On-line Trajectory Adaptation for Active Lower Limbs Orthoses based on Neural Networks

This work deals with on-line gait-pattern adaptation algorithms for an active lower limbs orthosis based on neural networks (NN). Stable trajectories are generated during the optimization process, considering a trajectory generator based on the Zero Moment Point criterion and the inverse dynamic model. Additionally, a set of neural networks are used to decrease the time-consuming computation of the model and ZMP optimization. The first neural network approximates the inverse dynamics and the ZMP optimization, while the second one works in the optimization procedure, giving an adapted desired trajectory according to orthosis-patient interaction. This trajectory adaptation is added directly to the trajectory generator, also reproduced by a set of neural networks. With this strategy, it is possible to adapt the trajectory during the walking cycle in an on-line procedure, instead of changing the trajectory parameter after each step as it is performed in previous works. The dynamic model of the actual exoskeleton, with interaction forces included, is used to generate simulation results. The results show the proposed algorithm can be applied to an actual orthosis, since the computational cost is reduced with the use of neural networks.