T-PFC — Trajectory-Optimized Perturbation Feedback Control

Published in IEEE Robotics and Automation Letters and presented at IEEE/RSJ IROS 2019, Macau, China.

Overview

Robot motion controllers typically handle trajectory generation and disturbance rejection as separate concerns. T-PFC unifies them: it jointly optimizes the nominal reference trajectory and the perturbation feedback gains, so the planned path is inherently shaped around the system’s ability to handle uncertainty.

The key insight is that a trajectory that looks optimal in the nominal (deterministic) sense may be brittle under real-world noise, while a slightly suboptimal nominal path may enable a much more robust closed-loop response. T-PFC captures this trade-off explicitly in the optimization objective.

Method

Results

IEEE Paper