Abstract: 

In this paper, we propose Mobility-aware Energy Efficient Scheduling (MEES), a novel scheduling framework for Cyber-Physical Systems (CPS) in the computing continuum. MEES addresses the challenges posed by dynamic mobility patterns and the heterogeneous nature of resources across edge, fog, and cloud layers. MEES includes a task scheduling and resource management mechanism that adapts to network topology changes using mobility patterns of Levy and random walks 5-tier architecture. It employs Dynamic Event base Heterogeneous Scheduling (DEHS), an energy-efficient ensemble scheduling algorithm optimized for heterogeneous architectures to minimize energy consumption while meeting deadlines Directed Acyclic Graph (DAG)-based and scientific workflows and periodic tasks with stipulated deadlines. The framework also supports preemptive scheduling with lazy and eager strategies. Additionally, MEES integrates Reinforcement Learning Resource management (RLRM), an intelligent decision-making capability that adapts to historical data and real-time system dynamics. The Kubernetes-based evaluations show that MEES can reduce energy consumption by an average of 35.9% up to 50.07%, decrease latency by an average of 42.76% up to 62.62%, and reduce makespan by an average of 22.58% up to 74.23% as compared to state-of-the-art algorithms while providing deadline guarantees under various mobility scenarios.