Motion control for dynamic mobile robots

In this thesis, we present research results on sensor-based motion planning and nonlinear control for mobile robotic systems. In sensor-based motion planning, we use ideas from game theory to deal with the uncertainties accompanying real sensors and moving obstacles. We show that this idea can be successfully applied to both open loop and closed loop motion planning and control algorithms. With the emphasis on the use of a vision sensor, we extend the concept of sensor-based motion planning to motion planning in the image plane, which can help us to bypass the calibration errors associated with vision-based control, and achieve faster response speeds. Meanwhile, we address the effect of dynamics in vision-based motion control, or visual servoing, and expanded our ability to control a dynamic robotic system, such as the blimp robot. We also study the control aspects of underactuated mechanical systems, such as blimps or satellites. We show that for these types of systems, we can formulate the problem on a principal fiber bundle. We illustrate how to use mechanical connections in doing Lagrangian reduction for such systems, and compare the two descriptions of a dynamic system. When coupled with the case of under-actuation, the control of this type of mechanical system becomes even more challenging. We also introduce an averaging theorem to the control on fiber bundles, and show that we can generate desired motions along the directions which are not available directly. In the blimp system, this theorem means we can parallel park a flying vehicle. To illustrate the theoretical research, we apply our algorithms to 2D and 3D robots, including an exciting one to control an unmanned blimp in either simulations or experiments