Design of Routing Protocols for Heterogeneous WSN Based on Multi-Agent Reinforcement Learning

The proliferation of IoT and wireless sensor networks is affecting many different industries. The primary concern in designing IoT-integrated heterogeneous WSNs is energy efficiency since nodes often operate with limited battery units. Software-defined heterogeneous WSNs integrated with the IoT provide an innovative design that can successfully handle these problems. The major concern in WSN is energy consumption by each node and the way of communication between the nodes in the form delivery ration and the reliability need to be more concentrated. This paper presents an enhanced routing method to resolve the two interrelated issues of large-scale IoT-based WSNs. Systems that address clustering and routing independently do not function well together, leading to inefficient solutions when considering energy usage and network longevity. There are several cluster-based routing algorithms for usage with homogeneous WSNs, but very few cater to heterogeneous WSNs regarding energy efficiency. The study's proposed routing protocol, MARL-HWSN, uses multi-agent reinforcement learning to address the heterogeneous WSN issue. The routing protocol design would improve clustering, reduce energy consumption, and provide reliable routing in the network. Modelling the network's architecture with multiple agents and developing a routing protocol with residual power and connectivity quality in mind improves adaptation to changing circumstances and helps extend the network's lifetime. The optimization approach incorporates a global reward function as part of the networked system. Compared to previous approaches, the experimental results showed that MARL-HWSN simulations employing Software-Defined Nets (SDN) enhanced the packet delivery ratio and lowered data delivery time while demonstrating excellent energy efficiency and quick response to unanticipated adjustments to the network.