Design and implementation of Compliant Control Strategies for a Series Elastic Actuated upper limb exoskeleton for neurorehabilitation

Stroke is a cerebrovascular disease that occurs when there is a blockage or haemorrhage that reduces the blood supply to the brain; consequently, the brain does not receive sufficient oxygen and nutrients, leading to the death of brain cells. Stroke is the second leading cause of death in the world and in Italy. Stroke's survivors usually are left with impairments and motor problems, which make daily living activities difficult. Rehabilitation plays a fundamental role in the patient's motor recovery: the need for repetitive and precise assistive movements makes robotics an ideal application in rehabilitation. Technology-supported training is emerging as a solution to support physiotherapists by providing high-intensity, repetitive and activity-specific treatments, to improve the recovery process and facilitate restoration of upper limb function. Upper limb exoskeletons are mechatronic and anthropomorphic systems designed to interact with the patient. Series Elastic Actuators (SEAs) are actuator systems suitable for interacting with the environment: the compliant element is used to measure and control the torque interaction between motor and load. One of the main features of the exoskeleton is transparency, i.e., the possibility of creating a torque control that is quite accurate to create a safe environment for the patient. In this regard, SEAs are a good solution to implement a complaint control. The purpose of this thesis is to design and implement different compliant control strategies for a multi-joint system that make human-machine interaction user-friendly by modifying the virtual dynamic of the system. To work on the multi degrees of freedom (DOF) system it was necessary to characterize and validate the operation of the Series Elastic Actuators. Thanks to the aid of a real time machine, Speedgoat, different types of SEAs were tested in real time to validate frequency behaviour and study the main characteristics of the joint. After validating the SEA mechatronics, this thesis has been focused on compliant control design and implementation. To obtain such control, first it was implemented a single DOF position control based on a PID controller properly tuned to satisfy different input profiles (steps, ramps, and sine waves at different frequencies). Secondly, it has been worked on compliant control strategies that allow the implementation of partially assistive modalities of the upper limb exoskeleton. Since rehabilitation robots interact with the human body, it is necessary to consider the manipulator and the patient as a single coupled mechanical system. In order to reach the desired, appropriate and safe dynamic interaction between man and robot, the following compliant control techniques were investigated: impedance control (force loop) and its dual, admittance control (position loop). The two techniques were implemented on a multi-joint system with two DOF, the admittance was developed in the joint space, while the impedance was developed in the task space. The results show how, thanks to the compliance and correct tuning of the impedance/admittance filter parameters, it is possible to interact with the end-effector by modifying the virtual dynamics of the system.