MAC protocol for multi-hop wireless sensor networks utilizing integrated reinforcement learning with joint frame and slot optimization

This paper introduces a distributed reinforcement learning-based MAC protocol designed for high-density educational IoT environments. In smart campuses, the reliability of real-time data from student wearable sensors and classroom environmental monitors is often hampered by hidden-node interference as well as network collisions. This phenomenon disrupts the synchronicity required for effective Human-AI collaboration. Current research utilizes an adaptive slot-swapping strategy to ensure stable, low-latency communication, facilitating seamless human-AI collaboration in data-driven pedagogical settings. The suggested protocol changes both temporal and spatial scheduling at the same time to keep up with changing traffic loads and changing multi-hop network topologies, compared to traditional methods, which often optimize these dimensions separately. Simulation results show that the proposed protocols maintain stable performance across diverse network densities and traffic loads of 0.1 to 1.0 Erlangs. Moreover, collision rates were reduced to 60% compared to Q-learning ALOHA in dense interference scenarios, thereby effectively mitigating the pseudo steady state trap. A 30–50% faster convergence time was demonstrated, ensuring that AI-driven educational feedback loops are established almost instantaneously. Significant enhancement in energy efficiency to 35%, reducing end-to-end delay. These results confirm that joint optimization of frame and slot parameters is a robust strategy for building the reliable, high-throughput communication backbones necessary for the next generation of AI-driven pedagogical tools.