card: document the recorder timestamp artefact (ur_joint_states +0.900 s, tcp_pose/wrench ~+0.41 s; per-take tau; fix rule); depth now opt-in in the recorder

DATA_DICTIONARY.md

@@ -19,6 +19,9 @@ version see the LeRobot dataset
 - Every figure in this document was measured from the files themselves by
   `make_carrot_raw_stats.py`. The machine-readable version is
   [`dataset_stats.json`](dataset_stats.json).
+- ⚠️ **Read §3.6 before you align streams.** Three of the nine groups — `ur_joint_states`,
+  `tcp_pose` and `wrench` — carry timestamps that are **late by a constant** (0.900 s and ≈0.41 s).
+  It is a recorder artefact, it is a pure delay, and it is **not** corrected in these files.
 
 ---
 
@@ -69,10 +72,10 @@ dataset = seconds since take start. All datasets are **1-D `float64`**, one arra
 |---|--:|---|---|
 | `cam1_frames` | 18,557 | 30.0 Hz | Timestamp + frame index for each `cam1.mp4` frame |
 | `cam2_frames` | 18,557 | 30.0 Hz | Timestamp + frame index for each `cam2.mp4` frame |
-| `command` | 40,640 | **~67 Hz** | Commanded UR7e joint targets (the **action**), radians |
-| `ur_joint_states` | 40,629 | **~67 Hz** | Measured UR7e joint state: position, velocity, effort |
-| `tcp_pose` | 40,619 | **~67 Hz** | Measured tool-center-point pose (position + quaternion) |
-| `wrench` | 40,621 | **~67 Hz** | 6-axis force/torque at the TCP |
+| `command` | 40,640 | **~67 Hz** | Commanded UR7e joint targets (the **action**), radians. `t_rel_s` **fresh** |
+| `ur_joint_states` | 40,629 | **~67 Hz** | Measured UR7e joint state: position, velocity, effort. ⚠️ `t_rel_s` **0.900 s late** (§3.6) |
+| `tcp_pose` | 40,619 | **~67 Hz** | Measured tool-center-point pose (position + quaternion). ⚠️ `t_rel_s` **≈0.41 s late** (§3.6) |
+| `wrench` | 40,621 | **~67 Hz** | 6-axis force/torque at the TCP. ⚠️ `t_rel_s` **≈0.41 s late** (§3.6) |
 | `gripper` | 24,015 | **~38.8 Hz** | Gripper command, measured position, leader trigger |
 | `gello_joint_states` | 18,573 | 30.0 Hz | GELLO **leader** joint pos + vel (teleop only — see §3.4) |
 | `synchronized` | **0 (EMPTY)** | — | Intended fused/aligned table; **empty in all 54 takes** — ignore |
@@ -143,7 +146,7 @@ All datasets `float64`, shape `(N,)` where `N` = that group's row count for the
 #### `command` — commanded UR7e joint targets ➜ **the ACTION**
 | Channel | dtype | Unit | Meaning | min | max | mean |
 |---|---|---|---|--:|--:|--:|
-| `t_rel_s` | float64 | s | Timestamp | — | — | — |
+| `t_rel_s` | float64 | s | Timestamp — **fresh** (this table did not queue; §3.6) | — | — | — |
 | `cmd1` | float64 | rad | Target angle, joint 1 (base) | −3.7024 | −2.6804 | −3.1542 |
 | `cmd2` | float64 | rad | Target angle, joint 2 (shoulder) | −1.7302 | −0.9591 | −1.2976 |
 | `cmd3` | float64 | rad | Target angle, joint 3 (elbow) | 1.2936 | 2.0710 | 1.7272 |
@@ -154,7 +157,7 @@ All datasets `float64`, shape `(N,)` where `N` = that group's row count for the
 #### `ur_joint_states` — measured UR7e joint state ➜ core of **observation.state**
 | Channel | dtype | Unit | Meaning | min | max | mean |
 |---|---|---|---|--:|--:|--:|
-| `t_rel_s` | float64 | s | Timestamp | — | — | — |
+| `t_rel_s` | float64 | s | Timestamp — ⚠️ **stamped 0.900 s late; subtract it (§3.6)** | — | — | — |
 | `q1` | float64 | rad | Measured angle, joint 1 | −3.7028 | −2.6809 | −3.1325 |
 | `q2` | float64 | rad | Measured angle, joint 2 | −1.7284 | −0.9593 | −1.3023 |
 | `q3` | float64 | rad | Measured angle, joint 3 | 1.3382 | 2.0708 | 1.7311 |
@@ -176,14 +179,23 @@ All datasets `float64`, shape `(N,)` where `N` = that group's row count for the
 
 **Follower tracking.** The median per-take `|cmd_i − q_i|` (nearest-timestamp aligned onto the
 command clock) is **0.023 / 0.061 / 0.033 / 0.080 / 0.029 / 0.031 rad** for joints 1–6, with a
-worst single sample of **0.510 rad** (joint 6). This is ordinary servo following error on a
-hand-driven teleop stream; it is *not* a defect, but it means `command` is a **target**, not a
-measurement — if you need where the arm actually was, use `ur_joint_states`.
+worst single sample of **0.510 rad** (joint 6).
+
+> ⚠️ **Those figures are computed on the raw timestamps, and most of that gap is the §3.6 timestamp
+> artefact, not tracking error.** Aligned on the *corrected* clock (`ur_joint_states/t_rel_s`
+> − 0.900 s) the same residual collapses to **0.0007–0.0103 rad** (median 0.0062) — and to
+> 0.0004–0.0009 rad on the robust objective of §3.6. What is left there *is* the genuine servo
+> following error, and it lives in the high-velocity transients. The uncorrected numbers are kept
+> above because they describe the files as they are shipped, which is what a consumer who skips
+> §3.6 will see.
+
+`command` is a **target**, not a measurement — if you need where the arm actually was, use
+`ur_joint_states` (on its corrected clock).
 
 #### `tcp_pose` — measured tool-center-point pose (robot base frame)
 | Channel | dtype | Unit | Meaning | min | max | mean |
 |---|---|---|---|--:|--:|--:|
-| `t_rel_s` | float64 | s | Timestamp | — | — | — |
+| `t_rel_s` | float64 | s | Timestamp — ⚠️ **stamped ≈0.41 s late; subtract it (§3.6)** | — | — | — |
 | `x` | float64 | m | TCP position x in robot base frame | 0.4148 | 0.7095 | 0.5674 |
 | `y` | float64 | m | TCP position y | −0.2141 | 0.3900 | 0.1307 |
 | `z` | float64 | m | TCP position z | 0.0175 | 0.3066 | 0.1424 |
@@ -199,7 +211,7 @@ pick-and-place. The z floor of **0.0175 m** is the fingers reaching the table to
 #### `wrench` — 6-axis force/torque at TCP
 | Channel | dtype | Unit | Meaning | min | max | mean |
 |---|---|---|---|--:|--:|--:|
-| `t_rel_s` | float64 | s | Timestamp | — | — | — |
+| `t_rel_s` | float64 | s | Timestamp — ⚠️ **stamped ≈0.41 s late; subtract it (§3.6)** | — | — | — |
 | `fx` | float64 | N | Force along x | −18.711 | 36.708 | −1.179 |
 | `fy` | float64 | N | Force along y | −20.573 | 35.621 | −0.536 |
 | `fz` | float64 | N | Force along z | −114.883 | 7.142 | 0.987 |
@@ -301,6 +313,134 @@ tcp_x, tcp_y, tcp_z, tcp_qx, tcp_qy, tcp_qz, tcp_qw,
 cam1_frame_idx, cam2_frame_idx
 ```
 
+### 3.6 ⚠️ Recorder timestamp artefact — three groups are stamped **late**
+
+**The arm was not late; the recording is.** In this release the `t_rel_s` of three groups is behind
+the rest of the take by a *constant*: their rows describe the robot as it was a fraction of a second
+**earlier** than they claim. Nothing else in the file is affected, and nothing here has been
+rewritten.
+
+| group | stamped late by | per-take spread | how it was measured |
+|---|--:|---|---|
+| `ur_joint_states` | **0.900 s** | **0.895 – 0.900 s**, median **0.900 s**; most takes land exactly on 0.900 and the rest one 5 ms grid step below (38 of 54 at 0.900 in this release's own measurement, 41 of 54 in an independent re-measurement — same lag, different tie-break) | joint-space residual: shift the `ur_joint_states` clock by −τ, linearly interpolate `q1..q6` onto the `command` timestamps, take the τ on a **5 ms grid** over 0–1.5 s that minimises mean \|ur_q − cmd\| |
+| `tcp_pose` | **≈ 0.41 s** (= 0.900 − 0.495; speed cross-correlation against `command` independently gives 0.41. Its *absolute* queue age is ≈0.50 s — the difference is `command`'s own residual ≈0.08 s) | **0.495 – 0.500 s** *ahead of* `ur_joint_states` (median 0.495) — exactly the half-queue prediction | forward-kinematics residual: `fk(q) ⊕ 0.174 m` against `tcp_pose`, **11.2 – 23.3 mm** at zero shift and **0.52 – 2.89 mm** at the optimum |
+| `wrench` | **≈ 0.41 s** | measured **0.000 s** from `tcp_pose` in 7 of 8 takes | velocity cross-correlation; `wrench` shares `tcp_pose`'s QoS depth (50) and publisher, so it carries the same age |
+
+**Fresh — do not shift these:** `cam1_frames`, `cam2_frames`, both `depth.h5` streams,
+`gello_joint_states`, `gripper`, and `command`. ⚠️ For all of them but `command` that is the queue
+model *and* the measurement; for **`command` it is the measurement alone** — it is published at 250 Hz behind a
+depth-50 queue, so the model predicts ~0.20 s, while the scene-camera cross-correlation (evidence 2
+below) puts it at τ ≈ 0 and the queue algebra at ≈0.08 s. Treat `command` as fresh to within
+≈0.08 s — under three camera frames — not to zero.
+
+Three independent observations make "the robot was slow" impossible and pin the blame on the
+recorder:
+
+1. **Depth frames carry the driver's own `header.stamp`** (`depth.h5/<cam>/stamp_s`, §5), and their
+   measured age at the moment they were written is **0.022 s**. The images are fresh; that is the
+   absolute reference point the other streams are compared against.
+2. **The scene camera sees the arm move when `command` says it moved.** Frame-to-frame motion energy
+   in `cam1.mp4` correlates with the `command` TCP speed at **τ ≈ 0** (−0.055 s and +0.010 s on the
+   two takes checked) and with `ur_joint_states` at **+0.865 / +0.935 s**. If the *arm* had lagged,
+   the video would have lagged with it.
+3. **`tcp_pose` and `wrench` are 0.000 s apart.** Both are *measurements* published in the same
+   `controller_manager` cycle from the same RTDE packet, so their physical time difference is zero —
+   and that is exactly what is measured. Meanwhile `tcp_pose` → `ur_joint_states`, also two
+   measurements from that same packet, sit **0.495–0.500 s** apart (median 0.495, measured on all
+   54 takes by forward-kinematics residual; the sibling velocity-correlation analysis independently
+   reports 0.495 s, and the release script's speed cross-correlation 0.465 s). Two streams out of
+   one packet cannot physically disagree; only their *stamps* can.
+
+**Mechanism.** The recorder writes each row inside its ROS subscription callback and stamps it with
+the time that callback *ran* — message header stamps were not stored — and every callback shares one
+rclpy spin thread, which a `SingleThreadedExecutor` services one message per subscription per round.
