Return to search

Finding an Optimal Trajectory for Autonomous Parking Under Uncertain Conditions

Path planning that considers accurate vehicle dynamics and obstacle avoidance is an important problem in the area of autonomous driving. This paper describes a method of implementing trajectory planning for autonomous parking in conditions where the starting point and the position of fixed obstacles are uncertain. The narrow spaces and complicated manoeuvres required for parking demands a lot from the trajectory planning algorithm. It needs to have the ability to accurately model vehicle dynamics and find an efficient way around obstacles. Having obstacles in the way of the parking vehicle makes this a nonconvex problem the goal can usually not be reached by travelling in a straight line and finding a perfect trajectory around them is generally not computationally tractable. This paper reviews a two tiered approach to solving this problem. First a rough path is found using a modified Rapidly-exploring Random Tree (RRT) algorithm called Forward-Backward RRT, which runs two treebuilding processes in parallel and constructs a feasible path from where they intersect. Using optimisation this is then improved into a trajectory that is at least a local optimum. These methods will be demonstrated to produce efficient and feasible trajectories that respects the dynamic constraints of the vehicle and avoids collisions.

Identiferoai:union.ndltd.org:UPSALLA1/oai:DiVA.org:kth-254217
Date January 2019
CreatorsGreinsmark, Vidar, Hjertberg, Tommy
PublisherKTH, Skolan för elektroteknik och datavetenskap (EECS)
Source SetsDiVA Archive at Upsalla University
LanguageEnglish
Detected LanguageEnglish
TypeStudent thesis, info:eu-repo/semantics/bachelorThesis, text
Formatapplication/pdf
Rightsinfo:eu-repo/semantics/openAccess
RelationTRITA-EECS-EX ; 2019:118

Page generated in 0.0019 seconds