• upper limb prostheses
• stiffness control
• variable stiffness actuators

One of the main challenges in the design of prosthetic devices and bionic aids is to replicate the human motor behavior which can be integrated within the human sensory-motor architecture. The controllable elasticity embodied inside the human musculoskeletal systems appears to be a key point to generate their natural movements. In fact, this allows human beings to interact safely and robustly with the environment. In addition, they can then perform highly dynamic or cyclic tasks in our daily lives, such as running, jumping, hitting, or throwing.

Since the early eighties, the literature suggests that prostheses should have a user -controllable impedance. Due to the limited actuator and sensor technologies, only a few noticeable investigations on this subject had been made until the late 2000's. Advances in sensors technologies and in the processing of biosignals pave the way for investigations on new control strategies for prosthetic devices. Until now, using a user-controllable impedance prosthesis has not yet been proved to be significantly better. Yet, having several levels of impedance enhances the performances of the prostheses doing various types of tasks. It is with this in mind that I derive, in this thesis, an experiment to validate the potential of a user-controllable impedance in upper limb prostheses. The results suggest that this ability could enhance the versatility of the devices in activities of daily living requiring both precision motions and soft interactions.

Regarding the mechanical design of prosthetic devices, the first approach to get a controllable elasticity is to apply an impedance control. Yet a virtual control of the impedance properties can only simulate an elastic behavior and cannot fully reproduce some characteristics of a physical elastic system, such as the impact absorption or energy storage. In soft robotics, articulated soft robots are currently investigated to embed such inherent elasticity in the mechanisms to fully take advantage of their abilities. However, although this could enhance the performances of the systems and their integration within the human body, there is still a lack of a physical and controllable elasticity in the upper limb prosthetic devices. To address this issue, I propose, in this thesis, to design an upper limb prosthesis with two goals in mind. The first one is to physically implement a natural behavior inside, using an embedded elasticity. The second one is to provide a complete prosthesis with a user-controllable inherent elasticity for future investigations.

Using the articulated soft robotic technologies, I investigate several approaches to integrate an inherent and controllable elasticity for an elbow and wrist joints. I come up with two innovative solutions for the elbow joint. The first one is distributed along the mechanism on both sides of the joint and its torque function matches the shape of the human muscle model, suggesting a natural passive behavior. The second one is located only on one side to better fit the patient morphology in case of a lower level of amputation. In addition, I introduce a new configurable architecture to design two degree-of-freedom compliant joints with a user-controllable impedance to match the compliant behavior of human joints. This serves to design a compact and homogeneous three degree-of-freedom wrist mechanism that matches the passive stiffness of a human wrist. All of these articulations are joined together with an existing soft robotic hand to be the first soft articulated transhumeral prosthesis that has a passive behavior that matches the human one and a user-controllable impedance.