+Adding depth recording to this session (≈5.3 MB/s of HDF5 writes plus four extra 30 Hz callbacks on
+that same thread) dropped the round rate from ~98–100 Hz to **≈67 Hz — 41–70 Hz across takes
+(§3.1)**, so every publisher faster than that kept its `KEEP_LAST` queue permanently full and each
+row emerged already `queue_depth ÷ publish_rate` old:
+
+| topic | QoS depth | publish rate | predicted age | measured |
+|---|--:|--:|--:|--:|
+| `/joint_states` → `ur_joint_states` | **100** | ~100 Hz | **1.00 s** | 0.900 s vs `command` |
+| `/tcp_pose_broadcaster/pose` → `tcp_pose` | **50** | ~100 Hz | **0.50 s** | **0.495 s** *ahead of* `ur_joint_states` — half the queue, half the age |
+| `/force_torque_sensor_broadcaster/wrench` → `wrench` | **50** | ~100 Hz | **0.50 s** | 0.000 s from `tcp_pose` |
+| `/forward_position_controller/commands` → `command` | 50 | **250 Hz** (`publish_rate_hz`) | 0.20 s | **≈0.08 s** — the one row where the model over-predicts (see the *Fresh* note above) |
+| `/gello/joint_states`, gripper, colour, depth | 10–50 | 30–40 Hz | **~0** | 0.022 s (depth, header-stamped) |
+
+The **starvation signature is visible in §3.1**: `command`, `ur_joint_states`, `tcp_pose` and
+`wrench` have two *different* publish rates — 250 Hz for `command`, ~100 Hz for the three
+`controller_manager` broadcasters — yet converge on the same recorded rate to five
+decimals (`take_01`: 59.79218 / 59.79089 / 59.79282 / 59.79153 Hz). Streams that were *not* starved
+keep their own native rate (`gello_joint_states` 29.99–30.06 Hz, `cam1_frames` 29.96–30.01 Hz).
+
+**It is a pure delay, not a distortion.** The waveform survives the shift intact, which is both why
+one constant recovers it and how we know it is not servo lag — a real tracking lag is a low-pass and
+would smear the shape, so no shift could collapse the residual like this:
+
+| residual | uncorrected (τ = 0) | at the per-take optimum |
+|---|--:|--:|
+| `ur_joint_states` vs `command`, mean \|Δq\| over 6 joints | **0.049 – 0.090 rad** (median 0.063) | **0.0007 – 0.0103 rad** (median 0.0062) |
+| same, on the release script's median-over-time objective | 0.034 – 0.081 rad | **0.0004 – 0.0009 rad** |
+| `tcp_pose` vs `fk(q) ⊕ 0.174 m` | **11.2 – 23.3 mm** (median 17.6) | **0.52 – 2.89 mm** (median 1.61) |
+
+⚠️ **Quote the lag with its method attached.** Four estimators land within ±0.02 s of one another
+and none of them is wrong; published numbers that look contradictory are the same measurement:
+
+| estimator | grid | result |
+|---|---|--:|
+| joint-space residual, mean \|ur_q − cmd\| over 6 joints | 5 ms | **0.895 – 0.900 s** (median 0.900) — `dataset_stats.json` → `timestamp_lag.ur_joint_states_lag_s_mean_objective` |
+| same, median-over-time objective | 5 ms | **0.900 s in all 54 takes** — `dataset_stats.json` → `timestamp_lag.ur_joint_states_lag_s` |
+| velocity cross-correlation, `command` → `ur_joint_states` | 5 ms | **0.875 – 0.915 s** (median 0.900; the sibling analysis reports 0.910 s) |
+| TCP-position residual | 10 ms | **0.90 s** in 11 of 11 takes checked (residual median 0.42 – 0.64 mm, against 44 – 67 mm at τ = 0) |
+
+So: **0.89 – 0.91 s depending on method and grid.** The derived LeRobot release applies the
+**per-take optimum** from the first row — **0.900 s** for 41 takes, **0.895 s** for 13 — and records
+it episode by episode. A single 0.900 s for everything is within one 5 ms grid step of that, i.e.
+a seventh of a camera frame, and is perfectly usable.
+
+The second-order lags in the session are worth knowing so you do not mistake one for the other:
+`gello_joint_states` → `command` is **+0.25 s** and is *real* — the leader→command latency of the
+One-Euro filter and slew limiter, measured at +0.253 s in the joint-mode sibling session where
+nothing was starved and +0.260 s here. `grip_cmd` → `grip_pos` is **+0.37 s** and is also real: the
+Robotiq's open/close time (+0.34 s in the sibling). Neither is a stamping artefact and neither
+should be corrected.
+
+**Fix rule for consumers**
+
+```python
+import h5py, numpy as np
+
+UR_LAG_S, TCP_LAG_S = 0.900, 0.405         # 0.405 = 0.900 - 0.495; wrench uses TCP_LAG_S too
+with h5py.File("take_01_20260914_164811/vectors.h5", "r") as f:
+    cam_t = f["cam1_frames"]["t_rel_s"][:]                      # master clock, already correct
+    ur_t  = f["ur_joint_states"]["t_rel_s"][:] - UR_LAG_S       # <- the whole fix
+    ur_q  = np.stack([f["ur_joint_states"][f"q{i}"][:] for i in range(1, 7)], 1)
+ur_on_cam = np.stack([np.interp(cam_t, ur_t, ur_q[:, k]) for k in range(6)], 1)
+```
+
+**Do not shift anything else.** `command`, `gripper`, `gello_joint_states`, the colour frames and
+the depth frames already share one correct timebase; applying the same offset to them would *create*
+a 0.9 s misalignment that is not in the data — including the two real lags listed above, which are
+physics, not stamping.
+
+**What this means for the derived release.** The LeRobot conversion builds `observation.state[0:6]`
+from `ur_joint_states`, so it shifts that table by that take's measured τ (**−0.900 s** for 41
+takes, **−0.895 s** for 13) *before* resampling onto the 30 fps
+camera grid; `grip_pos`, `action` (`command` + `grip_cmd`), the RGB videos and both depth streams are
+resampled unchanged. The applied τ is recorded per episode there in `meta/source_takes.json`
+(`ur_joint_states_lag_s`). `tcp_pose` and `wrench` are **not** part of that release — take them from
+here and subtract ≈0.41 s yourself.
+
+**Status.** The raw files are published exactly as recorded: a documented constant is recoverable,
+a silently re-stamped file is not. Per-take values, both residual definitions and the
+cross-correlation checks are in [`dataset_stats.json`](dataset_stats.json) under `timestamp_lag`.
+The recorder itself was **fixed after this release** — header stamps are now stored alongside
+`t_rel_s`, the high-rate queue depths were shrunk, camera/depth writes were moved off the spin
+thread, and it now warns when the recorded rates of unrelated topics converge (the signature above);
+**depth recording is now opt-in (`ENABLE_DEPTH=1`), default RGB-only**.
+Later sessions in this family are not affected; the sibling joint-mode releases
+`cube_in_cup_raw` / `banana_in_pot_raw`, recorded before depth existed, show ≈0.05 s.
+
 ---
 
 ## 4. Complete pooled value ranges
@@ -595,7 +735,8 @@ with h5py.File(path, "r") as f:
     cols = json.loads(f["ur_joint_states"].attrs["columns"])   # -> ['t_rel_s','q1',...]
 
     # measured UR7e joint positions (N_ur, 6), radians, on the UR clock (~67 Hz)
-    ur_t = f["ur_joint_states"]["t_rel_s"][:]
+    # NOTE: this clock is stamped 0.900 s LATE -- subtract it before aligning (§3.6)
+    ur_t = f["ur_joint_states"]["t_rel_s"][:] - 0.900
     ur_q = np.stack([f["ur_joint_states"][f"q{i}"][:] for i in range(1, 7)], axis=1)
 
     # commanded joint targets = the action (N_cmd, 6), ~67 Hz
@@ -625,6 +766,17 @@ frame `k` corresponds to `cam1_frames/frame_idx[k]`. Because cam1 and cam2 diffe
 
 ## 9. Anomalies & data-quality notes
 
+- ⚠️ **Timestamp artefact — `ur_joint_states`, `tcp_pose` and `wrench` rows are stamped late**
+  (§3.6). `ur_joint_states` by **0.900 s** (per-take optimum 0.895–0.900 s, median 0.900;
+  **0.89–0.91 s** depending on estimator and grid), `tcp_pose` and `wrench` by **≈0.41 s**
+  (= 0.900 − 0.495, the tightest measured quantity here). It is a recorder artefact — a
+  single spin thread, `t_rel_s` = callback execution time, and `KEEP_LAST` queues of depth 100/50
+  kept permanently full once depth recording lowered the service rate to ≈67 Hz (41–70 Hz) — **not** robot or
+  controller lag. `command`, both cameras, both depth streams, `gello_joint_states` and `gripper`
+  are fresh (depth header-stamp age **0.022 s**). It is a pure delay: subtract **0.900 s** from
+  `ur_joint_states/t_rel_s` and **≈0.41 s** from `tcp_pose`/`wrench`, and shift **nothing else**.
+  The files here are as recorded; the derived LeRobot release has the `ur_joint_states` correction
+  already applied; the recorder was fixed after this release.
 - **Take-number gap `{12}` is by design.** Folders run `take_01` … `take_55` but `take_12` was
   discarded during recording, giving exactly **54** folders. It is not missing data and there is
   no partial `take_12` anywhere in the release.
@@ -686,5 +838,6 @@ script records which method it used in `aggregate.video_frame_count_method`). It
 folders read-only and writes `dataset_stats.json`.
 
 See [`dataset_stats.json`](dataset_stats.json) for exact per-take frame counts, durations, row
-counts, byte sizes, rate statistics, depth quality samples, leader/follower statistics, and
-per-channel min/max/mean/std.
+counts, byte sizes, rate statistics, depth quality samples, leader/follower statistics,
+per-channel min/max/mean/std, and — under `timestamp_lag` — the per-take timestamp offsets of §3.6
+with both residual definitions and the cross-correlation cross-checks.

README.md

@@ -106,10 +106,10 @@ rate. All datasets are **1-D `float64`** in a columnar layout (channel `foo` is
 |---|---|--:|---|---|
 | `cam1_frames` | 30.0 Hz | 18,557 | `frame_idx`, `t_rel_s` | index/time of each cam1 MP4 frame |
 | `cam2_frames` | 30.0 Hz | 18,557 | `frame_idx`, `t_rel_s` | index/time of each cam2 MP4 frame |
-| `command` | **~67 Hz** | 40,640 | `cmd1..cmd6`, `t_rel_s` | commanded **absolute** UR joint targets (rad) |
-| `ur_joint_states` | **~67 Hz** | 40,629 | `q1..q6`, `qd1..qd6`, `eff1..eff6`, `t_rel_s` | UR7e **follower** joint positions (rad), velocities (rad/s), efforts |
-| `tcp_pose` | **~67 Hz** | 40,619 | `x,y,z`, `qw,qx,qy,qz`, `t_rel_s` | TCP pose in robot **base frame**: position (m) + quaternion |
-| `wrench` | **~67 Hz** | 40,621 | `fx,fy,fz`, `tx,ty,tz`, `t_rel_s` | 6-axis end-effector force (N) + torque (N·m) |
+| `command` | **~67 Hz** | 40,640 | `cmd1..cmd6`, `t_rel_s` | commanded **absolute** UR joint targets (rad). Timestamps are **fresh** |
+| `ur_joint_states` | **~67 Hz** | 40,629 | `q1..q6`, `qd1..qd6`, `eff1..eff6`, `t_rel_s` | UR7e **follower** joint positions (rad), velocities (rad/s), efforts. ⚠️ `t_rel_s` is **0.900 s late** — [subtract it](#known-quirks) |
+| `tcp_pose` | **~67 Hz** | 40,619 | `x,y,z`, `qw,qx,qy,qz`, `t_rel_s` | TCP pose in robot **base frame**: position (m) + quaternion. ⚠️ `t_rel_s` is **≈0.41 s late** — [subtract it](#known-quirks) |
+| `wrench` | **~67 Hz** | 40,621 | `fx,fy,fz`, `tx,ty,tz`, `t_rel_s` | 6-axis end-effector force (N) + torque (N·m). ⚠️ `t_rel_s` is **≈0.41 s late** — [subtract it](#known-quirks) |
 | `gripper` | **~38.8 Hz** | 24,015 | `grip_pos`, `grip_cmd`, `gello_grip`, `t_rel_s` | measured opening, commanded, leader trigger |
 | `gello_joint_states` | 30.0 Hz | 18,573 | `q1..q6`, `qd1..qd6`, `t_rel_s` | **GELLO leader** joints (rad) + velocities (rad/s) |
 | `synchronized` | — | **0** | (56 declared keys, all length 0) | **EMPTY in every take** — a scaffold group; ignore |
@@ -137,8 +137,10 @@ rate. All datasets are **1-D `float64`** in a columnar layout (channel `foo` is
 Notes:
 - `command` (cmd1..6) are the joint targets sent to the UR7e; `ur_joint_states.q*` are the measured
   follower joints. The median per-take `|cmd_i − q_i|` is 0.023–0.080 rad depending on the joint
-  (worst single sample 0.510 rad) — ordinary servo following error on a hand-driven stream, but it
-  means `command` is a **target**, not a measurement.
