Adaptive Dynamics Movement Primitives, a control framework for wearable robotics

In the next few years, the field of wearable robotics is expected to spread in professional and personal environments. However, different open challenges need to be addressed to achieve a quasi-perfect human-robot symbiosis, implying a close mechanical and cognitive interaction between the robot and its human operator. Lightweight design and versatile control are compelling topics to be tackled for ensuring the user’s comfort and a successful interpretation of motion intentions, increasing the acceptability of the device and, therefore, boosting its usage in real case scenarios. The design of underactuated wearable robots is a promising design approach inspired by the human inter-joint coordination to accomplish complex tasks. Indeed, the underactuated design has been successful for hand and arm prostheses, but little research has been devoted for lower limb devices. However, the underactuated design can lead to a reduction of the weight of the device at expense of added complexity in both the transmission mechanisms and the control. Therefore, this thesis investigates the feasibility of controlling two different underactuated robots, namely (i) a multi-joint underactuated low-back exoskeleton for worker assistance and (ii) a transfemoral prosthesis. Although the achieved results showed that the control of an underactuated wearable robot is viable, further research is needed for a better comprehension of the added value of the underactuated design and the consequent functional benefits for a user. To date, exoskeleton controllers have excelled in rhythmic and quasi-rhythmic tasks, whereas control methods for assisting discrete movements remain limited by their task-specificity. Inspired by neurophysiological dynamic movement primitives (DMPs), in this thesis a novel controller has been formulated to facilitate discrete movements using a single adaptive DMP (aDMP), for wearable robotic assistance of discrete movements. The proposed aDMP-based control method can be potentially used in all domains of wearable robotics, enabling the possibility to provide functional and synchronous assistance to a great variety of movements. In particular, this thesis shows its use in wearable robotics for both human augmentation (specifically, assisting user in load lifting tasks), and motion replacement (as tool to perform locomotion mode recognition to control a lower limb prosthesis). The results of the experimental activities advances the current state of the art, thus supporting the aDMP as an enabling technology that can pave the way to further research topics in wearable robotics.