• dynamic modelling
• FPGA
• friction
• inverse dynamic control
• rehabilitation robotics
• series elastic actuator
• upper limb exoskeletons

Stroke is the principal cause of upper limb impairments. Stroke consequences can be treated or reduced by rehabilitation. Robotic exoskeletons, integrated in rehabilitation treatments, can provide considerable improvements in the rehabilitation clinical outcomes. To ensure a safe interaction, the exoskeleton should guarantee precise torque delivery and monitoring, without hindering the residual patient’s movements.
In order to improve the torque controller performance and robot transparency, this thesis aimed to model the dynamic contributions of a reaction force-sensing series elastic actuator (RFSEA) and of an upper-limb multi-link rigid exoskeleton robot. The friction of the actuators was identified with experimental data, and then mathematical models for friction compensation were implemented. The dynamic model of the RFSEA was directly derived from datasheet data, while the dynamic model of the robot was numerically derived using the Lagrangian formulation. Thus, an inverse dynamic control for dynamic compensation was first implemented and verified on the single RFSEA, and then extended on the multi-joint exoskeleton.
These compensation strategies were tested and evaluated on a bench test setup and with a healthy subject. The results showed that all the compensation strategies bring improvements both in the controller performance and in the exoskeleton output impedance. In particular, the results showed that the friction compensation proved to be considerably effective mainly for movements at lower frequencies, while the dynamic compensation plays a major role for movements at higher frequencies.