Penetration-Free Collision Detection and Response Algorithm Based on Extended Position-Based Dynamics

In the virtual simulation platform, frequent physical interactions between the robot and the environment poses extremely high demands on the handling of collisions and contacts in the simulation. Current mainstream virtual simulation platforms, in pursuit of real-time performance, have sacrificed some accuracy in simulation algorithms, leading to control strategies obtained in training that are difficult to directly transfer to actual physical machines. This virtual-physical discrepancy has significantly impacted the efficiency and effectiveness of mobile robot simulation. Existing work has attempted to use more accurate mathematical modeling and design more realistic solvers, but at the cost of uncontrollable problem scale and extremely high computational complexity. Starting from this perspective, this paper designs a non-penetration simulation algorithm based on extended Position-Based Dynamics (XPBD), which improves the physical processing accuracy during the simulation while ensuring real-time performance, enhances the fidelity of the simulation, and reduces the virtual-physical discrepancy. Simulation experiments have been conducted to validate the reliability of our proposed algorithm.