Tuer, Steven (University of Waterloo), Panahandeh, Pouya (University of Waterloo), Ahmad, Alghooneh (University of Waterloo), Sun, Chen (University of Waterloo), Zhang, Ruihe (University of Waterloo), Ning, Minghao (University of Waterloo), Khajepour, Amir (University of Waterloo)

Real-Time MPC for WATonoBus Path Tracking

Scheduled for presentation during the Invited Session "Emerging Data-driven Technologies and Machine Intellection for Smart Traffic Applications" (FrAT6), Friday, September 27, 2024, 12:10−12:30, Salon 14

2024 IEEE 27th International Conference on Intelligent Transportation Systems (ITSC), September 24- 27, 2024, Edmonton, Canada

This information is tentative and subject to change. Compiled on October 8, 2024

Abstract

In this paper, a real-time Model Predictive Controller (MPC) tailored for precise path tracking in congested urban traffic environments is developed. Leveraging a front and rear steering dynamic bicycle model, along with a linear tire model, the controller aims to reduce system dynamic complexity to ensure computational feasibility for real-time operation. The proposed MPC formulation features a quadratic cost function, solved using a standard quadratic programming solver, to optimize steering inputs over a prediction horizon. Experimental validation on the WATonoBus platform demonstrates the controller's effectiveness in guiding the vehicle along desired trajectories with precision. Precise tracking is especially important in urban driving scenarios such as merging, and pullover, where safety and avoidance of pedestrians and curbs is paramount. Additionally, a framework for augmenting the vehicle prediction model using data-driven techniques is explored. This framework provides the ability to account for unmodelled system dynamics while maintaining the real-time capability of the controller. Overall, this work showcases the practical applicability of an MPC controller for real-world autonomous driving scenarios.