On Maximizing Lateral Clearance of an Autonomous Vehicle in Urban Environments

TL;DR

This paper develops an MPC-based control strategy to maximize lateral clearance of autonomous vehicles in urban environments, enhancing safety by increasing distance from road agents through modified control objectives.

Contribution

It introduces a novel MPC formulation that explicitly encodes clearance maximization, linking safety metrics with control design, and demonstrates real-world applicability on public roads.

Findings

MPC controller effectively increases lateral clearance in urban driving scenarios.

Modified MPC can incorporate safety metrics into control objectives.

Experimental validation shows feasibility and improved safety margins.

Abstract

We consider the problem of maximizing distance to road agents for a self-driving car. To this extent, we employ a Model Predictive Control (MPC) approach for the steering tracking control of an Autonomous Vehicle (AV). Specifically, we first present a traditional MPC controller, which is then extended to encode the clearance maximization goal by manipulating its cost function and constraints. We provide insights on the additional information needed to achieve such goal, and how this modifies the structure of the original controller. Furthermore, a connection between commonly used safety metrics and clearance to road users is established. We implement the MPC controller using two off-the-shelf numerical solvers, assessing its computational feasibility. Finally, we show experimental results of the proposed approach on public roads in Boston and in Singapore.