Grant: Discovery Grants Program - Individual

Control of multi-agent systems has attracted considerable attention in engineering and science research literature in the past two decades. Various coordination objectives for multi-agent systems, such as consensus, rendezvous, formation, flocking, and coverage, have been investigated extensively, and efficient control strategies have been developed accordingly. Recent advancements in emerging technologies, such as a platoon of autonomous vehicles in a smart city, a group of unmanned aerial vehicles (UAVs) deployed for surveillance and rescue missions, and a network of mobile sensors employed to monitor an environment, have signified the importance of feasible agent coordination techniques to address various problems arising in real-world applications. The proposed research aims to tackle existing problems in controlling and coordinating a network of agents to perform a prescribed mission. Given a group of stationary and mobile agents capable of communicating in a limited range, this work aims to develop distributed control algorithms to cooperatively achieve multiple objectives subject to trade-offs, where improvement in one objective may oppositely affect another. In particular, the coverage problem in sensor networks, which is the main focus of this proposal, involves multiple conflicting objectives, such as network lifetime. On the one hand, it is desired to move the mobile sensors to the proper places in the field to maximize sensing coverage. On the other hand, mobility, sensing and communication are energy-consuming, which is of practical importance because sensors (mobile or stationary) in real-world applications are often battery-powered. Furthermore, given the typically large number of mobile agents and their geographical distribution, each agent needs to be driven by a control law that uses only the available information of that agent and its neighbors. Therefore, the agents must operate in a distributed fashion, which means they must maintain a certain level of connectivity for an efficient flow of information between them. These introduce potentially conflicting performance criteria, and hence, the problem is formulated as a constrained multi-objective optimization. As the first step in solving the above optimization problem, all features such as obstacle avoidance, nonidentical sensing capabilities, presence of both stationary and mobile sensors, prioritized sensing field, probabilistic sensing model, reliable coverage and measurement error will be divided into different categories based on their mathematical description. An effective analytical method for each category will then be used in combination with an appropriate learning-based approach to address the problem. The proposed framework enables one to model the most general setting for a coverage maximization problem, making it applicable across numerous real-world scenarios and practical realizations. The results will be validated by Monte Carlo simulations and experiments.