Personalizing Assistive Robots: A Reconfigurable and Adaptable Shared-autonomy Framework and its Applications

The quality of life of people with motor impairments, often caused by spinal cord injuries, is severely reduced. Assistive robotic devices offer an opportunity to improve the quality of life of these people and help them regain a certain level of independence. A completely autonomous solution is often not desired since people have a need to keep their remaining autonomy or recover some of their lost autonomy, while feeling in control as much as possible. Therefore, robotic assistive solutions that share the autonomy between human and computer have been proposed. These solutions assist by granting the users only partial control over the robot. However, these solutions often provide limited possibilities for personalization and customization, ignoring individual mobility difficulties and preferences. This trend limits people's access to suitable assistive devices that comply with their needs and preferences, since personalization of such devices requires high technical expertise, development time, and therefore high associated costs. Research shows that the lack of personalization also causes that users of these technologies often abandon them, either because the device never fulfilled the user's requirements, or because the requirements changed over time. The lack of user-centered and personalizable robotic devices leaves a clear gap between the features of the available assistive devices and the users' particular needs and preferences. In order to fill this gap, as a first contribution, a constraint-based reactive assistive framework for the control of assistive robots is proposed, which facilitates the personalization of assistive robotic devices such as table-mounted or wheelchair-mounted robot arms. This enables the robot to comply with the particular needs of users with different levels of mobility. For this purpose, the framework encapsulates reactive assistive behaviors into a set of building blocks that can be created and reused in different assistive scenarios. For example, these building blocks enable a continuous reaction to input from human-machine interfaces (HMI) and sensor measurements, enabling human-robot interactivity while providing robustness against changes in the environment.These building blocks are also (re-)combined to create personalized assistive strategies and are modulated to shift the level of the provided assistance according to the situation. As a second contribution, the thesis presents several of reusable and combinable bui