| In this dissertation, we present a solution to the motion planning and control problem that combines three different, and at times conflicting, control inputs: deliberative plans, local reactive behaviors, and unpredictable user-initiated goals. We systematically bring together these different inputs and develop a shared motion control framework. This framework can be used for semi-autonomous robotic systems. Our experimental platform is a robotic wheelchair where the user and robot share control of the motion.; We begin by describing our test bed, University of Pennsylvania's S MARTCHAIR. By adding motors, encoders, sensors, and onboard processing to a regular wheelchair, we have designed an intelligent robotic system. We also developed a theory for shared motion control, which led to the development of a hierarchical control architecture for semi-autonomous systems. In this hierarchy, user input ranges from continuous interaction (via joystick) to a single interaction (user selects the destination and the chair autonomously drives to the target).; Controllers at each level have been derived and implemented. For example, we can drive the SMARTCHAIR to a user-specified location on an omnidirectional camera image. After developing basic navigation abilities, we compose together these modular behaviors to perform more complex tasks, which require multiple controllers. At an even higher level of the hierarchy, the system chooses appropriate behaviors to drive to a user-specified location. In our completed semi-autonomous system, users are able to interact with the autonomous system at any given time during the execution of a task.; We are also concerned with the framework's impact on the advancement of assistive technology. Therefore, we have conducted extensive usability tests which suggest that our semi-autonomous system performs as well as, or better than the human user when functioning in unpredictable, dynamic environments. A detailed analysis of the study can be found at the end of the thesis. |
