Development of a vehicle dynamics controller for obstacle avoidance

TLDR: In this paper, an automatic obstacle avoidance controller that integrates steering and braking action was developed to perform an ISO 3888-2 emergency obstacle avoidance manoeuvre in an industrial environment using a two-track vehicle model.

Abstract: As roads become busier and automotive technology improves, there is considerable potential for driver assistance systems to improve the safety of road users. Longitudinal collision warning and collision avoidance systems are starting to appear on production cars to assist drivers when required to stop in an emergency. Many luxury cars are also equipped with stability augmentation systems that prevent the car from spinning out of control during aggressive lateral manoeuvres. Combining these concepts, there is a natural progression to systems that could assist in aiding or performing lateral collision avoidance manoeuvres. A successful automatic lateral collision avoidance system would require convergent development of many fields of technology, from sensors and instrumentation to aid environmental awareness through to improvements in driver vehicle interfaces so that a degree of control can be smoothly and safely transferred between the driver and vehicle computer. A fundamental requirement of any collision avoidance system is determination of a feasible path that avoids obstacles and a means of causing the vehicle to follow that trajectory. This research focuses on feasible trajectory generation and development of an automatic obstacle avoidance controller that integrates steering and braking action. A controller is developed to cause a specially modified car (a Mercedes `S' class with steer-by-wire and brake-by-wire capability) to perform an ISO 3888-2 emergency obstacle avoidance manoeuvre. A nonlinear two-track vehicle model is developed and used to derive optimal controller parameters using a series of simulations. Feedforward and feedback control is used to track a feasible reference trajectory. The feedforward control loops use inverse models of the vehicle dynamics. The feedback control loops are implemented as linear proportional controllers with a force allocation matrix used to apportion braking effort between redundant actuators. Two trajectory generation routines are developed: a geometric method, for steering a vehicle at its physical limits; and an optimal method, which integrates steering and braking action to make full use of available traction. The optimal trajectory is obtained using a multi-stage convex optimisation procedure. The overall controller performance is validated by simulation using a complex proprietary model of the vehicle that is reported to have been validated and calibrated against experimental data over several years of use in an industrial environment.