GPU Accelerated Collision Detection for Robotic Manipulators
Daniel Attila Szabo, Emese Gincsainé Szádeczky-Kardoss · 2022
This paper presents a collision detection method for robotic manipulators. In this work, three methods are compared with each other. In the first one, both the segments of the manipulator and the obstacles are modeled with polyhedrons, and the collision between them is detected by checking the feasibility of the union of inequality systems describing each of them. The second one is solving the collision detection problem between meshes and polyhedrons, while the third one uses the same process, but parallelizing it by using GPU. These are general methods to determine which robot configurations are causing collisions, so they can be used in path-planning or in collision-avoidance algorithms as well. The presented GPU-based parallelized method overperformed the previously examined algorithms and it is fast enough to use in online applications.