Research on Multi-Objective Task Scheduling Optimization and Reinforcement Learning Decision-Making Methods for Edge-Cloud Collaborative Scalable Information Systems

INTRODUCTION: Edge-cloud schedulers must coordinate latency, energy, load balance, and deadline compliance under changing demand while keeping task-arrival and throughput units physically consistent. OBJECTIVE: This study evaluates MORL-ECSO under an auditable, paired-seed simulation protocol and compares it with tuned heuristic, metaheuristic, value-based, actor-critic, and entropy-regularized baselines. METHODS: A custom Python discrete-event simulator processes individual tasks in 0.1-s event windows. Offered load is 500-2000 tasks/s, with 4-20 edge nodes and four cloud nodes. DQN, A2C, discrete SAC, and MORL-ECSO receive the same 12,000 training transitions; GA and PSO use a population of 40, 30 iterations, and a common objective. Each reported test point uses 30 independent workload seeds after a 10-s warm-up and a 60-s measurement window. Means, standard deviations, 95% confidence intervals, paired Wilcoxon tests, Holm correction, and rank-biserial effects are reported. RESULTS: At 2000 tasks/s, MORL-ECSO produced 103.16 ms mean latency, 88.80 kJ energy over 60 s, 94.18% on-time success, and 1947.56 tasks/s throughput. Relative to validation-tuned GA, latency was 18.53% lower, and success was 0.89 percentage points higher.  CONCLUSION: Compared to the optimal baseline, MORL-ECSO improves latency and deadline success rate under high loads without significantly altering energy consumption or throughput. In low-capacity scenarios with four nodes, the genetic algorithm remains superior.