Robot motion model improves action choices and speed under uncertainty
Energy-Score Stabilized Attentive Mixture-Density Policy: A Lightweight Approach to Multimodal Robot Motion
Robotics
Summary
Robots often face situations where multiple ways to move are possible and good. The authors present a new method that teaches robots to pick from many possible actions quickly using less data and with clearer predictions. Their approach uses a special kind of probability model that avoids common training problems and focuses attention on important visual features. Tests show it helps a robotic arm avoid obstacles better and catch moving objects more reliably.
What this means in practice
- •For robot motion planners: Generate multiple safe robot paths efficiently with better handling of uncertain visual inputs and quick decision-making.
- •For industrial robot operators: Improve picking and obstacle avoidance tasks for robotic arms working in environments with variable and moving objects.$Commercial implications: This enables more reliable and faster robotic arm controllers for manufacturing and logistics, improving productivity and safety.
Authors
Sachiya Fujita, Hideyuki Ichiwara, Shigeki Sugano, Tetsuya Ogata
Abstract
For practical motion generation, it is essential to handle multimodality (multiple actions can be correct for the same observation) while maintaining low inference cost and strong generalization from limited demonstrations. We introduce an attentive mixture-density policy trained with the Energy Score that represents future visuomotor trajectories as explicit probability distributions with a single forward pass. Concretely, we output Gaussian mixture parameters for joint commands (and optionally attention points) and optimize them with the Energy Score (computed as a sum of 1D CRPS under a diagonal covariance assumption) instead of the negative log-likelihood (NLL) to avoid MDN singularities and gradient-scale instabilities. Our architecture couples explicit spatial attention with mixture-density outputs, enabling robust multimodal behavior selection under visual noise and small-data regimes. On an 8-DOF manipulator, the method generates distinct avoidance trajectories and stabilizes learning in obstacle avoidance. Moreover, in a randomized moving-object grasping task, where spatiotemporal multimodality and fast timing decisions are required, the proposed model outperforms Diffusion Policy by 15-25% under several unseen conditions, demonstrating superior generalization and real-time feasibility.