2026-10-26 and 2026-10-27 (Asia/Singapore)** toward reference-map
keyframe **1012**, using the Record3D map `2026-10-22--15-48-58`.

## Cases

| Case | Main finding | RGB/depth pairs | VGGT passes | Details |
|---|---|---:|---:|---|
| **20261026_goal1012_02** | Measured-depth checks reject all local graph attachments during a head survey | 680 | 380 | [Evidence, download and replay](cases/20261026_goal1012_02/README.md) |
| **20261026_goal1012_03** | Transient false BEV occupancy triggers the added path guard; a separate braking handoff bug stops the survey | 500 | 240 | [Evidence, download and replay](cases/20261026_goal1012_03/README.md) |
| **20261026_goal1012_05** | False forward-left occupancy despite survey floor evidence; historical-reference lifting ablation; complete head survey followed by a separate depth-attachment stop | 749 | 376 | [Evidence, download and replay](cases/20261026_goal1012_05/README.md) |
| **20261027_goal1012_01** | False occupancy on currently measured clear floor; historical-view head/context controls; leftward goals independently explained by graph lookahead | 1,061 | 34 archived / 612 total | [Evidence, training notes and replay](cases_oct27/20261027_goal1012_01/README.md) |

Each case is self-contained, including original RGB/depth, odometry and image-time transforms, exact
VGGT input tensors and numeric outputs for its documented scope, BEV replay, planner evidence, the reference map, recorded
source and offline inspection tools. The case reports distinguish native CROSS behavior from the
local adapter's checks. See each case's README for findings, limitations and reproduction commands.


The October 27 collection is under [`cases_oct27/`](cases_oct27/README.md), with the same self-contained
case layout. Its first case retains the full sensor/replay sequence and a focused raw-model window;
the distinction between source-run totals and uploaded model data is explicit in its manifest.

## Repository layout

```text
cases/
  20261026_goal1012_02/
    README.md
    archives/
    code/
    environment/
    manifest.json
    previews/
    reports/
    tools/
  20261026_goal1012_03/
    README.md
    archives/
    code/
    environment/
    manifest.json
    previews/
    reports/
    tools/
    post_run_fixes/        # later changes, explicitly separate from recorded source
  20261026_goal1012_05/
    README.md
    archives/
    code/
    environment/
    manifest.json
    previews/
    reports/
    tools/
```

```text
cases_oct27/
  README.md
  20261027_goal1012_01/     # full sensor/replay sequence, focused raw-model window
    README.md
    training_notes.json
    archives/
    code/
    environment/
    manifest.json
    reports/
    tools/
    post_run_fixes/
```

Archives always belong to their containing case. Archive/member SHA-256 hashes and extraction paths
are in each case's `manifest.json`. Cases 03 and 05 also include file-level `SHA256SUMS`.
The cases include identical reference-map/source archives so either can be downloaded independently.
Pretrained model weights are referenced by repository/revision/hash in the case provenance.

## Download one case

```python
from huggingface_hub import snapshot_download

case_id = "20261026_goal1012_05"  # or "20261026_goal1012_02" / "20261026_goal1012_03"
snapshot_download(
    repo_id="jizhuochen/CROSS_nav_debug",
    repo_type="dataset",
    allow_patterns=f"cases/{case_id}/**",
    local_dir="CROSS_nav_debug",
)
```

Change into the downloaded case folder, then run its verifier/unpacker:

```bash
cd CROSS_nav_debug/cases/20261026_goal1012_05
python tools/unpack.py --output extracted
```

Case 02 needs approximately **3.5 GB** for download plus **3.5 GB** for extraction; case 03 needs
approximately **2.3 GB** for each; case 05 needs approximately **3.6 GB** to download and **3.7 GB** to extract.
Replay and saved-output inspection need no robot or GPU.

## Layout migration

The original case 02 files formerly at the repository root (`archives/`, `code/`, `reports/`, etc.)
now live under `cases/20261026_goal1012_02/`. Its formerly nested report directory is now simply
`reports/` inside the case. Recorded payloads are unchanged. Historical revisions retain the old
paths; use the case folders and their current README instructions for new downloads.

Sharing copies have credentials and private machine-specific text removed. Numerical sensor data,
transforms and model predictions are preserved. Source retains its included upstream licenses;
no additional recording license is asserted here.