Global path planning for competitive robotic cars

In this paper, we consider the optimal motion planning problem for an autonomous race car. A competitive autonomous car must acquire environmental and opponent information to compute, in real time, the minimum time collision free path and the low level control to track the chosen path. To cope with those requirements, we first solve the problem for a car running in isolation considering the optimal sequence of manoeuvres to approach bends and straight stretches of track. We then propose a discrete abstraction to derive a problem of graph optimisation that has a very efficient, albeit suboptimal, solution. In this context, an overtake manoeuvre against a slower car will be obtained excluding from the path the arcs that could potentially generate a collision. Finally, a control algorithm is used to ensure that the car always remains close to the planned path.