In the paper we present the module which uses triangle mesh of the robot to detect collisions between parts of the robot. To this end, the triangle to triangle intersection test is applied. To speed up the computation the bounding box test is carried out at the beginning. We show the properties and performance of the collision detection module. Then, we propose the method which uses Gaussian mixture to determine self- collision model. The method is significantly faster than the method which uses triangle meshes but less precise. The collision detection methods can be applied during motion planning as well as during execution of the planned motion to detect infeasible configurations of the robot.
Tomasz Augustyn, Dominik Belter, Fast Self-Collision Detection Method for Walking Robots, Recent Advances in Automation, Robotics and Measuring Techniques, vol. 351, R. Szewczyk, C. Zieliński, M. Kaliczyńska (Eds.), pp. 549-559, 2016 (pdf)