A flexible hierarchical control method for optimal collision avoidance

Modern active safety systems on road vehicles are capable of sophisticated motion control, e.g. for emergency braking, collision avoidance etc. - assisting or potentially overriding the driver to make speed and/or path corrections. The availability of multiple actuators - especially individual wheel braking, active front steering - enables an agile response from the vehicle, even compared to that of the most skilled human driver. For collision avoidance, a typical control approach is to: (a) define a reference geometric path that avoids collision; (b) apply low level control to perform path following. However there are a number of limitations in this approach, addressed in the current paper. First, it is typically unknown whether the reference path is feasible or over-conservative. Secondly, the control scheme is not well suited to avoiding a moving object, e.g. another vehicle. Further, any incorrect choice of reference path may degrade performance, fast adaptation to friction change is not easy to implement and the associated low-level control allocation may be computationally intensive. In this paper we make use of a particle model for initial path planning and guidance, coupled with a simplified optimal controller, used for control integration and low-level actuation. The particle trajectory is only used as a starting point for control integration; the trajectory is not required to be followed. Instead, motion is maximized in a preferred direction away from possible collision, so the particle trajectory is used for prioritization rather than strict guidance. The aim of the present paper is to show the general feasibility of a simple control algorithm based on a linear Hamiltonian function.