+  (worst single sample 0.510 rad) — but ⚠️ **that is measured on the raw timestamps and is mostly
+  the 0.900 s timestamp artefact, not tracking error**: on the corrected clock the residual falls to
+  0.0007–0.0103 rad (see [Known quirks](#known-quirks)). Either way `command` is a **target**, not a
+  measurement.
 - `gello_joint_states.q*` are the **leader** joints. In this session they do **not** mirror the
   follower — see [Known quirks](#known-quirks). They are not a valid inference input either way
   (the deployed robot cannot see the leader).
@@ -276,7 +278,17 @@ fingers meeting each other with nothing between them, which happens in exactly t
 | [LeRobot joint + depth](https://huggingface.co/datasets/Bigenlight/carrot_in_pot_lerobot_v3) | `observation.state`(7)=`ur_q1..6`+`grip_pos`; `action`(7)=`cmd1..6`+`grip_cmd`; 2 RGB videos + **2 native depth video features**, **54** episodes |
 
 The LeRobot dataset resamples every stream onto the **30 fps `cam1` timestamp grid** (nearest
-timestamp), drops the `gello_*` leader streams, and re-encodes the videos. Depth is carried as
+timestamp), drops the `gello_*` leader streams, and re-encodes the videos.
+
+> **The LeRobot release is time-corrected; this one is not.** Because `observation.state[0:6]` comes
+> from `ur_joint_states`, the conversion first shifts that table's `t_rel_s` by that take's measured
+> τ (**−0.900 s** in 41 of the 54 episodes, **−0.895 s** in the other 13) and only
+> then resamples onto the camera grid — see the [timestamp artefact](#known-quirks) above. Nothing
+> else is shifted: `grip_pos`, `action` (`command` + `grip_cmd`), the RGB videos and both depth
+> streams were already on the camera timebase and are resampled unchanged. The applied τ is recorded
+> per episode in that dataset's `meta/source_takes.json` (`ur_joint_states_lag_s`). **`tcp_pose` and
+> `wrench`, the other two late tables, are not part of the LeRobot release at all** — if you need
+> them, take them from here and subtract ≈0.41 s yourself. Depth is carried as
 first-class LeRobot depth features — `observation.images.cam1_depth` and
 `observation.images.cam2_depth`, `(480, 848, 1)`, `is_depth_map: true`, HEVC `gray12le` lossless
 encoding of **linearly** 12-bit quantized codes over **0 – 10.0 m** (step = 10000/4095 ≈
@@ -307,6 +319,9 @@ with h5py.File(f"{take}/vectors.h5", "r") as f:
 cap = cv2.VideoCapture(f"{take}/cam1.mp4")   # 1280x720 colour @ 30 fps
 ```
 
+> ⚠️ **Before you align anything to `ur_q`, `tcp` or `wrench` above, subtract their timestamp
+> offset** — 0.900 s and ≈0.41 s respectively. See [Known quirks](#known-quirks).
+
 For depth see the snippet in [Depth](#depth) above. To go straight to training, use the ready
 LeRobot dataset instead (`LeRobotDataset("Bigenlight/carrot_in_pot_lerobot_v3")`).
 
@@ -317,6 +332,89 @@ LeRobot dataset instead (`LeRobotDataset("Bigenlight/carrot_in_pot_lerobot_v3")`
 
 ## Known quirks
 
+- ⚠️ **Timestamp artefact — `ur_joint_states`, `tcp_pose` and `wrench` rows are stamped late.**
+  **The arm was not late; the recording is.** Three of the nine groups carry timestamps that are
+  behind the rest of the take by a *constant* amount, so their samples describe the robot as it was
+  a fraction of a second **earlier** than their `t_rel_s` claims.
+
+  | group | how late | per-take spread | how it was measured |
+  |---|--:|---|---|
+  | `ur_joint_states` | **0.900 s** | **0.895 – 0.900 s**, median **0.900 s** — most takes land exactly on 0.900 and the rest one 5 ms grid step below | joint-space residual: shift the `ur_joint_states` clock by −τ, interpolate the six joints onto `command`, minimise mean \|ur_q − cmd\| on a **5 ms grid** |
+  | `tcp_pose` | **≈ 0.41 s** (0.900 − 0.495; speed cross-correlation against `command` gives 0.41 independently) | **0.495 – 0.500 s** *ahead of* `ur_joint_states`, median 0.495 | forward-kinematics residual `fk(q) ⊕ 0.174 m` against `tcp_pose`: **11.2 – 23.3 mm** at zero shift, **0.52 – 2.89 mm** at the optimum |
+  | `wrench` | **≈ 0.41 s** — same queue depth and publisher as `tcp_pose` | measured **0.000 s** from `tcp_pose` (7 of 8 takes) | velocity cross-correlation |
+
+  Everything else is **fresh** and must not be touched: `command`, both colour cameras, both depth
+  streams, `gello_joint_states` and `gripper` (⚠️ `command`'s own residual age is ≈0.08 s rather than
+  exactly zero — see **Mechanism** below). The evidence is direct — the depth frames store the
+  driver's own header stamp, and their measured age at the moment they were written is **0.022 s**;
+  scene-camera motion energy correlates with `command` at **τ ≈ 0** (−0.055 s / +0.010 s on the two
+  takes checked) while the *same* motion correlates with `ur_joint_states` at **+0.865 / +0.935 s**;
+  and `tcp_pose` ↔ `wrench`, two tables that come out of the same driver cycle, are **0.000 s**
+  apart, which no "the robot lagged" explanation can produce.
+
+  **Mechanism.** The recorder writes every row inside its ROS subscription callback and stamps it
+  with the time that callback *ran* — header stamps were not stored — and all callbacks share a
+  single rclpy spin thread that serves one message per subscription per round. Adding depth
+  recording to this session dropped that round rate to **≈67 Hz — 41–70 Hz across takes**, so
+  every publisher faster than it kept its `KEEP_LAST` queue permanently full and each row came out
+  already `queue_depth ÷ publish_rate` old: `/joint_states` depth **100** at ~100 Hz → **1.0 s**,
+  `tcp_pose` / `wrench` depth **50** at ~100 Hz → **0.5 s**. The signature is in the rate table above — `command`,
+  `ur_joint_states`, `tcp_pose` and `wrench` have two *different* publish rates (the ~100 Hz
+  `controller_manager` broadcasters and the 250 Hz `command` stream) yet all four converge on the
+  same recorded rate (e.g. 59.792 / 59.791 / 59.793 / 59.792 Hz in `take_01`); that is what
+  starvation looks like. Streams at or below the round rate (30–40 Hz) never queued, which is why
+  they are fresh. ⚠️ **`command` is the one stream the queue model does not predict** — at 250 Hz
+  behind a depth-50 queue it should have emerged ~0.2 s old, and it measures ≈0.08 s. It is treated
+  as fresh on the *measurements* above (scene-camera cross-correlation at τ ≈ 0), not on the model.
+
+  **It is a pure delay, not a distortion** — the waveform is intact, which is why one constant shift
+  recovers it. Independently re-measured on all 54 takes, the `ur_joint_states` ↔ `command`
+  residual drops from **0.049–0.090 rad** (median 0.063) uncorrected to **0.0007–0.0103 rad**
+  (median 0.0062) at the per-take optimum; on the release script's own more robust objective it
+  drops from **0.034–0.081 rad** to **0.0004–0.0009 rad**. For `tcp_pose` the same story in
+  millimetres: `fk(q) ⊕ 0.174 m` versus `tcp_pose` goes from **11.2–23.3 mm** to **0.52–2.89 mm**.
+  A servo lag would smear the waveform and could not collapse a residual by that factor; only a
+  timestamp offset can.
+
+  ⚠️ **Quote the number with its method attached, because four estimators land within ±0.02 s of
+  each other and none of them is "wrong":** joint-space residual on a 5 ms grid → **0.895–0.900 s**
+  (median 0.900; 41 of 54 takes land exactly on 0.900 in the independent re-measurement the LeRobot
+  release applies, 38 of 54 in `dataset_stats.json`'s own run of the same estimator — same lag,
+  different tie-break); the same residual with a median-over-time objective
+  → **0.900 s in all 54 takes**; velocity cross-correlation of `command` against `ur_joint_states`
+  → **0.875–0.915 s** (median 0.900; the sibling analysis reports 0.910 s); TCP-position residual on
+  a 10 ms grid → **0.90 s** (11 of 11 takes checked, residual median 0.42–0.64 mm against 44–67 mm
+  at τ = 0). Say **0.89–0.91 s depending on method and grid**; the derived LeRobot
+  release applies the **per-take optimum** — **0.900 s** for 41 takes and **0.895 s** for 13 —
+  recorded episode by episode in its `meta/source_takes.json`. Both objectives are in
+  [`dataset_stats.json`](dataset_stats.json) → `timestamp_lag` (`ur_joint_states_lag_s` and
+  `ur_joint_states_lag_s_mean_objective`) so you can see for yourself that they differ by at most
+  one grid step. **A single 0.900 s for everything is within 5 ms of correct and perfectly usable**
+  — 5 ms is a seventh of a camera frame.
+
+  **Fix rule for consumers.** Subtract **0.900 s** from `ur_joint_states/t_rel_s` and **≈0.41 s**
+  from `tcp_pose/t_rel_s` and `wrench/t_rel_s` (the two constants must differ by the measured
+  **0.495 s**, which is the tightest number here — 0.900 − 0.495 = 0.405). **Do not shift anything
+  else** — the camera,
+  depth, gripper and `command` tables already share one timebase, and shifting them would *create*
+  a misalignment that is not in the data.
