First, Learn What You Don't Know: Active Information Gathering for Driving at the Limits of Handling
Alexander Davydov, Franck Djeumou, Marcus Greiff, Makoto Suminaka, Michael Thompson, John Subosits, Thomas Lew
AI summary
Problem
Online model adaptation is often too slow to prevent destabilization when controlling unstable nonlinear systems like drifting vehicles. The paper addresses how to quickly and safely identify uncertain dynamics to ensure reliable control.
Approach
The authors combine a Bayesian last-layer meta-learning dynamics model with an optimal control framework that actively plans pre-deployment trajectories to maximize information gain and rapidly refine the model.
Key results
- Active information gathering enables reliable control of a vehicle at the edge of stability during dynamic drifting.
- Online adaptation alone is insufficient for zero-shot control and can lead to large transient errors or spin-outs.
- Information-rich trajectories can be safely collected off-track to rapidly reduce model uncertainty.
- The framework successfully identifies nonlinear dynamics on a physical Toyota Supra with minimal experimentation.
Why it matters
This framework provides a practical method for safely deploying data-driven model predictive control in high-risk autonomous driving and robotics applications where immediate model accuracy is critical.
Abstract
Combining data-driven models that adapt online and model predictive control (MPC) has enabled effective control of nonlinear systems. However, when deployed on unstable systems, online adaptation may not be fast enough to ensure reliable simultaneous learning and control. For example, a con- troller on a vehicle executing highly dynamic maneuvers—such as drifting to avoid an obstacle—may push the vehicle’s tires to their friction limits, destabilizing the vehicle and allowing modeling errors to quickly compound and cause a loss of control. To address this challenge, we present an active information gathering framework for identifying vehicle dynamics as quickly as possible. We propose an expressive vehicle dynamics model that leverages Bayesian last-layer meta-learning to enable rapid online adaptation. The model’s uncertainty estimates are used to guide informative data collection and quickly improve the model prior to deployment. Dynamic drifting experiments on a Toyota Supra show that (i) the framework enables reliable control of a vehicle at the edge of stability, (ii) online adaptation alone may not suffice for zero-shot control and can lead to undesirable transient errors or spin-outs, and (iii) active data collection helps achieve reliable performance.