Enhancing Safety in Mixed Traffic: Learning-Based Modeling and Efficient Control of Autonomous and Human-Driven Vehicles

TL;DR

This work tackles safety in mixed traffic by explicitly accounting for HV uncertainty in longitudinal car-following. It introduces a learning-based HV model that hybrids a time-delayed first-principles ARX nominal model with a Gaussian process correction, enabling both accurate velocity prediction and uncertainty quantification. A sparse GP-enhanced MPC (GP-MPC) with an adaptive, chance-constrained safe-distance constraint demonstrates improved safety margins and higher platoon speeds, while remaining computationally feasible in real time (approximately a 4.6% overhead and ~100× faster than the authors\' earlier GP-MPC iterations). Field data and field trials validate the approach, showing robust HV handling under emergency braking and WLTP driving profiles, with real-time performance achievable on standard hardware. The results indicate a practical path toward uncertainty-aware AV-HV interaction in mixed-traffic environments, highlighting both safety gains and efficiency improvements for autonomous platoons.

Abstract

With the increasing presence of autonomous vehicles (AVs) on public roads, developing robust control strategies to navigate the uncertainty of human-driven vehicles (HVs) is crucial. This paper introduces an advanced method for modeling HV behavior, combining a first-principles model with Gaussian process (GP) learning to enhance velocity prediction accuracy and provide a measurable uncertainty. We validated this innovative HV model using real-world data from field experiments and applied it to develop a GP-enhanced model predictive control (GP-MPC) strategy. This strategy aims to improve safety in mixed vehicle platoons by integrating uncertainty assessment into distance constraints. Comparative simulation studies with a conventional model predictive control (MPC) approach demonstrated that our GP-MPC strategy ensures more reliable safe distancing and fosters efficient vehicular dynamics, achieving notably higher speeds within the platoon. By incorporating a sparse GP technique in HV modeling and adopting a dynamic GP prediction within the MPC framework, we significantly reduced the computation time of GP-MPC, marking it only 4.6% higher than that of the conventional MPC. This represents a substantial improvement, making the process about 100 times faster than our preliminary work without these approximations. Our findings underscore the effectiveness of learning-based HV modeling in enhancing both safety and operational efficiency in mixed-traffic environments, paving the way for more harmonious AV-HV interactions.

Figures (13)

• Figure 1: A mixed vehicle platoon is composed of $N_a$ connected AVs, denoted as ${A^1, A^2, \cdots, A^{N_a}}$, trailed by a HV $H$. The AVs employ a sequential bidirectional communication topology for data sharing, but no direct communication takes place between the connected AVs and the HV. This configuration is motivated by recent studies highlighting that most accidents in mixed traffic involve HVs rear-ending AVs, leading to the specific platoon arrangement showcased.
• Figure 2: An aerial image of the test track to conduct the field experiments of a human-driven vehicle following an autonomous vehicle.
• Figure 3: The path for the field experiments shown in a GPS photo of the field test track.
• Figure 4: The autonomous vehicle platform, UW Moose, and a following human-driven vehicle on the test track.
• Figure 5: The pre-defined velocity profile for the autonomous vehicle to follow.