UC Santa Barbara

There has been considerable interest in the area of communication-aware robotics in recent years, where the sensing, communication and motion objectives of robotic systems are jointly optimized. One particular open problem in this area is that of exploiting the mobility of unmanned vehicles in order to improve or satisfy communication objectives in realistic communication environments. Progress in this field could not only affect robust networked operation of unmanned vehicles but also would improve communication systems of the future (e.g. 5G), thus contributing to both areas of robotics and communications. This mobility-enabled connectivity and communication is the main area of interest in this dissertation.

This dissertation is focused on path planning and communication strategies for robotic systems seeking to satisfy certain communication objectives in realistic communication environments experiencing path loss, shadowing and multipath fading. We consider realistic communication environments by leveraging and incorporating a probabilistic channel prediction framework that allows the robots to predict the channel quality at unvisited locations. This thesis then contributes to the area of mobility and connectivity through three main topics 1) energy-optimal distributed beamforming, 2) finding the statistics of the distance traveled until connectivity, and 3) path planning for connectivity. First, in energy-optimal distributed beamforming, we utilize the motion of a group of initially unconnected mobile robots to enable new forms of connectivity. More specifically, we co-optimize their locations and transmission powers to cooperatively enable connectivity through distributed beamforming. We further bring a foundational theoretical understanding to robotic distributed beamforming. Next, in finding the statistics of the distance traveled until connectivity, we analytically characterize the probability density function of the distance traveled by an initially unconnected robot until it gets connected to a remote node as it moves along a given path. We utilize tools from the stochastic differential equation literature to develop this characterization. Finally, in path planning for connectivity, we actively plan the path of a mobile robot such that it finds a connected spot with a minimum expected traveled distance (i.e., energy). The scenario considered in this part is in fact a more general one, and tackles the problem of path planning on a graph to minimize the expected cost incurred until the successful completion of a task. This framework has applications beyond path planning for connectivity, in areas such as celestial body imaging, human-robot collaboration, and search scenarios. We bring a foundational understanding to this problem. We show how this problem is inherently hard to solve (NP-complete) and also propose a path planner, based on a game-theoretic framework, that provides an asymptotic optimality guarantee.

Overall, this thesis proposes novel strategies for utilizing the mobility of unmanned vehicles and enabling connectivity while considering the underlying energy constraints. We also provide a rigorous theoretical analysis of the aforementioned problems using a wide range of tools from communications theory, game theory, optimal control and time series literature. Moreover, through extensive realistic numerical studies using real channel parameters/data, we show the efficiency and performance of our proposed approaches.