Under-constrained end-effector path following: sampling-based planning algorithms and benchmarking framework

This research focuses on offline, collision-free planning for path-following tasks. Robot tasks with end-effector constraints are common in industrial processes such as robot welding, spraying or gluing. There exists a wide variety of motion planning algorithms to calculate collision-free paths for robot manipulators. Sampling-based motion planning is a large and successful category of such motion planning algorithms. Incorporating path constraints into sampling-based motion planning algorithms is an active area of research. In contrast, grid search can be effective for this type of problems. This thesis builds on this initial grid search success and addresses its limitations by proposing and evaluating a novel hybrid approach in between grid search and sampling-based planning. In addition, the thesis proposes a sampling-based local optimisation algorithm. Finally, the thesis presents a novel benchmarking framework specifically focusing on motion planning problems with many different segments with different types of constraints, where the robot does not manipulate objects in the environment. The novel algorithms in this thesis have been implemented and evaluated in simulation and are evaluated on path planning tasks largely based on robot arc welding. All implementations are freely available online (https://gitlab.kuleuven.be/ACRO/jdm).

Jaar van publicatie:2021

Toegankelijkheid:Open