From: 2025-06-05 17:47 to 20:47
Place: Seminar Room M 3170-73 in the M-building, LTH
Contact: karl-erik [dot] arzen [at] control [dot] lth [dot] se

Date & Time: June 9th, 14:00-15:00
Location: Seminar Room M 3170-73 in the M-building, LTH
Author: Fabian Feltenmark
Title: Applying and proving suitable planning principles on RRT* for autonomous trucks
Supervisor: Ofa Ismail (Scania) and Björn Olofsson
Examiner: Karl-Erik Årzén

Abstract: This thesis investigates how high-level planning principles can be deterministically and algorithmically incorporated into a sampling-based motion planner by modifying the structure of the RRT* algorithm. In particular, the requirement to always end a planned route by a reversing maneuver can be directly imposed through the expansion and connection logic of bidirectional search-trees, rather than relying on heuristic biases or cost shaping. A modified bidirectional RRT* algorithm is proposed, in which two search trees are grown from the start and goal configurations of a planning task, where the goal-rooted sub-tree is directionally constrained using path validation and sample rejection. These sub-trees are connected via a pre-defined node called the waypoint node, enabling the enforcement of a fixed motion structure. Experimental results in simulation and on the real autonomous trucks verify the capability of the presented planner to reliably provide feasible paths under the stated constraints. The proposed framework provides a deterministic and modular approach to the integration of planning principles into sampling-based algorithms for structured environments.