Strategies, algorithms, and analysis for autonomous mobile sensor deployment

Mobile sensor deployment is one of the important technological problems in modern science and engineering. The primary objectives include locating and maintaining optimal positions for a set of sensors and finding suitable paths to deploy them to accomplish their missions, such as rescue and surveillance. The environment can be complex, such as in fluids, in cluttered regions, or in dynamic environments with changing obstacles and landscapes. The problem can be even more challenging if the environmental information is partially known so the sensors must adjust their motions and positions when new information becomes available. The mobile sensors may have limitations, such as short communication range, low power supply and weak computation capability. Considering those factors together, the mobile sensor deployment problem becomes a large system engineering problem that calls for new mathematical models, efficient algorithms and analysis strategies. The main objective of this project is developing a set of algorithms and their mathematical foundations to accomplish different tasks, such as optimal sensor placements and motion planning, while handling various uncertainties encountered in reality. More specifically, we proposed to investigate the following problems: - Path-planning strategies that overcome unfavorable starting positions, -Decentralized techniques for re-deployment and re-distribution when individual sensors experience operational failure, -Probability based path planning in uncertain environments. The proposed study is based on the PI's previous results on optimal sensor placements to achieve complete surveillance in a cluttered region. The new developments include 1) efficient and robust algorithms to compute solutions of the problems, 2) the mathematical analysis for the models and algorithms, and 3) numerical simulations of various scenarios. The investigations are conducted based on recent advancements in the optimal transport theory, numerical simulations of stochastic differential equations (SDEs), and artificial neural networks.