Multi-Agent Reachability Calibration with Conformal Prediction

TL;DR

This paper presents a method combining conformal prediction and reachability analysis to provide probabilistic safety guarantees for autonomous agents predicting other agents' trajectories, validated in driving and Boeing vehicle experiments.

Contribution

It introduces a novel approach integrating quantile regression, conformal prediction, and reachability analysis for probabilistic safety assurances in autonomous planning.

Findings

Generated safe, probabilistic prediction sets for autonomous agents.

Validated safety guarantees in simulation and real-world Boeing vehicle experiments.

Demonstrated improved safety certification for planning algorithms.

Abstract

We investigate methods to provide safety assurances for autonomous agents that incorporate predictions of other, uncontrolled agents' behavior into their own trajectory planning. Given a learning-based forecasting model that predicts agents' trajectories, we introduce a method for providing probabilistic assurances on the model's prediction error with calibrated confidence intervals. Through quantile regression, conformal prediction, and reachability analysis, our method generates probabilistically safe and dynamically feasible prediction sets. We showcase their utility in certifying the safety of planning algorithms, both in simulations using actual autonomous driving data and in an experiment with Boeing vehicles.