Wheel based quadruped racing improves with learned residual dynamics

RACER: Residual-Adaptive Closed-Loop Estimation for Sampling-Based Planning in Wheeled-Quadruped Racing

Robotics

Summary

Controlling wheeled four-legged robots to race quickly and accurately is hard because their movements are complex and influenced by many factors. The authors created a new control framework called RACER that combines a traditional planning method with a learned model that corrects for tracking errors, plus a low-level controller to follow velocity commands. They developed a way to train this learned correction model efficiently by first using lots of simulated data and then fine-tuning with a small amount of real-world data. Their experiments show that this approach helps the robot handle sharp turns better and perform more reliably across different surfaces.

What this means in practice

  • For robotics engineers: Develop high-speed wheeled quadruped controllers that maintain accurate path tracking during aggressive racing maneuvers on varied terrain.
  • For autonomous vehicle software teams: Improve planning and tracking control in vehicles with complex or hybrid dynamics by integrating learned residual models for better real-world performance.

Authors

Yuxiang Liu, Marla Eisman, Lizhi Yang, Aaron Ames, Francesco Borrelli

Abstract

We present RACER, a hierarchical control framework for wheel-based quadruped racing that combines an MPPI planner with a learned residual dynamics model and a low-level RL velocity tracker. The planner augments a nominal unicycle kinematic model with a neural residual term to capture the closed-loop tracking behavior of the RL policy. To train this residual model under limited real-world data, we propose Low-Rank Residual Adaptation (LoRRA), a two-stage approach that pre-trains on large-scale simulation data for broad coverage and then fine-tunes on a small real-world dataset with a low-rank constraint. In simulation, we empirically validate our engineering choices by showing (A) Residual dynamics improve the overall performance of our pipeline by capturing the tracking error of RL velocity tracker at high-speed cornering. (B) Residual dynamics trained with both source-domain and target-domain data gives racing performance significantly better than the residual dynamics trained with only target-domain data. (C) Low-rank constraint at target-domain adaptation gives higher success rates and higher performance than full-tune and from-scratch when domain gap in ground coefficient or joint gain increases.
View PDFOpen arXiv