Resource interleaving with reinforcement learning for task scheduling in Internet of Things

The rapid growth of the Internet of Things (IoT), coupled with increasing demands for task service quality such as low latency, has positioned task scheduling as a fundamental area of study within IoT systems. Owing to resource competition among tasks reaching the IoT network node, queuing delays often occur across heterogeneous resources. To address this problem, we propose RIRL, a task-scheduling method that combines resource interleaving and reinforcement learning, to enhance resource utilization and improve the efficiency of IoT systems. Using resource interleaving, we dynamically schedule network node tasks and adjust the task distribution order to optimize waiting delays across heterogeneous resources after task forwarding. By aggregating the requirements for computing, storage, and networking resources of different tasks in the IoT system, we formalize the task optimization scheduling problem at the network forwarding node. To implement task scheduling in complex and dynamic IoT environments, we use an RL approach based on the Soft Actor-Critic to derive a real-time task scheduling policy. Simulation experiments confirm the effectiveness and superiority of RIRL in task scheduling, demonstrating that it reduces task response time.