GPU powered robotic control learns and adapts safely in real time
Planning to Learn: Real-Time Robust Adaptive Control on the GPU with Guaranteed Future Learning
RoboticsMachine Learning
Summary
Robots often struggle to stay safe while learning about new or uncertain situations because existing control methods are too cautious and slow. This paper presents a new way to plan robot movements that forecasts how much the robot will learn and adjusts control to be less conservative but still safe. The method uses fast GPU computing to work in real time, even on robots with many parts, like a four-legged robot or a drone. The authors showed their approach works well both in simulations and on a real flying drone that adapts on the fly.
What this means in practice
- •For robotics engineers: Implement fast adaptive controllers for high-dimensional robots that learn uncertain dynamics safely while operating in real time.
- •For drone developers: Stabilize drones online by learning dynamics parameters during flight to improve safety and performance without stopping operation.
Authors
Anutam Srinivasan, Siddhant Panse, Glen Chou
Abstract
Robust adaptive control is vital for keeping robots safe while actively reducing uncertainty in unfamiliar settings. However, traditional methods are often conservative since they fail to account for guaranteed future parameter learning during planning and struggle to scale to high-dimensional systems. In this work, we present A-SLS, a learning-aware robust adaptive model predictive control (RAMPC) framework based on system-level synthesis. A-SLS forecasts future parameter uncertainty and preserves correlations between process disturbances, parameter error, and estimate updates to construct tight, guaranteed reachable sets that account for future information gain. We also develop a GPU-accelerated A-SLS solver to enable real-time performance. Across nonlinear robotic systems, A-SLS reduces reachable-tube conservativeness, maintains real-time performance, and scales to whole-body control for a quadruped with 61 states, 8 uncertain parameters, and 12 control inputs. We further demonstrate the approach on hardware, where a physical Crazyflie drone learns an uncertain dynamics parameter online to stabilize on the fly.