* training/bc/collect.py: --use-webots-preset now picks the
drive-matched variant. Mecanum drives get HERDING_MEC_WEBOTS
(with the Webots-calibrated strafe efficiency and bleed) so the
collected demos reflect the imperfect physical mecanum the
deployed policy will see. Differential drives still use
HERDING_WEBOTS (no behaviour change there).
* training/rl/train.py: mecanum fine-tune now *unconditionally*
applies the HERDING_MEC_WEBOTS robot config to the PPO env (the
policy must update against the same imperfect kinematics it
deploys on). Diff fine-tune unchanged.
To retrain a mecanum policy end-to-end against the new proto:
python -m training.bc.collect --drive-mode mecanum --world field \
--use-webots-preset \
--out training/bc/demos_mecanum_field_v2.npz
python -m training.bc.pretrain --demos training/bc/demos_mecanum_field_v2.npz \
--out training/runs/bc_mecanum_field_v2 ...
python -m training.rl.train --bc training/runs/bc_mecanum_field_v2 \
--out training/runs/rl_mecanum_field_v2 \
--drive-mode mecanum --world field --use-webots-preset
The same flow for field_round / mecanum/round.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Adds RecurrentPPO-based training as an alternative to MLP+frame-stack.
The LSTM gives the policy unbounded temporal memory, addressing the
partial-obs failure mode of the 140° Webots LiDAR (tracker briefly
empties when the dog turns; sporadic phantom tracks confuse decisions).
* training/rl/train_lstm.py: from-scratch RecurrentPPO trainer (no BC
init, no KL term since there's no reference). Uses HERDING_WEBOTS
preset so the obs distribution matches deployment.
* training/eval.py: auto-detects RecurrentPPO zips, maintains LSTM
hidden state across steps, resets between episodes.
* controllers/shepherd_dog/policy_loader.py: PolicyHandle supports
recurrent policies — state managed inside, reset_recurrent() exposed.
Result on diff/field after 3M steps:
- Gym (default 360°): 69% avg success across n=1..10
- Gym (HERDING_WEBOTS preset, training env): 2% — penning 3-4/5 but
rarely all 5
- Webots LiDAR 140°: 0/5 (same wall as DAgger and v1 policies)
Conclusion: architectural changes (LSTM vs MLP) don't close the
perception sim-to-real gap. The gym LiDAR sim doesn't faithfully
reproduce Webots phantom-track distribution; any policy trained on the
gym proxy fails to handle real Webots phantoms regardless of
architecture. Closing this gap requires either modeling Webots phantom
patterns in the gym sim (multi-day work) or Webots-in-the-loop
training (very slow). See memory/lstm_results.md for details.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Today's session worked across the full Webots delivery stack — found and
fixed a cluster of bugs blocking the BC/RL transfer, then explored
training-side mitigations for the residual perception gap.
Bug fixes:
- Makefile FP_RATE default 2.0 → 0.0: BC demos used fp_rate=0 but RL
fine-tune defaulted to fp_rate=2, poisoning the BC obs distribution
and stalling PPO at 0% success across 1.46M+ steps.
- controllers/{shepherd_dog,sheep}/runtime.ini: Webots was launching
controllers under system python3 (no numpy) and they were crashing
silently. Pinned to the conda tir env.
- herding/config.py HERDING_WEBOTS preset: pen_latch_depth 0.5 → 2.0,
max_new_tracks_per_step 3 → 1, static_reject 0.8 → 1.2. Stops phantom
FPs near the gate from latching as permanently-penned tracks.
- herding/perception/sheep_tracker.py: penned tracks now decay at
forget_steps × 8 instead of living forever. Adds get_positions
min_freshness filter for deploy-time use.
Training/eval matches deployment:
- training/bc/collect.py: --dagger-policy flag for DAgger rollouts
(policy drives, teacher labels) + --use-webots-preset for matched
140° tracker + DR config.
- controllers/shepherd_dog/shepherd_dog.py: scan-fallback (0, 0.6) when
BC/RL sees empty sheep_positions — recovers from FOV gaps.
Tooling:
- tools/dagger_round.sh: one-shot DAgger round (collect + concat + bc).
- tools/webots_sweep_gt.sh: full sweep with HERDING_USE_GT=1 for the
perception-gap diagnosis matrix.
- protos/ShepherdDog360.proto: 360° FOV variant for the FOV-ablation
comparison. Canonical proto stays at 140° per project spec.
Artifacts: v1 BC/RL policies for all 4 (drive × world) combos trained
in clean gym (success: diff/field 90-100%, diff/round 58%, mec/field
60-100%, mec/round 50-100%). DAgger r1/r2 BCs for diff/field show
12%→38% progression on gym HERDING_WEBOTS proxy but did not close
to actual Webots LiDAR (0/5 throughout). Next: LSTM policy or
learned tracker per the project-state memory.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Johnny Fernandes