Contributions to autonomous robots working in dynamic environments.
详细信息
文摘
Robots are becoming more and more prevalent today and one of the challenges for robots is navigating around dynamic environments. Dynamic environments are those where a robot has no a priori information and the environmental conditions can change at any time. This can apply to a self-driving robot on the road,a rescue robot at a disaster site or battlefield,or an exploration robot in outer space. For a robot to be truly robust,it must not assume it has access to a human operator for instructions or a GPS system to guide it. In this dissertation,we focus on autonomous robots and assume they are non-holonomic wheeled mobile robots (WMR). We also assume the robots may have limited storage or computation capabilities,such as micro robots or swarm robots,which traditionally are very small. It is under all such considerations that we have design our models and algorithms. We are not aware of many state-of-the-arts focusing on all three constraints- environment,autonomy and limited capability,as each poses its own challenges. In this dissertation,we made three contributions: Our first contribution is some navigation models. We present the three layer motion control and introduce our top layer path planner control that enables a robot to learn how its component functions in dynamic environments and construct optimal paths. We then present our middle layer reactive-plan control model that can avoid collision in real time and discuss an enhancement that speeds up our reactive-plan controller. Our second contribution is an adaptive robotic framework for dynamic environments. Multiple robots working together are generally completing one or more complex tasks that require a robot to switch between navigation,communication,coordination and so forth with other robots. There are some autonomous robotic frameworks in existence but they are not quite suitable for dynamic environments. We discuss the issues of some of these frameworks and then present our framework. Our third contribution is a fast object tracker. For time constraint task or mission,having a fast tracker is critical to a robot,whether to avoid collision with other robots or to track a moving target. We present a tracker that utilizes our recently developed ultra-fast and cheap template matcher.