A Potential Field Framework for Active Vehicle Lanekeeping Assistance
In 2001, over 18000 deaths in the United States were caused by automobiles leaving the lane and colliding with a fixed object in the environment. Despite recent advances in vehicle safety, the responsibility for avoiding collisions with objects in the environment remains solely with the driver. Though humans are quite adept at this task, they are far from infallible. This work presents a potential field framework for active lanekeeping assistance that seeks to prevent accidents from lane departures. This control concept assumes by-wire technology to add control inputs (steering and braking) on top of the driver commands. The potential field concept passively couples the vehicle to the environment, seamlessly adding control inputs when necessary to aid in the lanekeeping task. For a lanekeeping assistance system, safety is of primary importance. The potential field framework provides mathematical safety guarantees for the lanekeeping performance while creating a system that works cooperatively with the driver.
This thesis covers the general control structure, a method of incorporating stability and performance objectives in the potential field framework, and a Lyapunov-based method for bounding the lateral motion of the vehicle subjected to time-varying disturbances. This bounding technique is extremely general and is applicable to a wide variety of dynamic systems. The technique provides an excellent bound on the lateral motion of the vehicle in the presence of road curvature disturbances. The potential field controller is implemented on a 1997 Corvette C5 modified to include steer-bywire. The experimental results validate the theoretical lanekeeping performance of the system. The results also show that the theoretical bounds are a useful tool for the design of a potential field controller and quantitatively guarantee the nominal safety of the lanekeeping system.