Neural network modelling of a planar 5-dof biped

A new way of dynamic modelling a biped system using neural networks (NN) is presented. Such NN models have been shown to be good at describing the dynamic behaviour of systems in many applications sectors. The NN biped model is developed for the double support (DS) phase of a planar 5-link biped. Standard identification techniques are used to generate suitable data from an ODE model, which is used for the training of the NN model. The configuration of the biped, methods used for obtaining the training data and the topology of the NN are discussed. The results show that it is possible to model the DS phase using such heuristic techniques that can yield performance superior to those achieved using traditional modelling methods.