Publication

Abstract : The Internet of Things (IoT) explosive growth necessitates high-performance networks that can satisfy demanding specifications, including low latency, high reliability, and energy economy. A crucial part of IEEE 802.15.4, the Time-Slotted Channel Hopping (TSCH) protocol is extensively used in IoT networks because of its ability to guarantee dependable communication and its resistance to interference. Nevertheless, traditional TSCH scheduling methods frequently find it difficult to meet the various requirements of applications in dynamic environments that are power-constrained and latency-sensitive. Prioritized Multi-Agent Reinforcement Learning (PMRL) and OPTIMA Prioritized Multi-Agent Reinforcement Learning (OPMRL) are two innovative scheduling models that we suggest as solutions to this problem. Multi-Agent Reinforcement Learning (MRL) and Double Deep Q-Networks (DDQN) are combined with traffic priority awareness in these models to allow for adaptive scheduling decisions. A smart home scenario is used to evaluate the suggested method first, laying the groundwork for later implementation in more intricate IoT environments. Comparing our models to conventional TSCH scheduling techniques, extensive simulations show that they greatly improve Packet Delivery Ratio (PDR), lower latency and power consumption, and limit collisions. The foundation is laid by this study for autonomous, scalable scheduling techniques that can be applied to industrial and smart home settings.