+
+  ```python
+  import h5py, numpy as np
+  with h5py.File("take_01_20260914_164811/vectors.h5", "r") as f:
+      cam_t = f["cam1_frames"]["t_rel_s"][:]                       # master clock, fresh
+      ur_t  = f["ur_joint_states"]["t_rel_s"][:] - 0.900           # <- the whole fix
+      ur_q  = np.stack([f["ur_joint_states"][f"q{i}"][:] for i in range(1, 7)], 1)
+  ur_on_cam = np.stack([np.interp(cam_t, ur_t, ur_q[:, k]) for k in range(6)], 1)
+  ```
+
+  **The files here are exactly as recorded — nothing was rewritten** — because a documented constant
+  is recoverable and a silently re-stamped file is not. Per-take values, residuals and the
+  cross-correlation checks are in [`dataset_stats.json`](dataset_stats.json) under `timestamp_lag`.
+  The recorder itself was **fixed after this release** (header stamps are now stored, the high-rate
+  queue depths were shrunk, camera/depth writes were moved off the spin thread, and it now warns
+  when the recorded rates of unrelated topics converge), and **depth recording is now opt-in
+  (`ENABLE_DEPTH=1`), default RGB-only**, so later sessions in this family are not affected.
 - **`take_12` is absent by design.** Folders run `take_01` … `take_55` but take 12 was discarded
   during recording, giving exactly **54** folders. Do not treat the gap as missing data; there is no
   partial `take_12` anywhere in the release.
@@ -390,9 +488,14 @@ Audited over all 54 takes:
   (`unit`, `depth_scale_m`, `width`, `height`, `aligned_to_color`, `source_topic`) likewise.
 - **Uniform schema:** identical 9 groups, identical channel names, all `float64`, in all 54
   `vectors.h5`; identical video format (`mpeg4 1280×720 yuv420p 30/1`) in all 108 MP4s.
-- Timing is well behaved: the largest single gap in any robot stream is **44.1 ms**, the largest in
-  any colour camera stream **61.0 ms** (~2 frame periods), and the largest in any depth stream
-  **69.2 ms**.
+- **Timestamp *spacing* is well behaved; timestamp *offset* is not, for three groups.** No stream
+  drops out: the largest single gap in any robot stream is **44.1 ms**, the largest in any colour
+  camera stream **61.0 ms** (~2 frame periods), and the largest in any depth stream **69.2 ms**.
+  But `ur_joint_states` (**0.900 s**), `tcp_pose` and `wrench` (**≈0.41 s**) are stamped *late* as a
+  whole — a constant offset, not jitter, and **not** corrected in these files. Cross-stream
+  alignment is therefore wrong by that much until you apply the shift in [Known
+  quirks](#known-quirks). `command`, the cameras, depth, `gello_joint_states` and `gripper` are on a
+  common, correct timebase.
 - Cleanliness: exactly four files per take folder, no stray files, no hidden files, no
   sub-directories, and **no `/home/` absolute-path string anywhere in the 216 data files** (54
   `vectors.h5` + 54 `depth.h5` + 108 MP4s, full byte scan).
@@ -410,6 +513,10 @@ carrot in the pot.
 - **No trained policy exists for this task yet.**
 - Streams are **asynchronous** (each has its own clock); you must resample/align them yourself —
   the `synchronized/` group is empty.
+- **Three groups need a constant timestamp correction before they can be aligned to anything else**
+  (`ur_joint_states` −0.900 s, `tcp_pose` / `wrench` −≈0.41 s; see [Known quirks](#known-quirks)).
+  It is one subtraction, it is documented per take in `dataset_stats.json`, and the derived LeRobot
+  release has it applied — but it is a real trap if you skip it.
 - Sampling is bursty (see the rate note above), so nearest-timestamp alignment is required rather
   than index arithmetic.
 - **Depth is unaligned to colour and the colour intrinsics are not included** (see [Depth](#depth)),

dataset_stats.json

@@ -1,6 +1,6 @@
 {
  "dataset": "Bigenlight/carrot_in_pot_raw",
- "generated_utc": "2026-09-14T09:11:23+00:00",
+ "generated_utc": "2026-09-14T14:28:17+00:00",
  "generated_by": "make_carrot_raw_stats.py \u2014 h5py + cv2 + ffprobe audit over all 54 take folders (vectors.h5, depth.h5, cam1.mp4, cam2.mp4). mean rate = (N-1)/(t_last-t_first) per take, then median across takes; dt percentiles pooled over all takes. Video frame counts from `ffprobe -count_frames`. Value ranges pooled over all takes. Depth quality measured on ONE sampled mid frame per camera per take (decoded with cv2.imdecode(..., IMREAD_UNCHANGED)).",
  "aggregate": {
   "n_takes": 54,
@@ -371,7 +371,7 @@
   }
  },
  "command_tracking": {
-  "note": "|command cmd_i - ur_joint_states q_i| after nearest-timestamp alignment of the follower onto the command clock. This is the servo following error, pooled over all takes.",
+  "note": "|command cmd_i - ur_joint_states q_i| after nearest-timestamp alignment of the follower onto the command clock. Measured on the RAW (uncorrected) timestamps, so it is mostly the ~0.9 s timestamp artefact documented in `timestamp_lag`, NOT servo following error: on the corrected clock the 6-joint mean residual falls to 0.0007-0.0103 rad. Pooled over all takes.",
   "per_joint": {
    "j1": {
     "median_of_take_medians_rad": 0.0227,
@@ -405,6 +405,947 @@
    }
   }
  },
+ "timestamp_lag": {
+  "note": "RECORDER ARTEFACT, not robot behaviour: `ur_joint_states`, `tcp_pose` and `wrench` rows are stamped LATE in this release. The GUI recorder writes every row inside its subscription callback and sets t_rel_s to the time that callback RAN; ROS header stamps are not stored. All callbacks share one rclpy spin thread, which a SingleThreadedExecutor services one message per subscription per round. Adding depth recording to this session dropped the round rate to ~67 Hz (41-70 Hz across takes; see `rates`), so every publisher faster than that keeps its KEEP_LAST queue permanently full and each row leaves the queue already `queue_depth / publish_rate` seconds old: /joint_states depth 100 at ~100 Hz -> ~1.0 s, tcp_pose and wrench depth 50 at ~100 Hz -> ~0.5 s. The starvation signature is visible in `rates`: command / ur_joint_states / tcp_pose / wrench, four topics with two different publish rates (250 Hz for command, ~100 Hz for the three controller_manager broadcasters), all converge on the same recorded rate. Streams at or below the round rate never queue and are fresh: cameras (depth header-stamp age measured at 0.022 s), depth, gello_joint_states and gripper. `command` is the one exception to the model rather than an example of it: it publishes at 250 Hz behind a depth-50 queue, so the model predicts ~0.20 s, while the scene-camera cross-correlation puts it at tau ~ 0 and the queue algebra at ~0.08 s. It is treated as fresh on that measurement, to within ~0.08 s. The lag is a PURE DELAY - the waveform is intact - so subtracting the constant below from t_rel_s restores the timebase. NOTHING IS REWRITTEN HERE: the raw files are as recorded. FIX RULE for consumers: subtract ur_joint_states_lag_s from ur_joint_states/t_rel_s and ~0.41 s from tcp_pose/t_rel_s and wrench/t_rel_s (0.900 - 0.495 = 0.405, and speed cross-correlation against command independently gives 0.41); do NOT shift anything else - the gripper and camera tables are already on the camera timebase and shifting them would create a misalignment that is not there. The recorder has since been fixed (header stamps stored, queue depths shrunk, camera/depth writes moved off the spin thread, starvation warning) and depth recording is now opt-in (ENABLE_DEPTH=1, default RGB-only) - this release predates that fix.",
+  "method": "ur_joint_states_lag_s: joint-space residual. For each candidate lag on a 5 ms grid over 0..1.5 s the ur_joint_states clock is shifted EARLIER by that lag, the six joints are linearly interpolated onto the `command` timestamps, and the lag minimising the median over time of the 6-joint mean |ur_q - cmd| is reported. residual_rad_at_lag / residual_rad_at_zero give that residual with and without the correction; the *_mean variants are the plain mean over all samples and joints, which is larger because it is dominated by the high-velocity transients where the genuine servo following error lives. ur_joint_states_lag_s_mean_objective is the argmin of that plain mean instead of the median: it is the SAME measurement and agrees to within one 5 ms grid step, and it is reported because independent re-measurements of this dataset use it and land on 0.895 s for a minority of takes. Quote the lag as 0.89-0.91 s depending on method and grid. The derived LeRobot release applies the PER-TAKE optimum from the independent re-measurement (carrot_in_pot_lag.json): 0.900 s for 41 takes and 0.895 s for 13, recorded episode by episode in its meta/source_takes.json. Note the two runs of the mean objective split the ties differently (38 of 54 at 0.900 here, 41 of 54 there); they never differ by more than one 5 ms grid step. xcorr_*: independent check by speed-profile cross-correlation on the same grid (speed, not position, so a constant spatial offset such as the 0.174 m tool offset cannot bias it); a positive value means the second stream is stamped that many seconds later than the first.",
