Intelligent IoT-Based Network Clustering and Camera Distribution Algorithm Using Reinforcement Learning

The advent of a wide variety of affordable communication devices and cameras has enabled IoT systems to provide effective solutions for a wide range of civil and military applications. One of the potential applications is a surveillance system in which several cameras collaborate to monitor a specific area. However, existing surveillance systems are often based on traditional camera distribution and come with additional communication costs and redundancy in the detection range. Thus, we propose a smart and efficient camera distribution system based on machine learning using two Reinforcement Learning (RL) methods: Q-Learning and neural networks. Our proposed approach initially uses a geometric distributed network clustering algorithm that optimizes camera placement based on the camera Field of View (FoV). Then, to improve the camera distribution system, we integrate it with an RL technique, the role of which is to dynamically adjust the previous/existing setup to maximize target coverage while minimizing the number of cameras. The reinforcement agent modifies system parameters—such as the overlap distance between adjacent cameras, the camera FoV, and the number of deployed cameras—based on changing traffic distribution and conditions in the surveilled area. Simulation results confirm that the proposed camera distribution algorithm outperforms the existing methods when comparing the required number of cameras, network coverage percentage, and traffic coverage.