Path tracking control using model predictive control with on GPU implementation for autonomous driving

Abstract

Autonomous driving is one of the technologies that will become the next milestone in mobility advancement. Before reaching public roads, autonomous driving is expected in areas where driving is considered to be hard and unsafe, like arid and desert terrains. We are trying to achieve fully automated driverless vehicle in transportation sector. Demand of high performance computation using limited space and minimum energy is a pressing need for the mass application of autonomous driving systems. This paper presents a control system for autonomous driving based on model predictive control (MPC) in which a graphics processing unit (GPU) was used to enhance the computation speed for a better real-time performance. The motion controls for the lateral direction is formulated as a model predictive control problem. The lateral MPC was partly executed using GPU for speed enhancement. The implemented MPC package and its adaptation to GPU will make autonomous driving feasible on relatively smaller and cheaper on-board computers, it will also be beneficial to the developers and researchers in various fields other than control engineering field.