The goal of the thesis is to develop and implement a detection of motion intention (DoI) algorithm to assist walk for the Autonomyo exoskeleton. The used parameters are the net CoP position and velocity and the vertical ground reaction forces (VGRF). The CoP and VGRFs are obtained from the values of force measured by the load cells allocated in each foot sole. To evaluate the capacity of the instrumented soles several experiments are performed. These are necessary as the parameters for the DoI algorithm must be validated. The first tests are done by comparing the values measured by a dynamometer with the cells’ ones. Unfortunately, the results are not good because of the incorrectness of the applied methodology. Therefore, another experiment must be approached. Thus, CoP values calculated by the cells are compared with the ones obtained by a force plate. In this case, results were good enough to accept the reliability of the instrumented sole. Therefore, the CoP could be validated as well. The calculation of the net CoP is performed since it is a parameter of the algorithm. This requires to have all the cells under the same reference. Regarding that the cells are positioned individually per foot, a reference in the exoskeleton center is set. Thus, the cells location can be expressed under the same reference and hence calculate the net CoP. Some gait events are defined to then correlate them with the candidate parameters for the DoI algorithm. These are hip flexion, toe off, knee extension and heel strike, for each foot. Moreover, an initial gait event to describe the start state is contemplated. Found correlations are heel strike to net CoP velocity in x direction and hip flexion, toe off and knee extension to net CoP velocity direction (expressed as an angle). The start state is correlated to net CoP velocity in y direction and to percentage of VRGF in each foot. Finally, the algorithm is developed as a finite state machine (FMS) and tested. The states correspond to the gait events previously described. The transitions are designed regarding the precedent found correlations. As for its evaluation, the algorithm is tried with data recorded during a walk with the exoskeleton. The success rate is between 76% and 84%. Even though these values are not good enough for a DoI algorithm, they reflect that the seen correlations exist. Therefore, improvements in the robustness of the algorithm or finding better correlations must be carried out.
