The robot control strategies are remarkably insufficient stabilization classical systems (conventional PID control), or nonlinear robust controls where no significant changes in the control gains based on the dynamic of interaction. In this research, the design and implementation of a Wavenet control (Multiresolution PID control based on wavelet neural networks of second generation as a identification scheme) and first time is evaluated in a device with high technological performance (Cyberforce) designed for positioning an exoskeleton CyberGrasp with the human operator in the loop. The performance is remarkable, to verify the energy flow in the interaction, experimental evidence can be observed that there is stability in the control loop with the human operator also convergence position in cartesian space, another important factor is that in spite of uncertainty generated by the human operator retains the controller path tracking. Validation from the point of view of the user and defining the workload in the development of the task is evaluated with the NASA TLX protocol. Experiments with different users were developed, which corroborated the high performance of the algorithm proposed in this contribution.