stage1_pickplace_om@v1 — Stage-1 pick-and-place — observation-modality (OM) control
The same tabletop pick-and-place viewed through heterogeneous observation modalities: a vision-only condition masks proprioception and force-torque, while the augmented condition adds synthesised force-torque and proprioception. The OM axis was never fired at the gate and ADR-026 closed its promotion. The task is retained as a Tier-1 control for observation-modality claims.
Tier 1 — controls · Status: CONTROL · Suite: CHAMBER-Bench v1.0 (adr/ADR-027-chamber-bench-v1-protocol.md)
Spaces
- Action space. Dict over ego (panda_wristcam) + partner (fetch).
- Observation. rgb+depth+state_dict; the vision-only condition masks proprio + force-torque via Stage1OMChannelFilter, the vision+FT+proprio condition enables synthesised force-torque (condition_id selects).
- Simulator backend.
maniskill3· Agents. 2
Stress channel
None — no force coupling channel.
Axis validity
Per-cell status under the validity matrix (adr/ADR-027-chamber-bench-v1-protocol.md):
| AS | OM | CR | CM | PF | SA |
|---|---|---|---|---|---|
untested |
untested |
untested |
untested |
untested |
untested |
Industrial analogue
A camera-guided picking station versus one with joint encoders and a wrist force-torque sensor — the standard sensing-tier trade-off in industrial picking, here kept as a control rather than a cooperation measurement.
Evidence
How to run one episode
import chamber.tasks
env = chamber.tasks.make("stage1_pickplace_om")
obs, info = env.reset(seed=0)
# Requires a SAPIEN-capable host (Tier-2).
Notes
The OM axis was never fired at the gate — ADR-026 closed its promotion, so the OM cell stays untested and the task is retained as a Tier-1 control (ADR-027 §Tier ladder). Default condition is 'stage1_pickplace_vision_only'; pass condition_id='stage1_pickplace_vision_plus_force_torque_plus_proprio' for the FT-augmented cell.