+  "ur_joint_states_lag_s": {
+   "min": 0.9,
+   "median": 0.9,
+   "max": 0.9,
+   "n_takes": 54
+  },
+  "ur_joint_states_lag_s_mean_objective": {
+   "min": 0.895,
+   "median": 0.9,
+   "max": 0.9,
+   "n_takes": 54
+  },
+  "takes_at_modal_lag": 54,
+  "modal_lag_s": 0.9,
+  "residual_rad_at_lag": {
+   "min": 0.0004,
+   "median": 0.0007,
+   "max": 0.0009,
+   "n_takes": 54
+  },
+  "residual_rad_at_lag_mean": {
+   "min": 0.0016,
+   "median": 0.0103,
+   "max": 0.0167,
+   "n_takes": 54
+  },
+  "residual_rad_at_zero": {
+   "min": 0.034,
+   "median": 0.0515,
+   "max": 0.0814,
+   "n_takes": 54
+  },
+  "residual_rad_at_zero_mean": {
+   "min": 0.0447,
+   "median": 0.0581,
+   "max": 0.0782,
+   "n_takes": 54
+  },
+  "xcorr_command_to_ur_joint_states_s": {
+   "min": 0.58,
+   "median": 0.905,
+   "max": 0.94,
+   "n_takes": 54
+  },
+  "xcorr_command_to_tcp_pose_s": {
+   "min": 0.375,
+   "median": 0.41,
+   "max": 1.395,
+   "n_takes": 54
+  },
+  "xcorr_tcp_pose_to_ur_joint_states_s": {
+   "min": 0.0,
+   "median": 0.465,
+   "max": 0.8,
+   "n_takes": 54
+  },
+  "fresh_streams": [
+   "cam1_frames",
+   "cam2_frames",
+   "depth.h5 cam1",
+   "depth.h5 cam2",
+   "gello_joint_states",
+   "gripper",
+   "command"
+  ],
+  "late_streams": [
+   "ur_joint_states",
+   "tcp_pose",
+   "wrench"
+  ],
+  "wrench_note": "`wrench` shares tcp_pose's QoS depth (50) and publisher, so it carries the same ~0.5 s lag. It is not cross-correlated here because the force signal's derivative is too noisy to time reliably; the sibling velocity-correlation analysis measured tcp_pose -> wrench at 0.000 s over 7 of 8 takes, which is the sharpest confirmation of the queue-depth model (equal depth -> equal lag), and tcp_pose -> ur_joint_states at +0.495 s (unequal depth).",
+  "per_take": [
+   {
+    "take": "take_01_20260914_164811",
+    "ur_joint_states_lag_s": 0.9,
+    "ur_joint_states_lag_s_mean_objective": 0.9,
+    "residual_rad_at_lag": 0.0006,
+    "residual_rad_at_lag_mean": 0.0016,
+    "n_compared": 503,
+    "residual_rad_at_zero": 0.0581,
+    "residual_rad_at_zero_mean": 0.0527,
+    "xcorr_command_to_ur_joint_states_s": 0.915,
+    "xcorr_command_to_ur_joint_states_r": 0.824,
+    "xcorr_command_to_tcp_pose_s": 0.42,
+    "xcorr_command_to_tcp_pose_r": 0.746,
+    "xcorr_tcp_pose_to_ur_joint_states_s": 0.46,
+    "xcorr_tcp_pose_to_ur_joint_states_r": 0.75
+   },
+   {
+    "take": "take_02_20260914_164845",
+    "ur_joint_states_lag_s": 0.9,
+    "ur_joint_states_lag_s_mean_objective": 0.9,
+    "residual_rad_at_lag": 0.0007,
+    "residual_rad_at_lag_mean": 0.0086,
+    "n_compared": 642,
+    "residual_rad_at_zero": 0.0417,
+    "residual_rad_at_zero_mean": 0.0511,
+    "xcorr_command_to_ur_joint_states_s": 0.875,
+    "xcorr_command_to_ur_joint_states_r": 0.606,
+    "xcorr_command_to_tcp_pose_s": 0.435,
+    "xcorr_command_to_tcp_pose_r": 0.708,
+    "xcorr_tcp_pose_to_ur_joint_states_s": 0.4,
+    "xcorr_tcp_pose_to_ur_joint_states_r": 0.643
+   },
+   {
+    "take": "take_03_20260914_164930",
+    "ur_joint_states_lag_s": 0.9,
+    "ur_joint_states_lag_s_mean_objective": 0.9,
+    "residual_rad_at_lag": 0.0008,
+    "residual_rad_at_lag_mean": 0.0042,
+    "n_compared": 650,
+    "residual_rad_at_zero": 0.0745,
+    "residual_rad_at_zero_mean": 0.078,
+    "xcorr_command_to_ur_joint_states_s": 0.94,
+    "xcorr_command_to_ur_joint_states_r": 0.79,
+    "xcorr_command_to_tcp_pose_s": 0.42,
+    "xcorr_command_to_tcp_pose_r": 0.651,
+    "xcorr_tcp_pose_to_ur_joint_states_s": 0.475,
+    "xcorr_tcp_pose_to_ur_joint_states_r": 0.604
+   },
+   {
+    "take": "take_04_20260914_165011",
+    "ur_joint_states_lag_s": 0.9,
+    "ur_joint_states_lag_s_mean_objective": 0.9,
+    "residual_rad_at_lag": 0.0007,
+    "residual_rad_at_lag_mean": 0.0078,
+    "n_compared": 592,
+    "residual_rad_at_zero": 0.0689,
+    "residual_rad_at_zero_mean": 0.0674,
+    "xcorr_command_to_ur_joint_states_s": 0.88,
+    "xcorr_command_to_ur_joint_states_r": 0.664,
+    "xcorr_command_to_tcp_pose_s": 0.41,
+    "xcorr_command_to_tcp_pose_r": 0.497,
+    "xcorr_tcp_pose_to_ur_joint_states_s": 0.53,
+    "xcorr_tcp_pose_to_ur_joint_states_r": 0.471
+   },
+   {
+    "take": "take_05_20260914_165156",
+    "ur_joint_states_lag_s": 0.9,
+    "ur_joint_states_lag_s_mean_objective": 0.895,
+    "residual_rad_at_lag": 0.0006,
+    "residual_rad_at_lag_mean": 0.0128,
+    "n_compared": 659,
+    "residual_rad_at_zero": 0.0471,
+    "residual_rad_at_zero_mean": 0.0522,
+    "xcorr_command_to_ur_joint_states_s": 0.91,
+    "xcorr_command_to_ur_joint_states_r": 0.692,
+    "xcorr_command_to_tcp_pose_s": 0.415,
+    "xcorr_command_to_tcp_pose_r": 0.66,
+    "xcorr_tcp_pose_to_ur_joint_states_s": 0.435,
+    "xcorr_tcp_pose_to_ur_joint_states_r": 0.583
+   },
+   {
+    "take": "take_06_20260914_165238",
+    "ur_joint_states_lag_s": 0.9,
+    "ur_joint_states_lag_s_mean_objective": 0.895,
+    "residual_rad_at_lag": 0.0009,
+    "residual_rad_at_lag_mean": 0.0119,
+    "n_compared": 666,
+    "residual_rad_at_zero": 0.0649,
+    "residual_rad_at_zero_mean": 0.067,
+    "xcorr_command_to_ur_joint_states_s": 0.885,
+    "xcorr_command_to_ur_joint_states_r": 0.626,
+    "xcorr_command_to_tcp_pose_s": 0.375,
+    "xcorr_command_to_tcp_pose_r": 0.654,
+    "xcorr_tcp_pose_to_ur_joint_states_s": 0.505,
+    "xcorr_tcp_pose_to_ur_joint_states_r": 0.503
+   },
+   {
+    "take": "take_07_20260914_165316",
+    "ur_joint_states_lag_s": 0.9,
+    "ur_joint_states_lag_s_mean_objective": 0.9,
+    "residual_rad_at_lag": 0.0006,
+    "residual_rad_at_lag_mean": 0.0129,
+    "n_compared": 519,
+    "residual_rad_at_zero": 0.0398,
+    "residual_rad_at_zero_mean": 0.0447,
+    "xcorr_command_to_ur_joint_states_s": 0.905,
+    "xcorr_command_to_ur_joint_states_r": 0.526,
+    "xcorr_command_to_tcp_pose_s": 0.375,
+    "xcorr_command_to_tcp_pose_r": 0.675,
+    "xcorr_tcp_pose_to_ur_joint_states_s": 0.2,
+    "xcorr_tcp_pose_to_ur_joint_states_r": 0.638
+   },
+   {
+    "take": "take_08_20260914_165353",
+    "ur_joint_states_lag_s": 0.9,
+    "ur_joint_states_lag_s_mean_objective": 0.895,
+    "residual_rad_at_lag": 0.0008,
+    "residual_rad_at_lag_mean": 0.0137,
+    "n_compared": 708,
+    "residual_rad_at_zero": 0.0814,
+    "residual_rad_at_zero_mean": 0.0782,
+    "xcorr_command_to_ur_joint_states_s": 0.9,
+    "xcorr_command_to_ur_joint_states_r": 0.62,
+    "xcorr_command_to_tcp_pose_s": 0.43,
+    "xcorr_command_to_tcp_pose_r": 0.473,
+    "xcorr_tcp_pose_to_ur_joint_states_s": 0.0,
+    "xcorr_tcp_pose_to_ur_joint_states_r": 0.493
+   },
+   {
+    "take": "take_09_20260914_165429",
+    "ur_joint_states_lag_s": 0.9,
+    "ur_joint_states_lag_s_mean_objective": 0.895,
+    "residual_rad_at_lag": 0.0007,
+    "residual_rad_at_lag_mean": 0.0167,
+    "n_compared": 529,
+    "residual_rad_at_zero": 0.0386,
+    "residual_rad_at_zero_mean": 0.0525,
+    "xcorr_command_to_ur_joint_states_s": 0.925,
+    "xcorr_command_to_ur_joint_states_r": 0.624,
+    "xcorr_command_to_tcp_pose_s": 0.405,
+    "xcorr_command_to_tcp_pose_r": 0.584,
+    "xcorr_tcp_pose_to_ur_joint_states_s": 0.48,
+    "xcorr_tcp_pose_to_ur_joint_states_r": 0.595
+   },
+   {
+    "take": "take_10_20260914_165510",
+    "ur_joint_states_lag_s": 0.9,
+    "ur_joint_states_lag_s_mean_objective": 0.895,
+    "residual_rad_at_lag": 0.0008,
+    "residual_rad_at_lag_mean": 0.0117,
+    "n_compared": 623,
+    "residual_rad_at_zero": 0.0586,
+    "residual_rad_at_zero_mean": 0.0638,
+    "xcorr_command_to_ur_joint_states_s": 0.89,
+    "xcorr_command_to_ur_joint_states_r": 0.476,
+    "xcorr_command_to_tcp_pose_s": 0.44,
+    "xcorr_command_to_tcp_pose_r": 0.632,
+    "xcorr_tcp_pose_to_ur_joint_states_s": 0.0,
+    "xcorr_tcp_pose_to_ur_joint_states_r": 0.504
+   },
+   {
+    "take": "take_11_20260914_165557",
+    "ur_joint_states_lag_s": 0.9,
+    "ur_joint_states_lag_s_mean_objective": 0.895,
+    "residual_rad_at_lag": 0.0007,
+    "residual_rad_at_lag_mean": 0.014,
+    "n_compared": 822,
+    "residual_rad_at_zero": 0.0429,
+    "residual_rad_at_zero_mean": 0.049,
+    "xcorr_command_to_ur_joint_states_s": 0.58,
+    "xcorr_command_to_ur_joint_states_r": 0.584,
+    "xcorr_command_to_tcp_pose_s": 0.47,
+    "xcorr_command_to_tcp_pose_r": 0.513,
