Design and Development of Robot Control Systems for Biomedical Applications

This thesis is focussed on the design and development of control strategies for applications in rehabilitation and assistive robotics. The first part of the work is dedicated to rehabilitation robotics. According to the World Health Organization, by 2050, the number of persons over 65 years old will increase by 73 percent in the industrialized countries and by 207 percent worldwide. This segment of population is particularly prone to suffer a cerebrovascular accident or stroke, since the relative incidence of stroke doubles every decade after age 55. Stroke is a growing social problem in the most industrialized countries all over the world and cause of serious long-term disability. An efficient and effective recovery process after stroke in chronic patients should stimulate the neuronal plasticity by promoting the active involvement of the patient throughout repetitive movements with limited assistance. Additionally, an objective assessment of the patient during the treatment period should be guaranteed. In the presented scenario, robot manipulators can be exploited as tools for therapists to increase productivity and quality of the care. In fact, the robot can deliver a quantifiable input and measure the patient's output objectively, monitoring the evolution of the therapy from admission to discharge of the patient. The control systems for robot-aided rehabilitation proposed in this work tries to provide assistance to the patients exploiting human motion generation mechanism. The main purposes to achieve are a tailored therapy on the patient and a 'physiological assistance' according to patient's need. For 'physiological assistance' is intended the robot capability to replicate the therapist interaction with the patient. To this aim, a body of neuroscientific evidences is identified which supports the existence of discrete elementary units, called submovements, underlying continuous human movement and on this neuroscientific assumption a novel control approach for patient assistance is proposed. According to these evidences, a complex movement, such as pointing tasks can be decomposed in a sequence of discrete units of movements, which are usually denoted as submovements. In the proposed approach (named submovement-based control), robot motion is generated by dynamic sequences of elementary motion units (submovements) that, on-line modulated, can achieve accurate motion or else force regulation. The expected benefit of the control approach is to provide robotic assistance as similar as possible to human therapist. A feasibility study on the predetermined static sequence of submovements and the dynamic sequence generator implementation are presented. The dynamic submovement sequence generation is grounded on the adaptive behaviour of four oscillators. Control performance in free space and in interaction with a purely elastic environment can be managed by dynamically changing features of motion units in the current sequence. The theoretical formulation of the control strategy and the application to robot point-to-point motion in free and in constrained space are provided in simulated environment and by means trials on available robot. A comparative analysis with a traditional PD control is also carried out. Additionally and in complementary way in the key area of robot-aided rehabilitation, in order to enhance robot dependability in human-robot interaction, highly nonlinear phenomena able to strongly degrade robot performance such as joints friction are investigated and compensated in an impedance control based on inverse dynamics of the considered robot. The proposed control system (named current-based impedance control) relies further on measured electric currents to close the control loop, in lieu of traditionally used torque/force sensors, in addition to position feedback. It does not require force sensors and solves problems related to the increase of the apparent inertia perceived by a human user (due to force sensor mass), wiring issues and costs of the sensors. The control law has been applied to the CBM-Motus upper-limb rehabilitation machine by modelling robot kinematics and dynamics, whereas static and dynamic friction properties are experimentally estimated. A workspace characterization of the CBM-Motus has been carried out in order to retrieve the experimental relationship between applied force to the end-effector and motor currents. The current-based impedance control performance in free space and in interaction with external environment have been compared with impedance control purely based on position feedback. The second part of thesis is focused on the field of assistive robotics. In particular, in view of an application to prosthetic hands, the main objective of this work is determining the optimal hand configuration that fits object characteristics and ensures a stable grasp. In fact, for manipulation aids, robotic and prosthetic hands the ability to realize smooth movements and to obtain a stable grasp has a primary importance. For this reason, one issue that the present work tries to address is to develop a bio-inspired control approach for determining posture, contact points with the object and trajectory of the fingers of a robotic hand during the grasping action. In order to reduce the complexity of the control algorithm and ensure grasp stability, the optimal hand grasp posture is searched. The proposed algorithm is based on the minimization of an objective function expressed by the sum of the distances of the hand joints from the object surface for diagonal and transverse volar grasping. Algorithm effectiveness has preliminarily been tested by means of simulation trials. Finally, experimental trials on a real arm-hand robotic system (consisting of MIT-Manus planar arm and DLR-HIT\Hand II robotic hand) have been carried out in order to validate the approach and assess algorithm performance.