Evaluation of MPC-based Imitation Learning for Human-like Autonomous Driving

TL;DR

This paper evaluates the integration of imitation learning with differentiable MPC for human-like autonomous driving, demonstrating improved robustness and alignment with human driving behaviors through experimental validation.

Contribution

It introduces a hierarchical learning-based policy combined with MPC, and shows how augmenting behavioral cloning with closed-loop training enhances performance.

Findings

Imitative policies closely match human driving style.

Closed-loop training improves robustness of the learned policy.

The approach performs well in lane keeping tasks in simulation.

Abstract

This work evaluates and analyzes the combination of imitation learning (IL) and differentiable model predictive control (MPC) for the application of human-like autonomous driving. We combine MPC with a hierarchical learning-based policy, and measure its performance in open-loop and closed-loop with metrics related to safety, comfort and similarity to human driving characteristics. We also demonstrate the value of augmenting open-loop behavioral cloning with closed-loop training for a more robust learning, approximating the policy gradient through time with the state space model used by the MPC. We perform experimental evaluations on a lane keeping control system, learned from demonstrations collected on a fixed-base driving simulator, and show that our imitative policies approach the human driving style preferences.