+    "xcorr_tcp_pose_to_ur_joint_states_s": 0.48,
+    "xcorr_tcp_pose_to_ur_joint_states_r": 0.401
+   },
+   {
+    "take": "take_13_20260914_165647",
+    "ur_joint_states_lag_s": 0.9,
+    "ur_joint_states_lag_s_mean_objective": 0.9,
+    "residual_rad_at_lag": 0.0007,
+    "residual_rad_at_lag_mean": 0.0131,
+    "n_compared": 563,
+    "residual_rad_at_zero": 0.0382,
+    "residual_rad_at_zero_mean": 0.0469,
+    "xcorr_command_to_ur_joint_states_s": 0.92,
+    "xcorr_command_to_ur_joint_states_r": 0.568,
+    "xcorr_command_to_tcp_pose_s": 0.4,
+    "xcorr_command_to_tcp_pose_r": 0.624,
+    "xcorr_tcp_pose_to_ur_joint_states_s": 0.5,
+    "xcorr_tcp_pose_to_ur_joint_states_r": 0.566
+   },
+   {
+    "take": "take_14_20260914_165719",
+    "ur_joint_states_lag_s": 0.9,
+    "ur_joint_states_lag_s_mean_objective": 0.895,
+    "residual_rad_at_lag": 0.0007,
+    "residual_rad_at_lag_mean": 0.0109,
+    "n_compared": 763,
+    "residual_rad_at_zero": 0.0702,
+    "residual_rad_at_zero_mean": 0.0644,
+    "xcorr_command_to_ur_joint_states_s": 0.91,
+    "xcorr_command_to_ur_joint_states_r": 0.757,
+    "xcorr_command_to_tcp_pose_s": 0.41,
+    "xcorr_command_to_tcp_pose_r": 0.592,
+    "xcorr_tcp_pose_to_ur_joint_states_s": 0.5,
+    "xcorr_tcp_pose_to_ur_joint_states_r": 0.619
+   },
+   {
+    "take": "take_15_20260914_165753",
+    "ur_joint_states_lag_s": 0.9,
+    "ur_joint_states_lag_s_mean_objective": 0.9,
+    "residual_rad_at_lag": 0.0006,
+    "residual_rad_at_lag_mean": 0.0063,
+    "n_compared": 901,
+    "residual_rad_at_zero": 0.0437,
+    "residual_rad_at_zero_mean": 0.0505,
+    "xcorr_command_to_ur_joint_states_s": 0.905,
+    "xcorr_command_to_ur_joint_states_r": 0.662,
+    "xcorr_command_to_tcp_pose_s": 0.415,
+    "xcorr_command_to_tcp_pose_r": 0.724,
+    "xcorr_tcp_pose_to_ur_joint_states_s": 0.5,
+    "xcorr_tcp_pose_to_ur_joint_states_r": 0.673
+   },
+   {
+    "take": "take_16_20260914_165825",
+    "ur_joint_states_lag_s": 0.9,
+    "ur_joint_states_lag_s_mean_objective": 0.9,
+    "residual_rad_at_lag": 0.0007,
+    "residual_rad_at_lag_mean": 0.0082,
+    "n_compared": 695,
+    "residual_rad_at_zero": 0.0441,
+    "residual_rad_at_zero_mean": 0.0533,
+    "xcorr_command_to_ur_joint_states_s": 0.9,
+    "xcorr_command_to_ur_joint_states_r": 0.659,
+    "xcorr_command_to_tcp_pose_s": 0.41,
+    "xcorr_command_to_tcp_pose_r": 0.677,
+    "xcorr_tcp_pose_to_ur_joint_states_s": 0.5,
+    "xcorr_tcp_pose_to_ur_joint_states_r": 0.716
+   },
+   {
+    "take": "take_17_20260914_165859",
+    "ur_joint_states_lag_s": 0.9,
+    "ur_joint_states_lag_s_mean_objective": 0.895,
+    "residual_rad_at_lag": 0.0007,
+    "residual_rad_at_lag_mean": 0.0131,
+    "n_compared": 651,
+    "residual_rad_at_zero": 0.0562,
+    "residual_rad_at_zero_mean": 0.0585,
+    "xcorr_command_to_ur_joint_states_s": 0.9,
+    "xcorr_command_to_ur_joint_states_r": 0.542,
+    "xcorr_command_to_tcp_pose_s": 0.38,
+    "xcorr_command_to_tcp_pose_r": 0.588,
+    "xcorr_tcp_pose_to_ur_joint_states_s": 0.4,
+    "xcorr_tcp_pose_to_ur_joint_states_r": 0.555
+   },
+   {
+    "take": "take_18_20260914_165926",
+    "ur_joint_states_lag_s": 0.9,
+    "ur_joint_states_lag_s_mean_objective": 0.895,
+    "residual_rad_at_lag": 0.0007,
+    "residual_rad_at_lag_mean": 0.0133,
+    "n_compared": 743,
+    "residual_rad_at_zero": 0.0521,
+    "residual_rad_at_zero_mean": 0.0608,
+    "xcorr_command_to_ur_joint_states_s": 0.88,
+    "xcorr_command_to_ur_joint_states_r": 0.715,
+    "xcorr_command_to_tcp_pose_s": 0.45,
+    "xcorr_command_to_tcp_pose_r": 0.672,
+    "xcorr_tcp_pose_to_ur_joint_states_s": 0.465,
+    "xcorr_tcp_pose_to_ur_joint_states_r": 0.57
+   },
+   {
+    "take": "take_19_20260914_170024",
+    "ur_joint_states_lag_s": 0.9,
+    "ur_joint_states_lag_s_mean_objective": 0.9,
+    "residual_rad_at_lag": 0.0008,
+    "residual_rad_at_lag_mean": 0.0138,
+    "n_compared": 787,
+    "residual_rad_at_zero": 0.0514,
+    "residual_rad_at_zero_mean": 0.059,
+    "xcorr_command_to_ur_joint_states_s": 0.905,
+    "xcorr_command_to_ur_joint_states_r": 0.576,
+    "xcorr_command_to_tcp_pose_s": 0.44,
+    "xcorr_command_to_tcp_pose_r": 0.515,
+    "xcorr_tcp_pose_to_ur_joint_states_s": 0.0,
+    "xcorr_tcp_pose_to_ur_joint_states_r": 0.459
+   },
+   {
+    "take": "take_20_20260914_170107",
+    "ur_joint_states_lag_s": 0.9,
+    "ur_joint_states_lag_s_mean_objective": 0.895,
+    "residual_rad_at_lag": 0.0005,
+    "residual_rad_at_lag_mean": 0.0082,
+    "n_compared": 643,
+    "residual_rad_at_zero": 0.0405,
+    "residual_rad_at_zero_mean": 0.0492,
+    "xcorr_command_to_ur_joint_states_s": 0.885,
+    "xcorr_command_to_ur_joint_states_r": 0.684,
+    "xcorr_command_to_tcp_pose_s": 0.41,
+    "xcorr_command_to_tcp_pose_r": 0.698,
+    "xcorr_tcp_pose_to_ur_joint_states_s": 0.48,
+    "xcorr_tcp_pose_to_ur_joint_states_r": 0.677
+   },
+   {
+    "take": "take_21_20260914_170135",
+    "ur_joint_states_lag_s": 0.9,
+    "ur_joint_states_lag_s_mean_objective": 0.9,
+    "residual_rad_at_lag": 0.0006,
+    "residual_rad_at_lag_mean": 0.0094,
+    "n_compared": 605,
+    "residual_rad_at_zero": 0.0476,
+    "residual_rad_at_zero_mean": 0.052,
+    "xcorr_command_to_ur_joint_states_s": 0.87,
+    "xcorr_command_to_ur_joint_states_r": 0.669,
+    "xcorr_command_to_tcp_pose_s": 0.44,
+    "xcorr_command_to_tcp_pose_r": 0.671,
+    "xcorr_tcp_pose_to_ur_joint_states_s": 0.0,
+    "xcorr_tcp_pose_to_ur_joint_states_r": 0.556
+   },
+   {
+    "take": "take_22_20260914_170211",
+    "ur_joint_states_lag_s": 0.9,
+    "ur_joint_states_lag_s_mean_objective": 0.9,
+    "residual_rad_at_lag": 0.0009,
+    "residual_rad_at_lag_mean": 0.0117,
+    "n_compared": 815,
+    "residual_rad_at_zero": 0.0582,
+    "residual_rad_at_zero_mean": 0.0639,
+    "xcorr_command_to_ur_joint_states_s": 0.91,
+    "xcorr_command_to_ur_joint_states_r": 0.799,
+    "xcorr_command_to_tcp_pose_s": 0.405,
+    "xcorr_command_to_tcp_pose_r": 0.7,
+    "xcorr_tcp_pose_to_ur_joint_states_s": 0.5,
+    "xcorr_tcp_pose_to_ur_joint_states_r": 0.654
+   },
+   {
+    "take": "take_23_20260914_170302",
+    "ur_joint_states_lag_s": 0.9,
+    "ur_joint_states_lag_s_mean_objective": 0.9,
+    "residual_rad_at_lag": 0.0007,
+    "residual_rad_at_lag_mean": 0.0128,
+    "n_compared": 653,
+    "residual_rad_at_zero": 0.0759,
+    "residual_rad_at_zero_mean": 0.0742,
+    "xcorr_command_to_ur_joint_states_s": 0.905,
+    "xcorr_command_to_ur_joint_states_r": 0.615,
+    "xcorr_command_to_tcp_pose_s": 0.44,
+    "xcorr_command_to_tcp_pose_r": 0.658,
+    "xcorr_tcp_pose_to_ur_joint_states_s": 0.43,
+    "xcorr_tcp_pose_to_ur_joint_states_r": 0.578
+   },
+   {
+    "take": "take_24_20260914_170339",
+    "ur_joint_states_lag_s": 0.9,
+    "ur_joint_states_lag_s_mean_objective": 0.9,
+    "residual_rad_at_lag": 0.0006,
+    "residual_rad_at_lag_mean": 0.007,
+    "n_compared": 361,
+    "residual_rad_at_zero": 0.0578,
+    "residual_rad_at_zero_mean": 0.0615,
+    "xcorr_command_to_ur_joint_states_s": 0.92,
+    "xcorr_command_to_ur_joint_states_r": 0.824,
+    "xcorr_command_to_tcp_pose_s": 0.39,
+    "xcorr_command_to_tcp_pose_r": 0.809,
+    "xcorr_tcp_pose_to_ur_joint_states_s": 0.53,
+    "xcorr_tcp_pose_to_ur_joint_states_r": 0.738
+   },
+   {
+    "take": "take_25_20260914_170407",
+    "ur_joint_states_lag_s": 0.9,
+    "ur_joint_states_lag_s_mean_objective": 0.9,
+    "residual_rad_at_lag": 0.0006,
+    "residual_rad_at_lag_mean": 0.0058,
+    "n_compared": 379,
+    "residual_rad_at_zero": 0.0701,
+    "residual_rad_at_zero_mean": 0.0756,
+    "xcorr_command_to_ur_joint_states_s": 0.91,
+    "xcorr_command_to_ur_joint_states_r": 0.769,
+    "xcorr_command_to_tcp_pose_s": 0.46,
+    "xcorr_command_to_tcp_pose_r": 0.686,
+    "xcorr_tcp_pose_to_ur_joint_states_s": 0.445,
+    "xcorr_tcp_pose_to_ur_joint_states_r": 0.705
+   },
+   {
+    "take": "take_26_20260914_170447",
+    "ur_joint_states_lag_s": 0.9,
+    "ur_joint_states_lag_s_mean_objective": 0.9,
+    "residual_rad_at_lag": 0.0006,
+    "residual_rad_at_lag_mean": 0.0092,
+    "n_compared": 456,
+    "residual_rad_at_zero": 0.0519,
+    "residual_rad_at_zero_mean": 0.0532,
+    "xcorr_command_to_ur_joint_states_s": 0.89,
+    "xcorr_command_to_ur_joint_states_r": 0.684,
+    "xcorr_command_to_tcp_pose_s": 0.44,
+    "xcorr_command_to_tcp_pose_r": 0.668,
+    "xcorr_tcp_pose_to_ur_joint_states_s": 0.045,
+    "xcorr_tcp_pose_to_ur_joint_states_r": 0.643
+   },
+   {
+    "take": "take_27_20260914_170519",
+    "ur_joint_states_lag_s": 0.9,
+    "ur_joint_states_lag_s_mean_objective": 0.9,
+    "residual_rad_at_lag": 0.0007,
