Navigation Control in an Urban Autonomous Ground Vehicle

Southwest Research Institute developed an Autonomous Ground Vehicle (AGV) capable of navigating in urban environments. The paper first gives an overview of hardware and software onboard the vehicle. The systems onboard are classified into perception, intelligence, and command and control modules to mimic a human driver. Perception deals with sensing from the world and translating it into situation awareness. This awareness is then fed into intelligence modules. Intelligence modules take inputs from the user to understand the need to navigate from its current location to another destination and, then, generate a path between them on urban, drivable surfaces using its internal urban database. Situational awareness helps intelligence to update the path in real time by avoiding any static/moving obstacles while following traffic rules. Control modules take the path command from intelligence and actuate the accelerator/brake pedal and steering to physically drive the vehicle from point A to point B. Sliding mode algorithms developed for controlling the steering are described. Performance improvement of the vehicle maneuvering curves is quantified relative to conventional PID algorithms.