Evolved Topology Generalized Multi-layer Perceptron (GMLP) for Anatomical Joint Constraint Modelling

The accurate simulation of anatomical joint models is becoming increasingly important for both medical diagnosis and realistic animation applications. Quaternion algebra has been increasingly applied to model rotations providing a compact representation while avoiding singularities. We propose the use of Artificial Neural Networks to accurately simulate joint constraints, by learning mappings in unit quaternion space. This paper describes the application of Genetic Algorithm approaches to neural network training in order to model corrective piece-wise linear/discontinuous functions required to maintain valid joint configurations. The results show that Artificial Neural Networks are capable of modeling constraints on the rotation of and around a virtual limb.