Skip to content Skip to navigation

Autonomous Vehicle Control at the Limits of Handling

June, 2012

Abstract

Many road accidents are caused by the inability of drivers to control a vehicle at its friction limits, yet racecar drivers routinely operate a vehicle at the limits of handling without losing control. If autonomous vehicles or driver assistance systems had capabilities similar to those of racecar drivers, many fatal accidents could be avoided. To advance this goal, an autonomous racing controller was designed and tested to understand how to track a path at the friction limits.

The controller structure was inspired by how racecar drivers break down their task into (i) finding a desired path, and (ii) tracking this desired path at the limits. Separating the problem in this way instead of integrating both path planning and path tracking into one problem results in an intuitive structure that is easy to analyze. Assuming that a desired path is known, the racing controller in this dissertation focuses on tracking this path at the friction limits. The controller is separated into steering and longitudinal modules, each module consisting of feedforward and feedback controller submodules. From the desired path, the longitudinal feedforward submodule uses the path geometry and friction information derived from a “g-g” diagram to execute trail-braking and throttle-on-exit driving techniques. These techniques maximize tire forces during cornering by using a combination of steering and brake/throttle inputs. To calculate the steering input, the feedforward steering submodule employs a nonlinear bicycle model. These feedforward submodules can adjust their commands in real-time to respond to any changes in the environment, such as changes in friction due to rain or changes in the desired path to avoid an obstacle.

To add path tracking ability and stability to the system, a fixed-gain full-state steering feedback submodule was combined with a longitudinal feedback submodule that regulates vehicle speed and minimizes tire slip through a slip circle feedback controller. For consistency in the steering controller, the same reference point at the center of percussion (COP) was used for both feedforward and feedback steering submodules. The COP was chosen because it simplifies the feedforward design process by eliminating the nonlinear and changing rear axle force from the lateral dynamics equation. Using the set of steering gains derived from lanekeeping steering with yaw damping feedback, the system was proven to be Lyapunov stable even when the rear tires are highly saturated.

The simulation and experimental results on various surfaces on oval tracks demonstrate that the submodules work collectively to robustly track a desired path at the friction limits. The experimental results highlight the challenges of trail-braking during a corner-entry phase, where a correct corner entry-speed and accurate model of longitudinal weight transfer are required. Thus, longitudinal weight transfer was incorporated into the feedforward longitudinal submodule to minimize oversteer caused by reduction in the rear normal load. In addition to performing well on oval tracks, the racing controller also showed its ability to operate in a challenging environment by driving 12.4 miles up Pikes Peak autonomously, where the path consists of both dirt and paved surfaces with significant bank and grade. The complex path at Pikes Peak also demonstrated the controller’s ability to plan the vehicle speed several corners in advance.

The racing controller’s ability to drive a vehicle at the friction limits can be applied to drive an autonomous vehicle while ensuring stability and tracking ability even in extreme conditions, such as driving on icy road. Alternatively, the submodules in the racing controller can be adapted to create driver assistance systems that work in conjunction with the driver, assisting the driver during emergency maneuvers.