Application of an artificial neural network to the control of an active external orthosis of the lower limb.

The object of this paper is to present a real-time application of an artificial neural network (ANN). The application for which this network is demonstrated is a motorised orthosis with six degrees-of-freedom for use by a paraplegic; a 'walking machine'. Theoretical networks and training methods need modification to function correctly with a real application. Several complex phenomena that are very difficult to model have to be accommodated; the starting threshold of the activators, non-linearity, noise, and the non-biunivocity between successive system states (position, velocity, actuator controls). The modifications made to the network and the associated training method partially alleviate these difficulties.