Dynamic Multi-Channel Access in Heterogeneous Industrial Wireless Sensor Networks: A MADRL Approach

Industrial Wireless Sensor Network (IWSN) is one of the key technologies for future industrial automation and intelligent manufacturing. With the increase in the number of wireless network devices, the coexistence of wireless networks becomes a challenge. Dynamic multi-channel access in IWSN is critical for optimizing spectrum utilization and ensuring real-time data freshness. This paper proposes a Model-free Actor-Critic based Multi-Agent Deep Reinforcement Learning (MADRL) approach to address the challenges posed by heterogeneous network environments. The method aims to improve spectrum efficiency and packet freshness by dynamically selecting channels for transmission, considering the Age of Information (AoI) metric. Key contributions include a novel multi-agent framework tailored for heterogeneous networks, harnessing MADRL to enhance spectrum utilization and reduce transmission delays. Simulation results demonstrate significant improvements in throughput and AoI management compared to conventional methods.