+    "residual_rad_at_lag_mean": 0.0142,
+    "n_compared": 510,
+    "residual_rad_at_zero": 0.0681,
+    "residual_rad_at_zero_mean": 0.0675,
+    "xcorr_command_to_ur_joint_states_s": 0.935,
+    "xcorr_command_to_ur_joint_states_r": 0.589,
+    "xcorr_command_to_tcp_pose_s": 0.39,
+    "xcorr_command_to_tcp_pose_r": 0.574,
+    "xcorr_tcp_pose_to_ur_joint_states_s": 0.0,
+    "xcorr_tcp_pose_to_ur_joint_states_r": 0.521
+   },
+   {
+    "take": "take_28_20260914_170557",
+    "ur_joint_states_lag_s": 0.9,
+    "ur_joint_states_lag_s_mean_objective": 0.9,
+    "residual_rad_at_lag": 0.0006,
+    "residual_rad_at_lag_mean": 0.0034,
+    "n_compared": 370,
+    "residual_rad_at_zero": 0.0562,
+    "residual_rad_at_zero_mean": 0.0604,
+    "xcorr_command_to_ur_joint_states_s": 0.915,
+    "xcorr_command_to_ur_joint_states_r": 0.808,
+    "xcorr_command_to_tcp_pose_s": 0.435,
+    "xcorr_command_to_tcp_pose_r": 0.819,
+    "xcorr_tcp_pose_to_ur_joint_states_s": 0.475,
+    "xcorr_tcp_pose_to_ur_joint_states_r": 0.747
+   },
+   {
+    "take": "take_29_20260914_170629",
+    "ur_joint_states_lag_s": 0.9,
+    "ur_joint_states_lag_s_mean_objective": 0.9,
+    "residual_rad_at_lag": 0.0008,
+    "residual_rad_at_lag_mean": 0.0094,
+    "n_compared": 669,
+    "residual_rad_at_zero": 0.0537,
+    "residual_rad_at_zero_mean": 0.0609,
+    "xcorr_command_to_ur_joint_states_s": 0.915,
+    "xcorr_command_to_ur_joint_states_r": 0.606,
+    "xcorr_command_to_tcp_pose_s": 0.465,
+    "xcorr_command_to_tcp_pose_r": 0.654,
+    "xcorr_tcp_pose_to_ur_joint_states_s": 0.0,
+    "xcorr_tcp_pose_to_ur_joint_states_r": 0.575
+   },
+   {
+    "take": "take_30_20260914_170700",
+    "ur_joint_states_lag_s": 0.9,
+    "ur_joint_states_lag_s_mean_objective": 0.9,
+    "residual_rad_at_lag": 0.0005,
+    "residual_rad_at_lag_mean": 0.0021,
+    "n_compared": 633,
+    "residual_rad_at_zero": 0.056,
+    "residual_rad_at_zero_mean": 0.0634,
+    "xcorr_command_to_ur_joint_states_s": 0.905,
+    "xcorr_command_to_ur_joint_states_r": 0.803,
+    "xcorr_command_to_tcp_pose_s": 0.425,
+    "xcorr_command_to_tcp_pose_r": 0.723,
+    "xcorr_tcp_pose_to_ur_joint_states_s": 0.5,
+    "xcorr_tcp_pose_to_ur_joint_states_r": 0.701
+   },
+   {
+    "take": "take_31_20260914_170732",
+    "ur_joint_states_lag_s": 0.9,
+    "ur_joint_states_lag_s_mean_objective": 0.9,
+    "residual_rad_at_lag": 0.0007,
+    "residual_rad_at_lag_mean": 0.0098,
+    "n_compared": 583,
+    "residual_rad_at_zero": 0.0663,
+    "residual_rad_at_zero_mean": 0.071,
+    "xcorr_command_to_ur_joint_states_s": 0.9,
+    "xcorr_command_to_ur_joint_states_r": 0.635,
+    "xcorr_command_to_tcp_pose_s": 0.42,
+    "xcorr_command_to_tcp_pose_r": 0.493,
+    "xcorr_tcp_pose_to_ur_joint_states_s": 0.435,
+    "xcorr_tcp_pose_to_ur_joint_states_r": 0.45
+   },
+   {
+    "take": "take_32_20260914_170803",
+    "ur_joint_states_lag_s": 0.9,
+    "ur_joint_states_lag_s_mean_objective": 0.895,
+    "residual_rad_at_lag": 0.0007,
+    "residual_rad_at_lag_mean": 0.0117,
+    "n_compared": 553,
+    "residual_rad_at_zero": 0.0446,
+    "residual_rad_at_zero_mean": 0.0544,
+    "xcorr_command_to_ur_joint_states_s": 0.89,
+    "xcorr_command_to_ur_joint_states_r": 0.558,
+    "xcorr_command_to_tcp_pose_s": 0.41,
+    "xcorr_command_to_tcp_pose_r": 0.606,
+    "xcorr_tcp_pose_to_ur_joint_states_s": 0.035,
+    "xcorr_tcp_pose_to_ur_joint_states_r": 0.516
+   },
+   {
+    "take": "take_33_20260914_170854",
+    "ur_joint_states_lag_s": 0.9,
+    "ur_joint_states_lag_s_mean_objective": 0.9,
+    "residual_rad_at_lag": 0.0005,
+    "residual_rad_at_lag_mean": 0.0112,
+    "n_compared": 488,
+    "residual_rad_at_zero": 0.0486,
+    "residual_rad_at_zero_mean": 0.0577,
+    "xcorr_command_to_ur_joint_states_s": 0.9,
+    "xcorr_command_to_ur_joint_states_r": 0.552,
+    "xcorr_command_to_tcp_pose_s": 0.395,
+    "xcorr_command_to_tcp_pose_r": 0.657,
+    "xcorr_tcp_pose_to_ur_joint_states_s": 0.435,
+    "xcorr_tcp_pose_to_ur_joint_states_r": 0.494
+   },
+   {
+    "take": "take_34_20260914_170932",
+    "ur_joint_states_lag_s": 0.9,
+    "ur_joint_states_lag_s_mean_objective": 0.895,
+    "residual_rad_at_lag": 0.0007,
+    "residual_rad_at_lag_mean": 0.014,
+    "n_compared": 782,
+    "residual_rad_at_zero": 0.0508,
+    "residual_rad_at_zero_mean": 0.0608,
+    "xcorr_command_to_ur_joint_states_s": 0.9,
+    "xcorr_command_to_ur_joint_states_r": 0.668,
+    "xcorr_command_to_tcp_pose_s": 0.425,
+    "xcorr_command_to_tcp_pose_r": 0.667,
+    "xcorr_tcp_pose_to_ur_joint_states_s": 0.43,
+    "xcorr_tcp_pose_to_ur_joint_states_r": 0.601
+   },
+   {
+    "take": "take_35_20260914_171001",
+    "ur_joint_states_lag_s": 0.9,
+    "ur_joint_states_lag_s_mean_objective": 0.895,
+    "residual_rad_at_lag": 0.0006,
+    "residual_rad_at_lag_mean": 0.008,
+    "n_compared": 867,
+    "residual_rad_at_zero": 0.0423,
+    "residual_rad_at_zero_mean": 0.0597,
+    "xcorr_command_to_ur_joint_states_s": 0.91,
+    "xcorr_command_to_ur_joint_states_r": 0.777,
+    "xcorr_command_to_tcp_pose_s": 0.39,
+    "xcorr_command_to_tcp_pose_r": 0.711,
+    "xcorr_tcp_pose_to_ur_joint_states_s": 0.495,
+    "xcorr_tcp_pose_to_ur_joint_states_r": 0.691
+   },
+   {
+    "take": "take_36_20260914_171037",
+    "ur_joint_states_lag_s": 0.9,
+    "ur_joint_states_lag_s_mean_objective": 0.9,
+    "residual_rad_at_lag": 0.0005,
+    "residual_rad_at_lag_mean": 0.004,
+    "n_compared": 1001,
+    "residual_rad_at_zero": 0.0641,
+    "residual_rad_at_zero_mean": 0.0611,
+    "xcorr_command_to_ur_joint_states_s": 0.905,
+    "xcorr_command_to_ur_joint_states_r": 0.782,
+    "xcorr_command_to_tcp_pose_s": 0.41,
+    "xcorr_command_to_tcp_pose_r": 0.725,
+    "xcorr_tcp_pose_to_ur_joint_states_s": 0.53,
+    "xcorr_tcp_pose_to_ur_joint_states_r": 0.627
+   },
+   {
+    "take": "take_37_20260914_171113",
+    "ur_joint_states_lag_s": 0.9,
+    "ur_joint_states_lag_s_mean_objective": 0.9,
+    "residual_rad_at_lag": 0.0007,
+    "residual_rad_at_lag_mean": 0.0091,
+    "n_compared": 843,
+    "residual_rad_at_zero": 0.0604,
+    "residual_rad_at_zero_mean": 0.0562,
+    "xcorr_command_to_ur_joint_states_s": 0.92,
+    "xcorr_command_to_ur_joint_states_r": 0.633,
+    "xcorr_command_to_tcp_pose_s": 0.435,
+    "xcorr_command_to_tcp_pose_r": 0.545,
+    "xcorr_tcp_pose_to_ur_joint_states_s": 0.0,
+    "xcorr_tcp_pose_to_ur_joint_states_r": 0.45
+   },
+   {
+    "take": "take_38_20260914_171144",
+    "ur_joint_states_lag_s": 0.9,
+    "ur_joint_states_lag_s_mean_objective": 0.9,
+    "residual_rad_at_lag": 0.0006,
+    "residual_rad_at_lag_mean": 0.0111,
+    "n_compared": 656,
+    "residual_rad_at_zero": 0.054,
+    "residual_rad_at_zero_mean": 0.0592,
+    "xcorr_command_to_ur_joint_states_s": 0.89,
+    "xcorr_command_to_ur_joint_states_r": 0.617,
+    "xcorr_command_to_tcp_pose_s": 0.435,
+    "xcorr_command_to_tcp_pose_r": 0.591,
+    "xcorr_tcp_pose_to_ur_joint_states_s": 0.43,
+    "xcorr_tcp_pose_to_ur_joint_states_r": 0.565
+   },
+   {
+    "take": "take_39_20260914_171227",
+    "ur_joint_states_lag_s": 0.9,
+    "ur_joint_states_lag_s_mean_objective": 0.9,
+    "residual_rad_at_lag": 0.0005,
+    "residual_rad_at_lag_mean": 0.0064,
+    "n_compared": 818,
+    "residual_rad_at_zero": 0.0388,
+    "residual_rad_at_zero_mean": 0.051,
+    "xcorr_command_to_ur_joint_states_s": 0.885,
+    "xcorr_command_to_ur_joint_states_r": 0.697,
+    "xcorr_command_to_tcp_pose_s": 0.42,
+    "xcorr_command_to_tcp_pose_r": 0.695,
+    "xcorr_tcp_pose_to_ur_joint_states_s": 0.435,
+    "xcorr_tcp_pose_to_ur_joint_states_r": 0.611
+   },
+   {
+    "take": "take_40_20260914_171256",
+    "ur_joint_states_lag_s": 0.9,
+    "ur_joint_states_lag_s_mean_objective": 0.9,
+    "residual_rad_at_lag": 0.0004,
+    "residual_rad_at_lag_mean": 0.0064,
+    "n_compared": 791,
+    "residual_rad_at_zero": 0.0392,
+    "residual_rad_at_zero_mean": 0.0484,
+    "xcorr_command_to_ur_joint_states_s": 0.91,
+    "xcorr_command_to_ur_joint_states_r": 0.669,
+    "xcorr_command_to_tcp_pose_s": 0.4,
+    "xcorr_command_to_tcp_pose_r": 0.68,
+    "xcorr_tcp_pose_to_ur_joint_states_s": 0.465,
+    "xcorr_tcp_pose_to_ur_joint_states_r": 0.618
+   },
+   {
+    "take": "take_41_20260914_171328",
+    "ur_joint_states_lag_s": 0.9,
+    "ur_joint_states_lag_s_mean_objective": 0.9,
+    "residual_rad_at_lag": 0.0007,
+    "residual_rad_at_lag_mean": 0.0083,
+    "n_compared": 683,
+    "residual_rad_at_zero": 0.0484,
+    "residual_rad_at_zero_mean": 0.0556,
+    "xcorr_command_to_ur_joint_states_s": 0.89,
+    "xcorr_command_to_ur_joint_states_r": 0.678,
