Dynamic optimization method of target coverage based on reinforcement learning for wireless visual sensor network

Wireless visual sensor network (WVSN)-based video surveillance systems have been widely used in a variety of monitoring fields. The problem of maximizing target coverage with the minimum visual sensors is receiving a great deal of attention in WVSN. Because video sensors’ sensing range is limited by their field of view (FoV), the different rotated angles cover the various target nodes. We propose a distributed adaptive learning algorithm for target coverage of a visual sensor network based on the direction coverage characteristics of the visual sensor nodes. First, we built a dynamic coverage optimization model of considering energy efficiency based on the different coverage directions of visual sensor network nodes, and then proposed a distributed target coverage optimization algorithm based on reinforcement learning (DTCRL). Experiment results show that the proposed DTCRL algorithm has a significantly higher average target coverage ratio than typical distributed algorithms. Furthermore, the proposed method uses fewer active sensor nodes and has a higher target coverage ratio than other algorithms. It also indicates that the proposed algorithm not only improves the utility of the target coverage but also reduces energy consumption. It provides technical support for the visual sensor network’s target coverage application.