This work deals with neural network-based gait-pattern adaptation algorithm for an active orthosis. The proposed device is developed for lower limbs and based on a commercially available orthosis, Figure 1. Active orthoses can be designed for helping physically weak or injured people during rehabilitation procedures [1]. The robotic orthosis Lokomat is being recently used for rehabilitation of patients with stroke or spinal cord injury individuals [2]. Gait-pattern adaptation algorithms are proposed by Riener, et. al [3], considering the human-machine interaction. The algorithms in Riener, et. al [3] were developed for a fixed base robotic system; they can not be applied directly in the proposed orthosis, since no stability of the gait pattern is considered. A trajectory generator for biped robots taking into account the ZMP (Zero Moment Point) criterion is presented in Huang, et al. [4]. This method presents suitable results with smooth and second-order differentiable curves.

You do not currently have access to this content.