B-spline Policy: Accelerating Manipulation Policies via B-spline Action Representations
For robot manipulation, BSP offers a practical way to accelerate policy execution without sacrificing performance, though the improvement is incremental over existing action chunking methods.
B-spline Policy (BSP) represents robot actions as continuous B-spline curves instead of discrete-time chunks, enabling smoother and faster execution. In experiments, BSP significantly reduces task completion time while maintaining high success rates compared to baselines.
In this work, we present B-spline Policy (BSP), an action representation designed for accelerating robot manipulation policies. Rather than predicting discrete-time action chunks, BSP parameterizes actions as continuous B-spline curves defined by a set of knots and control points. This representation yields smooth, time-continuous trajectories that can be temporally scaled and executed by low-level controllers at higher frequencies and speeds. We show that B-spline-parameterized actions can be seamlessly integrated into standard policy learning pipelines by directly predicting B-spline parameters. Experiments on simulated and real-world tasks demonstrate that BSP significantly reduces task completion time, achieving substantial improvements over baseline methods while maintaining strong success rates. More results: https://b-spline-policy.github.io