Enhancing Agricultural IoT Networks: Multi-Robot Wireless Charging Task Allocation through Reinforcement Learning

• João Pedro Gomes Carvalho UFPI
• Ranulfo Bezerra Tough Cyberphysical AI Research Center
• Ricardo De Andrade Lira Rabêlo UFPI
• João Pedro Soares Do Monte UFPI

Abstract

As climate change continues to endanger global food security, the adoption of innovative technologies like the Internet of Things (IoT) and robotics is becoming increasingly vital for sustainable agriculture. This evolution, often called Smart Agriculture or the Internet of Robotic Things (IoRT), utilizes IoT technology to optimize resource management, increase crop yields, and reduce environmental impact. One of the employed strategies is the use of a mobile robot capable of recharging IoT devices with low battery-level. However, as the network expands, a single robot may not be enough to sustain the workload. This paper proposes an approach to address these challenges by using multiple mobile robots equipped with wireless charging capabilities. By focusing on the efficient sharing of workload among several robots this approach ensures more effective charging of IoT devices, allowing for a higher number of nodes in a network. By employing a Reinforcement Learning (RL) strategy, the paper details the development of an optimized operational system for these robots, with the aim of extending network lifetime while minimizing individual robot workload. Extensive simulations conducted in a Unity3D-modeled agricultural environment demonstrate that effective task allocation among multiple robots, alongside metrics that account for potential device neglect, is critical for optimizing system performance. Proper task distribution ensures equitable workload allocation, while the metrics help assess and improve the sustained operation of agricultural IoT networks, enhancing both efficiency and reliability.

Keywords: Smart agriculture, System performance, Inductive charging, Stochastic processes, Reinforcement learning, Internet of Things, Resource management, Mobile robots, Reliability, Noise measurement