Research On Collision Detection For 3D Simulation Of Dual-Arm Mobile Robot

The project"Graphical Programming and tri-d Simulation for dual-arm mobile service robot"is cooperated by Shanghai Jiaotong University and Yaskawa electric corp., and this paper mainly researched collision detection in tri-d simulation based on the project. To realize real-time collision detection on the personal computer not work station, the system required different precision for SmartPal body, arm, gripper and table, cup, wall, shelf based on the given collision detection task, and also use local search to reduce time costing.The paper use different kinds of bounding box for SmartPal and simulation environment to meet the given precision requirement. The paper decompose the non-convex model into sub-model according to the standard hierarchical structure of VRML(Virtual Reality Modeling Language) based on the VRML data format of the virtual robot, AABB and OBB bounding box is built for every sub-model. Because construct OBB is harder than AABB, though OBB has high precision, a parameter k(5mm) is set for threshold value, when the distance is more than k, call AABB, when it's less than k, then call OBB, and Vclip was used to do the collision-computation for the bounding-boxes, which solves the difficulty of dealing with non-convex model for traditional collision-detection method. The paper built bounding box directly for every convex model. The main application of the collision detection system is to apply for SmartPal gripper to take the cup based on the distance information and arm kinematics equation.The experiment result shows that the algorithm can meet the design requirements well, the system can handle the collision detection between SmartPal and simulation environment. It does great contribution for service robot simulation.