Efficient Path Matching Motion Generation Algorithm for Multi Agent Environment

Abstract

This paper presents an efficient method to generate a large amount of continuous motion data from motion captured data. Given user defined way point lists for each agent, the algorithm can automatically generate collision free walking paths for them. The walking path data is then analyzed and transformed into a sequence of agent states such as walking or standing state. Based on the randomized depth first algorithm, the agent states are matched with a sequence of corresponding motion clips. The final motion is obtained by blending the motion clips to fit with the speed of the agents' walking paths. From our experiments, our algorithm generates natural looking motions.