Grammatical Evolution for Gait Retargeting

Abstract

Artists and scientists require tools to construct physics-based animal models. However, animating these models requires motion data for realistic movement. Motion data may either be measured from real-life animals-in-motion or generated using an optimisation approach. We propose a solution for retargeting gait data from one animal to another. The retargeted gait cycles are generated using a Grammatical Evolution optimisation approach and the search space is constrained based on dynamic similarity principles.