Model Predictive Control Using Stochastic Motion Prediction of Surrounding Vehicles in Uncontrolled Intersections

Abstract

This paper presents an autonomous driving algorithm for uncontrolled intersections based on Model Predictive Control (MPC) and Interacting Multiple Model (IMM) filters, proposing an innovative approach to addressing human driver uncertainty in mixed traffic scenarios. The autonomous vehicle receives real-time information about the target vehicle through Vehicle-to-Infrastructure (V2I) communication and leverages the IMM filter to accurately estimate the probabilities of various driving models and the future behavior of the human driver based on drivable paths. Specifically, by incorporating the collision probability calculated with the estimated driving probability and associated uncertainty into the MPC constraints, the system determines a safe distance, enabling the autonomous vehicle to maintain a safe distance from the target vehicle. This approach offers a significant technological advancement in handling atypical human drivers in Unstructured environments. The algorithm was validated through computer simulations in the MATLAB/Simulink environment, with case studies and Monte Carlo simulation results successfully demonstrating an optimal passing strategy that considers ride comfort, traffic efficiency, and safety. This study provides practical guidance on how autonomous driving systems should respond in complex traffic environments where autonomous vehicles and human drivers coexist and is expected to make a significant contribution to the commercialization and safety enhancement of autonomous driving technology.

Keywords

• Uncontrolled intersections
• vehicle to infrastructure (V2I)
• mixed traffic environment
• characteristics of human driver
• collision probability
• interacting multiple model filter (IMM Filter)