Neural Networks Applied to Gait Control of Physically Based Simulated Robots

This paper describes our experiments with autonomous robots, in which we use neural networks to generate and control stable gaits of simulated legged robots into a physically based simulation environment. In our approach, the gait is accomplished using an Elman network trained using a gradient descend method, more specifically, the RPROP algorithm, a improvement of the traditional Back-propagation. The model validation was performed by several experiments realized with a simulated four legged robot using the ODE physical simulation engine. The results showed that it is possible to generate stable gaits using neural networks in an efficient manner.