What it is
Robots improve a shared manipulation policy from fleet execution logs by retrieving successful peer trajectories near failure contexts, avoiding new expert corrections. Instead of relying on teleoperated corrections, failed executions query a shared buffer of successful fleet trajectories; retrieved segments are upweighted as supervision to fine-tune one shared policy asynchronously.
Key results
- Phone packaging: 79.2% to 98.3% success with 96 robots
- Across 5 tasks: 93.3-99.2% success over 120 eval trials per task
- Removing failure-guided retrieval drops phone packaging from 86.7% to 64.2%
Shared autonomous experience can materially improve deployed robot reliability without scaling expert corrective supervision.