Differentiable layer improves vehicle control safety and performance
A Differentiable Chassis Safety Layer for Intelligent Vehicle Motion Control
Robotics
Summary
Keeping self-driving cars safe while making them drive smoothly is tricky because safety rules can clash with the driving goals. The authors created a new way to mix safety rules directly into the car’s driving decisions so the car learns how to be both safe and effective. Their system uses math to make sure commands to the car don’t break physical limits, and it learns in two steps: copying experts and then adjusting to risky situations. Tests show their approach helps cars handle emergencies better while staying stable.
What this means in practice
- •For autonomous vehicle software teams: Combine safety constraints with control policies for better real-time vehicle motion decisions.
- •For automotive control engineers: Develop control systems that maintain vehicle handling stability during emergency maneuvers with explicit physical constraints.
Authors
Lu Xiong, Xuge Hu, Yufei Song, Weidong Liu, Chen Tang
Abstract
Learning-based vehicle motion control requires coordination between diverse task objectives and state-dependent chassis safety constraints. Conventional cascaded architectures may compromise task performance when safety filters modify nominal commands independently of policy learning. This article proposes a differentiable chassis safety layer that modularly integrates task-oriented control with explicit physical constraints. Actuator limits, individual-tire force constraints, and vehicle handling envelopes are incorporated into sequential quadratic programs. Each projected command propagates the predicted vehicle dynamics and updates subsequent constraints, enabling temporal gradient coupling through implicit differentiation. A two-stage learning strategy combines expert imitation with risk-guided supervision and conditional intervention guidance to coordinate nominal control decisions with chassis safety constraints. Closed-loop evaluations under unseen scenarios, emergency maneuvers, and ablation studies demonstrate that the proposed method improves safety-performance coordination and exploits vehicle dynamic potential while maintaining handling stability.