Gait optimisation for distinct horse models using grammatical evolution

Motion data is required for realistic animation of physics-based animal models. This data is expensive to acquire for a single animal and in herd situations, the large variation in animal shape and consequent motion increases this expense. We propose a method in which data measured from a single horse can be used to animate horses of different age, breed and conformation. The construction and animation of a physics-based horse is described. Details of an application, which automatically generates horse models of a user-specified age, are also presented. We compare two approaches in which Grammatical Evolution is used to optimise a generated model's motion parameters, to produce realistic motion. In one approach, the constant coefficients of a model's spring-damper based muscle system are optimised prior to the gait optimisation. We contrast this method with a parallel optimisation of both spring-damper constants and gait. The sequential approach was found to be the most successful for gait optimisation.