An Extensible Collision Avoidance Model for Realistic Self-Driven Autonomous Agents

We have developed a novel collision avoidance model and shown how the model can be used as a basic building block to generate various behaviors for intelligent agents. The main difference between our model and other path planning models is that our model is concerned with the execution of a planned path in a dynamic environment rather than the planning of the path itself. As a result, our model reflects more closely the decision process of humans in collision avoidance. We argue that collision avoidance is in fact a highly complex social interaction between two or more agents, and that our model has the potential of creating a truly heterogenous population for realistic crowd simulations. We have also conducted some basic experiments on the model to investigate the resultant behaviors. The results show that the proposed model is effective in generating various human-like collision avoidance behaviours.