Bilkent University
Department of Computer Engineering
M.S.THESIS PRESENTATION
MO-SeGMan: A Rearrangement Planning Framework for Multi-Objective Sequential and Guided Manipulation in Constrained Environments
Cankut Bora Tuncer
Master Student
(Supervisor: Asst. Prof. Özgür S. Öğüz)
Computer Engineering Department
Bilkent University
Abstract: MO-SeGMan presents a multi-objective sequential and guided manipulation planning framework for rearrangement tasks in highly constrained and cluttered environments. The framework addresses the dual challenges of planning efficiency and solution quality by generating object placement sequences that jointly minimize replanning effort and robot travel distance. A lazy evaluation strategy is employed to preserve critical dependency structures, ensuring feasibility across complex manipulation scenarios. To handle non-monotone and highly cluttered settings, a Selective Guided Forward Search (SGFS) algorithm is introduced that relocates only critical obstacles to feasible positions, thereby reducing unnecessary computational overhead. In addition, an adaptive subgoal selection and refinement strategy enables the decomposition of complex manipulation tasks into feasible intermediate steps, improving planning robustness. Extensive evaluations on 21 benchmark rearrangement tasks demonstrate that MO-SeGMan consistently generates feasible motion plans and outperforms state-of-the-art baselines in both solution time and solution quality. These results highlight the robustness, efficiency, and scalability of the proposed framework for complex rearrangement planning problems.
DATE: June 30, Tuesday @ 10:00
Place: EA 409