Autonomous Driving With Perception Uncertainties: Deep-Ensemble Based Adaptive Cruise Control

Xiao Li, H Eric Tseng, Anouck Girard, Ilya Kolmanovsky

Autonomous driving depends on perception systems to understand the environment and to inform downstream decision-making. While advanced perception systems utilizing black-box Deep Neural Networks (DNNs) demonstrate human-like comprehension, their unpredictable behavior and lack of interpretability may hinder their deployment in safety critical scenarios. In this paper, we develop an Ensemble of DNN regressors (Deep Ensemble) that generates predictions with quantification of prediction uncertainties. In the scenario of Adaptive Cruise Control (ACC), we employ the Deep Ensemble to estimate distance headway to the lead vehicle from RGB images and enable the downstream controller to account for the estimation uncertainty. We develop an adaptive cruise controller that utilizes Stochastic Model Predictive Control (MPC) with chance constraints to provide a probdbilistic safety guarantee. We evaluate our ACC algorithm using a high-fidelity traffic simulator and a real-world traffic dataset and demonstrate the ability of the proposed approach to effect speed tracking and car following while maintaining a safe distance headway. The out-of-distribution scenarios are also examined.

[arXiv]

Adaptive Cruise Control using Deep Ensemble under in-distribution and out-of-distribution perception

The perception model uses a heterogeneous set of Deep Neural Networks (Deep Ensembles) to predict the distance headway of the lead vehicle and generate estimation uncertainties from RGB images. A Stochastic Model Predictive Controller (SMPC) is proposed for Adaptive Cruise Control (ACC), ensuring probabilistic safety by accounting for these uncertainties. The Deep Ensemble is trained on a dataset collected in sunny weather with a black sedan as the lead vehicle. Observations similar to the training dataset are classified as in-distribution, while those differing from the training dataset (e.g., images collected in heavy rain with poor lighting or featuring an unfamiliar lead vehicle) are considered out-of-distribution (OOD). Our algorithm captures OOD perception by indicating large estimation uncertainties, which inform the SMPC to execute safer maneuvers.

Uncertainty Quantification of distance headway estimation using Deep Ensemble:

Adaptive Cruise Control using SMPC + Deep Ensemble (in-distribution perception):

Adaptive Cruise Control using real-world vehicle trajectory (High-D dataset):

Adaptive Cruise Control using SMPC + Deep Ensemble (Out-Of-Distribution perception):