A motion planning is an essential part of autonomous driving. It determines a safe driving course to control a vehicle from starting location to the desired destination. To compute this trajectory, several state of the art algorithms were proposed; however, the computation time and error exponentially grows with the complexity of environment. This limits the deployment of real time motion planning in complex environment. In this paper, we propose a heuristic exploring tree algorithm that locally grows a random exploring tree within potential driving area. With our method, in particular, the number of iterations and accuracy is improved in high complex environment. It can provide not only a real-time motion planning but also a prediction horizon of future motion. The method is an extension of heuristic A* search algorithm applied to continuous domain. The simulation shows improved results in complex path computation scenarios with respect to the state of the art Rapid exploring Random Tree (RRT) motion planning. Local Growing Rapid Tree (LGRT) gives a better performance in quality of the solution showing more than 10 × improvement in the accuracy - iteration ratio.
Locally Growing Rapid Tree (LGRT) motion planning for autonomous driving
2016-11-01
1318153 byte
Aufsatz (Konferenz)
Elektronische Ressource
Englisch
TREE POLICY PLANNING FOR AUTONOMOUS VEHICLE DRIVING SOLUTIONS
Europäisches Patentamt | 2021
|Search-Based Motion Planning for Performance Autonomous Driving
Springer Verlag | 2020
|MULTIMODAL MOTION PLANNING FRAMEWORK FOR AUTONOMOUS DRIVING VEHICLES
Europäisches Patentamt | 2022
|A Review of Motion Planning for Urban Autonomous Driving
IEEE | 2023
|