On the modeling, design, and control of compliant robotic manipulators

In the future of the aging society it becomes increasingly difficult to provide personal care to people who are in need of help due to physical impairments caused by various diseases or disorders. The number of caregivers is limited and the cost of individual personal care will increase. Although personal care is desired, the downside is that people in need of care become dependent on caregivers for accomplishing their activities of daily living. This dependence has a big influence on one’s quality of life and self esteem. A solution to this situation is the deployment of assistive robotic aids like multifunctional robotic arms mounted to a table or wheelchair. These systems can aid in various tasks like opening doors and picking up a drinking glass, as well as personal hygiene tasks. Because many of today’s wheelchair mounted robotic arms are based in some aspects on existing designs for industrial applications, they are not safe to be used in human care services. Therefore, there is a need for safe robotic arms that are applicable for this kind of care. This thesis describes various ways of accomplishing safety in robotic arms, and argues that using variable stiffness actuation, which has not been applied before in assistive robotic arms, is a way to accomplish that. These actuators provide the possibility to not only control the position of the actuator output, but also to influence the stiffness that can be perceived at this output. Various concepts of these actuators, using different principles, are investigated in Part I, and Part II treats the modeling of robotic arms that are driven by these actuators. Modularity was a specific goal, because of the large diversity of connections between the actuators and the arm. The result is a model based on graph theory, which consists of vertices and edges, in which the repositioning of an actuator means changing the connections of one or more edges. This thesis also presents a structured method of analyzing an arm's stiffness, and treats the control of the actuator stiffnesses to optimally approximate a desired end effector stiffness.