Robust motion planning for manipulation becomes significantly harder in constrained workspaces such as shelves and bins. The robot must move the arm through narrow passages while avoiding self-collisions and collisions with the environment, often with an attached object that changes the effective collision geometry. In such settings, different planning backends such as sampling-based, optimization-based, and GPU-accelerated motion generation exhibit distinct trade-offs in success rate, planning latency, and trajectory quality, particularly under clutter and narrow-passage conditions.

The focus of this thesis is the integration and systematic comparison of different motion planners across constrained manipulation scenarios, providing clear guidance on when each planning approach works best.