Learning Autonomous Vehicle Safety Concepts from Demonstrations

TL;DR

This paper introduces a data-driven methodology for autonomous vehicle safety that learns behavioral assumptions from demonstrations and synthesizes safety concepts using control theory techniques, improving interpretability and verification.

Contribution

It presents a novel approach combining data-driven learning with control theory to develop AV safety models based on behavioral assumptions derived from real-world data.

Findings

Learned safety concept effectively evaluates real-world driving logs.

Compared favorably to existing baseline safety models.

Demonstrated approach in highway traffic-weaving scenario.

Abstract

Evaluating the safety of an autonomous vehicle (AV) depends on the behavior of surrounding agents which can be heavily influenced by factors such as environmental context and informally-defined driving etiquette. A key challenge is in determining a minimum set of assumptions on what constitutes reasonable foreseeable behaviors of other road users for the development of AV safety models and techniques. In this paper, we propose a data-driven AV safety design methodology that first learns ``reasonable'' behavioral assumptions from data, and then synthesizes an AV safety concept using these learned behavioral assumptions. We borrow techniques from control theory, namely high order control barrier functions and Hamilton-Jacobi reachability, to provide inductive bias to aid interpretability, verifiability, and tractability of our approach. In our experiments, we learn an AV safety concept using demonstrations collected from a highway traffic-weaving scenario, compare our learned concept to existing baselines, and showcase its efficacy in evaluating real-world driving logs.