A path planning framework for autonomous vehicles

In this paper we present a framework for path planning of autonomous vehicles in static environments which allows for rapid prototyping and evaluation. We achieve this by decoupling search based path planning into search on the one hand and expansion strategies on the other hand. Thus we are able to combine arbitrary sampling methods to discretize the search space with arbitrary heuristic search algorithms. Sampling methods and searches can be connected at design-time without the need to recompile the code. Other search algorithms and sampling methods that are programmed against our interfaces can be easily added and combined with existing search algorithms and sampling methods. Furthermore, the framework is capable of creating more complex planners, by adding further sampling/search combinations to planning pipelines. We show the strength of our approach by applying our framework to one particular path planning problem and evaluating three different planning pipelines.