Distributed motion planning in robotic sensor networks

Recently there has been tremendous interest in sensor networks for its ubiquitous applications, and in many of these applications, robots have became an integral part of the system, and therein robot mobility and network communication are two deeply coupled components. In this thesis, we investigate some interesting interplays between communication and mobility. The first half of the thesis studies communication-assisted motion planning of robots, where a static sensor network deployed in the environment is used to navigate robots. We revisit some existing researches in wireless communication from the perspective of robot motion planning, and propose an effective and efficient distributed algorithm for robot navigation based on communication backbone. Toward another direction, we see the emerging trend of more sophisticated in more capable sensor networks, where sensors (such as cameras) give a spatial map rather than a single reading. We integrate the classic sampling-based planning techniques, and propose a distributed probabilistic roadmap algorithm for such applications. The proposed method is also able to deal with physical obstacles, and navigates robots through potential narrow passages, as traditional sensor networks with simple sensors are not able to. The second half the thesis discusses communication-constrained motion planning of robotic sensor networks, where a team of mobile robots form a mobile sensor network, and actively maintain connectivity of the system so that robots can always communicate with each other, either directly or via other robots. We propose a novel hierarchical distributed cooperative control scheme based on communication backbone of the network: Backbone-Based Connectivity Control (BBCC). Key advantages of BBCC are that it is completely general in that it can deal with arbitrary system topologies; it is a distributed method using only two-hop neighbor information; and finally, it has low communication cost. Further-more, we look into potential local minimum issues that can arise because multiple objectives are considered. Our empirical observations motivate a classification of local minima based on the underlying cause, and we outline strategies to escape these minima.