Robot world models improve task success using action perturbation insights
RAP-WAM: Learning World-Action Models through Predictive Rehearsal and Action Perturbations
Artificial Intelligence
Summary
Many robots learn to plan by predicting what will happen next from video data. This paper introduces a method where the robot practices predicting future outcomes while slightly changing its actions to see how predictions change. This helps the robot learn which actions matter most for achieving goals without needing extra training data or complex evaluation networks. The approach improved robot task success in both simulations and real-world tests compared to existing methods.
What this means in practice
- •For robotics engineers: Improve robot task success rates by integrating models that learn from video predictions and action influences without extra training data.
- •For automation system developers: Develop more reliable control systems by using action perturbations to guide policy learning in robots handling diverse tasks.
Authors
Yuanpeng He, Ceyao Zhang, Fangjing Li, Kefei Zhu, Lijian Li, Yueheng Li, Tianjia He, Wenpin Jiao, Zhi Jin, Yaodong Yang, Yuanpei Chen
Abstract
Recent world action models draw on video generation priors to jointly predict future observations and robot actions. Adapting these priors to control requires determining how future predictions should guide policy learning. RAP-WAM combines predictive Rehearsal and Action Perturbations (RAP) with future-supervised queries and Coupled Action Projection for policy learning. Predictive rehearsal trains on model-generated inputs while retaining recorded targets, and accumulates recurring errors into a spatial prior. We perturb actions and measure changes in video predictions. Action-response-based spatiotemporal credit assignment combines these responses, prediction error, and the rehearsal prior to allocate visual supervision; spatially aggregating the responses determines supervision intensity across future action steps. No separate scoring network or deployment-time probe is required. The base learns from a video-pretrained backbone without intermediate pretraining on external embodied data. We evaluate the fixed base and its agent-refined, task-conditioned system. On 42 RoboDojo tasks, their average success rates are 12.85% and 15.62%, with Scores of 17.86 and 20.59. Across 15 real-robot tasks, the complete system achieves 89.00% task-macro-average success, exceeding the strongest of five baselines by 37.67 percentage points. Across 315 fixed-output simulation windows, 45.21% of response-top20 positions fall outside error-top20, revealing information beyond prediction error.