Description

Caging analysis addresses the problem of preventing an object from escaping using a set of fixed points, such as a robot gripper [1]. This analysis is often solved via sampling-based motion planning methods, such as Probabilistic Road Maps (PRMs) or Rapidly exploring Random Trees (RRTs). However, these methods suffer from high computational complexity, restricting their use case for high-dimensional problems.

Recent work has proposed Optimal Transport (OT) theory as an efficient solution for motion planning [2]. This gradient-free method optimises motion trajectories in highly nonlinear costs for high-dimensional tasks. The goal of this thesis is to address the problem of caging analysis in object manipulation via Optimal Transport theory.

Deliverables
• Review relevant state-of-the-art literature.
• Formulate the caging analysis problem as an Optimal Transport problem.
• Develop python framework to perform the optimal transport caging analysis in robotic manipulation environments.

Practical information
Pre-requisites: Python (medium/basic).
Tools: Optimal Transport, PyTorch, Python programming language.
Start: Available immediately.

References:
[1] Dong, Yifei, and Florian T. Pokorny. "Quasi-static Soft Fixture Analysis of Rigid and Deformable Objects." arXiv preprint arXiv:2309.01224 (2023).
[2] Le, An T., Georgia Chalvatzaki, Armin Biess, and Jan R. Peters. "Accelerating Motion Planning via Optimal Transport." Advances in Neural Information Processing Systems 36 (2024)

The work is under the scope of the following projects:

• SoftEnable: Towards Soft Fixture-Based Manipulation Primitives Enabling Safe Robotic Manipulation in Hazardous Healthcare and Food Handling Applications (web)