+    "xcorr_command_to_tcp_pose_s": 0.435,
+    "xcorr_command_to_tcp_pose_r": 0.567,
+    "xcorr_tcp_pose_to_ur_joint_states_s": 0.5,
+    "xcorr_tcp_pose_to_ur_joint_states_r": 0.509
+   },
+   {
+    "take": "take_42_20260914_171358",
+    "ur_joint_states_lag_s": 0.9,
+    "ur_joint_states_lag_s_mean_objective": 0.9,
+    "residual_rad_at_lag": 0.0006,
+    "residual_rad_at_lag_mean": 0.0119,
+    "n_compared": 705,
+    "residual_rad_at_zero": 0.0486,
+    "residual_rad_at_zero_mean": 0.0513,
+    "xcorr_command_to_ur_joint_states_s": 0.9,
+    "xcorr_command_to_ur_joint_states_r": 0.676,
+    "xcorr_command_to_tcp_pose_s": 0.425,
+    "xcorr_command_to_tcp_pose_r": 0.548,
+    "xcorr_tcp_pose_to_ur_joint_states_s": 0.5,
+    "xcorr_tcp_pose_to_ur_joint_states_r": 0.553
+   },
+   {
+    "take": "take_43_20260914_171435",
+    "ur_joint_states_lag_s": 0.9,
+    "ur_joint_states_lag_s_mean_objective": 0.895,
+    "residual_rad_at_lag": 0.0006,
+    "residual_rad_at_lag_mean": 0.0116,
+    "n_compared": 647,
+    "residual_rad_at_zero": 0.0533,
+    "residual_rad_at_zero_mean": 0.0598,
+    "xcorr_command_to_ur_joint_states_s": 0.94,
+    "xcorr_command_to_ur_joint_states_r": 0.709,
+    "xcorr_command_to_tcp_pose_s": 0.41,
+    "xcorr_command_to_tcp_pose_r": 0.566,
+    "xcorr_tcp_pose_to_ur_joint_states_s": 0.43,
+    "xcorr_tcp_pose_to_ur_joint_states_r": 0.581
+   },
+   {
+    "take": "take_44_20260914_171505",
+    "ur_joint_states_lag_s": 0.9,
+    "ur_joint_states_lag_s_mean_objective": 0.9,
+    "residual_rad_at_lag": 0.0008,
+    "residual_rad_at_lag_mean": 0.0132,
+    "n_compared": 613,
+    "residual_rad_at_zero": 0.0466,
+    "residual_rad_at_zero_mean": 0.0503,
+    "xcorr_command_to_ur_joint_states_s": 0.905,
+    "xcorr_command_to_ur_joint_states_r": 0.606,
+    "xcorr_command_to_tcp_pose_s": 0.41,
+    "xcorr_command_to_tcp_pose_r": 0.604,
+    "xcorr_tcp_pose_to_ur_joint_states_s": 0.45,
+    "xcorr_tcp_pose_to_ur_joint_states_r": 0.467
+   },
+   {
+    "take": "take_45_20260914_171534",
+    "ur_joint_states_lag_s": 0.9,
+    "ur_joint_states_lag_s_mean_objective": 0.9,
+    "residual_rad_at_lag": 0.0006,
+    "residual_rad_at_lag_mean": 0.0091,
+    "n_compared": 821,
+    "residual_rad_at_zero": 0.0505,
+    "residual_rad_at_zero_mean": 0.0509,
+    "xcorr_command_to_ur_joint_states_s": 0.9,
+    "xcorr_command_to_ur_joint_states_r": 0.688,
+    "xcorr_command_to_tcp_pose_s": 0.405,
+    "xcorr_command_to_tcp_pose_r": 0.689,
+    "xcorr_tcp_pose_to_ur_joint_states_s": 0.4,
+    "xcorr_tcp_pose_to_ur_joint_states_r": 0.659
+   },
+   {
+    "take": "take_46_20260914_171603",
+    "ur_joint_states_lag_s": 0.9,
+    "ur_joint_states_lag_s_mean_objective": 0.9,
+    "residual_rad_at_lag": 0.0006,
+    "residual_rad_at_lag_mean": 0.0081,
+    "n_compared": 895,
+    "residual_rad_at_zero": 0.034,
+    "residual_rad_at_zero_mean": 0.0454,
+    "xcorr_command_to_ur_joint_states_s": 0.91,
+    "xcorr_command_to_ur_joint_states_r": 0.787,
+    "xcorr_command_to_tcp_pose_s": 0.445,
+    "xcorr_command_to_tcp_pose_r": 0.694,
+    "xcorr_tcp_pose_to_ur_joint_states_s": 0.43,
+    "xcorr_tcp_pose_to_ur_joint_states_r": 0.605
+   },
+   {
+    "take": "take_47_20260914_171640",
+    "ur_joint_states_lag_s": 0.9,
+    "ur_joint_states_lag_s_mean_objective": 0.895,
+    "residual_rad_at_lag": 0.0006,
+    "residual_rad_at_lag_mean": 0.0112,
+    "n_compared": 832,
+    "residual_rad_at_zero": 0.0489,
+    "residual_rad_at_zero_mean": 0.0569,
+    "xcorr_command_to_ur_joint_states_s": 0.905,
+    "xcorr_command_to_ur_joint_states_r": 0.559,
+    "xcorr_command_to_tcp_pose_s": 0.405,
+    "xcorr_command_to_tcp_pose_r": 0.516,
+    "xcorr_tcp_pose_to_ur_joint_states_s": 0.495,
+    "xcorr_tcp_pose_to_ur_joint_states_r": 0.526
+   },
+   {
+    "take": "take_48_20260914_171732",
+    "ur_joint_states_lag_s": 0.9,
+    "ur_joint_states_lag_s_mean_objective": 0.9,
+    "residual_rad_at_lag": 0.0007,
+    "residual_rad_at_lag_mean": 0.0097,
+    "n_compared": 591,
+    "residual_rad_at_zero": 0.0401,
+    "residual_rad_at_zero_mean": 0.0556,
+    "xcorr_command_to_ur_joint_states_s": 0.915,
+    "xcorr_command_to_ur_joint_states_r": 0.87,
+    "xcorr_command_to_tcp_pose_s": 0.38,
+    "xcorr_command_to_tcp_pose_r": 0.743,
+    "xcorr_tcp_pose_to_ur_joint_states_s": 0.5,
+    "xcorr_tcp_pose_to_ur_joint_states_r": 0.693
+   },
+   {
+    "take": "take_49_20260914_171812",
+    "ur_joint_states_lag_s": 0.9,
+    "ur_joint_states_lag_s_mean_objective": 0.9,
+    "residual_rad_at_lag": 0.0006,
+    "residual_rad_at_lag_mean": 0.0081,
+    "n_compared": 1181,
+    "residual_rad_at_zero": 0.053,
+    "residual_rad_at_zero_mean": 0.056,
+    "xcorr_command_to_ur_joint_states_s": 0.91,
+    "xcorr_command_to_ur_joint_states_r": 0.768,
+    "xcorr_command_to_tcp_pose_s": 0.38,
+    "xcorr_command_to_tcp_pose_r": 0.458,
+    "xcorr_tcp_pose_to_ur_joint_states_s": 0.8,
+    "xcorr_tcp_pose_to_ur_joint_states_r": 0.47
+   },
+   {
+    "take": "take_50_20260914_171849",
+    "ur_joint_states_lag_s": 0.9,
+    "ur_joint_states_lag_s_mean_objective": 0.9,
+    "residual_rad_at_lag": 0.0008,
+    "residual_rad_at_lag_mean": 0.004,
+    "n_compared": 999,
+    "residual_rad_at_zero": 0.0516,
+    "residual_rad_at_zero_mean": 0.0561,
+    "xcorr_command_to_ur_joint_states_s": 0.895,
+    "xcorr_command_to_ur_joint_states_r": 0.853,
+    "xcorr_command_to_tcp_pose_s": 0.405,
+    "xcorr_command_to_tcp_pose_r": 0.641,
+    "xcorr_tcp_pose_to_ur_joint_states_s": 0.53,
+    "xcorr_tcp_pose_to_ur_joint_states_r": 0.604
+   },
+   {
+    "take": "take_51_20260914_171924",
+    "ur_joint_states_lag_s": 0.9,
+    "ur_joint_states_lag_s_mean_objective": 0.9,
+    "residual_rad_at_lag": 0.0008,
+    "residual_rad_at_lag_mean": 0.0086,
+    "n_compared": 959,
+    "residual_rad_at_zero": 0.0547,
+    "residual_rad_at_zero_mean": 0.0653,
+    "xcorr_command_to_ur_joint_states_s": 0.925,
+    "xcorr_command_to_ur_joint_states_r": 0.753,
+    "xcorr_command_to_tcp_pose_s": 1.395,
+    "xcorr_command_to_tcp_pose_r": 0.491,
+    "xcorr_tcp_pose_to_ur_joint_states_s": 0.53,
+    "xcorr_tcp_pose_to_ur_joint_states_r": 0.399
+   },
+   {
+    "take": "take_52_20260914_171957",
+    "ur_joint_states_lag_s": 0.9,
+    "ur_joint_states_lag_s_mean_objective": 0.9,
+    "residual_rad_at_lag": 0.0006,
+    "residual_rad_at_lag_mean": 0.0119,
+    "n_compared": 961,
+    "residual_rad_at_zero": 0.0499,
+    "residual_rad_at_zero_mean": 0.0604,
+    "xcorr_command_to_ur_joint_states_s": 0.9,
+    "xcorr_command_to_ur_joint_states_r": 0.757,
+    "xcorr_command_to_tcp_pose_s": 0.405,
+    "xcorr_command_to_tcp_pose_r": 0.675,
+    "xcorr_tcp_pose_to_ur_joint_states_s": 0.5,
+    "xcorr_tcp_pose_to_ur_joint_states_r": 0.624
+   },
+   {
+    "take": "take_53_20260914_172033",
+    "ur_joint_states_lag_s": 0.9,
+    "ur_joint_states_lag_s_mean_objective": 0.9,
+    "residual_rad_at_lag": 0.0007,
+    "residual_rad_at_lag_mean": 0.0132,
+    "n_compared": 651,
+    "residual_rad_at_zero": 0.0448,
+    "residual_rad_at_zero_mean": 0.0668,
+    "xcorr_command_to_ur_joint_states_s": 0.91,
+    "xcorr_command_to_ur_joint_states_r": 0.699,
+    "xcorr_command_to_tcp_pose_s": 0.41,
+    "xcorr_command_to_tcp_pose_r": 0.67,
+    "xcorr_tcp_pose_to_ur_joint_states_s": 0.5,
+    "xcorr_tcp_pose_to_ur_joint_states_r": 0.625
+   },
+   {
+    "take": "take_54_20260914_172116",
+    "ur_joint_states_lag_s": 0.9,
+    "ur_joint_states_lag_s_mean_objective": 0.9,
+    "residual_rad_at_lag": 0.0009,
+    "residual_rad_at_lag_mean": 0.0117,
+    "n_compared": 1064,
+    "residual_rad_at_zero": 0.0422,
+    "residual_rad_at_zero_mean": 0.0483,
+    "xcorr_command_to_ur_joint_states_s": 0.88,
+    "xcorr_command_to_ur_joint_states_r": 0.691,
+    "xcorr_command_to_tcp_pose_s": 0.41,
+    "xcorr_command_to_tcp_pose_r": 0.586,
+    "xcorr_tcp_pose_to_ur_joint_states_s": 0.5,
+    "xcorr_tcp_pose_to_ur_joint_states_r": 0.468
+   },
+   {
+    "take": "take_55_20260914_172155",
+    "ur_joint_states_lag_s": 0.9,
+    "ur_joint_states_lag_s_mean_objective": 0.895,
+    "residual_rad_at_lag": 0.0006,
+    "residual_rad_at_lag_mean": 0.0136,
+    "n_compared": 569,
+    "residual_rad_at_zero": 0.0537,
+    "residual_rad_at_zero_mean": 0.0621,
+    "xcorr_command_to_ur_joint_states_s": 0.905,
+    "xcorr_command_to_ur_joint_states_r": 0.678,
+    "xcorr_command_to_tcp_pose_s": 0.38,
+    "xcorr_command_to_tcp_pose_r": 0.651,
+    "xcorr_tcp_pose_to_ur_joint_states_s": 0.03,
+    "xcorr_tcp_pose_to_ur_joint_states_r": 0.579
+   }
+  ]
+ },
  "depth_cameras": {
   "cam1": {
    "group_attrs": {