Path Planning in Heterogenous Environments: A Combined Approach
Ajinkya Krishnakumar · Research Repository (Delft University of Technology) · 2019
Autonomous vehicles are the inevitable future of the industry as theoretically they guarantee higher road throughput and a much safer means of transport compared to today’s ground vehicle. This has attracted the industries and universities making it a very important topic of research. The basic function of the autonomous vehicle boils down to transporting its passenger safety from door to door. This requires planning of a path that is obstacle free. Currently, sampling-based methods are widely used for path planning. Although these methods are proving to be successful in open environments, they are inefficient in heterogeneous environments. Planning in urban environments would be successful if this obstacle is tackled. In the current literature, it was found that there are many methods which focus on solving the path planning problem while planning around the obstruction, while there are other methods that focus on converging to optimal solutions. Hence, there is a need for methods that would plan optimal paths in heterogeneous environments. This thesis introduced a combined approach that shares elements of planning paths around obstacles and optimal path planning which are provided by the algorithms Adaptive RRT and Informed RRT* respectively. These methods are combined along with a Dubin’s motion model to simulate basic vehicle’s constraints. The proposed approach was compared with state of the art methods like RRT* to evaluate its performance via simulation. The simulation was carried out in three different scenarios with variable complexity depending upon the available free configuration space. The ability to find a solution and converge it were evaluated, an improvement of 50% was noticed in finding the initial solution and around 25% improvement was seen in convergence. This concluded that such a hybrid approach could be an important contribution to urban path planning.