Safe Adaptive Cruise Control Under Perception Uncertainty: A Deep Ensemble and Conformal Tube Model Predictive Control Approach

TL;DR

This paper introduces a robust adaptive cruise control method that combines deep ensemble neural networks with conformal prediction to quantify perception uncertainties, ensuring safety and reliability in autonomous driving.

Contribution

It presents a novel integration of deep ensemble and conformal prediction with model predictive control for safety-critical autonomous driving applications.

Findings

Effective speed tracking and safe distance maintenance demonstrated in simulations.

Robustness to Out-Of-Distribution scenarios confirmed.

Uncertainty quantification improves safety in perception-based control.

Abstract

Autonomous driving heavily relies on perception systems to interpret the environment for decision-making. To enhance robustness in these safety critical applications, this paper considers a Deep Ensemble of Deep Neural Network regressors integrated with Conformal Prediction to predict and quantify uncertainties. In the Adaptive Cruise Control setting, the proposed method performs state and uncertainty estimation from RGB images, informing the downstream controller of the DNN perception uncertainties. An adaptive cruise controller using Conformal Tube Model Predictive Control is designed to ensure probabilistic safety. Evaluations with a high-fidelity simulator demonstrate the algorithm's effectiveness in speed tracking and safe distance maintaining, including in Out-Of-Distribution scenarios.