Real-time system architecture for assistive and rehabilitation exoskeletons

Assistive and rehabilitation exoskeletons are a promising field of technology helping people with mobility disorders to accomplish daily activity tasks or to recover physical strength. Exoskeletons are used also in other fields that require human power augmentation including industry and military applications. Considering human body complexity and the nature of the required assistance, it has been acknowledged that user tailored, adaptable and adjustable devices are needed. This technology is still in an early stage due to the insufficient proof of exoskeleton effectiveness, with a small number of devices available, limited performed research and remaining unsolved challenges. Exoskeletons are naturally establishing a quasi-symbiotic relationship between robots and the humans wearing them. The close contact and natural dynamic interactions demand systems with novel hardware and software designs, acting as reactive machines, dealing with uncertainties rather than as pure servo pre-programmed mechanisms. One of the main goals of this dissertation is to investigate system requirements and propose an architecture solution that will allow building of high performance, user tailored exoskeletons, considering "assistance as needed" and "assistance where needed" paradigms. In this thesis hardware and software solutions are presented for all identified architectural layers. A special attention was attributed to the most challenging, the low level hardware and software layer, which physically contributes to the exoskeleton development. A key component developed for this reason is a versatile Low Level Controller unit that interfaces with all sensors and actuators required to control one exoskeleton joint in real-time. By combining the Low Level Controller unit with compliant actuation, redundant sensors, torque control strategies and real-time communication protocols, the Smart Variable Stiffness Actuator concept is introduced. It is meant to be used in modular systems to solve for issues such as adaptability, scalability and primarily safety. Towards safe and robust human-robot interaction sensing and control a smart multi axis force sensor has been developed. This sensor is meant to interact with the 6 Degrees of Freedom human pelvis used in balance control. Proposed architecture and the smart hardware modules have been developed and evaluated on ALTACRO, CORBYS and MIRAD case study exoskeletons. Provided solutions can be certainly adopted by other exoskeleton applications that share similar hardware and software challenges in physical human robot interaction.