Reinforcement learning-based framework for optimizing task allocation and service migration in edge-cloud systems

Addressing issues such as unreasonable resource allocation, limited computing power of edge devices, data privacy leakage, and difficulties in meeting low-latency and low-power consumption requirements in edge cloud computing models, a dynamic task allocation and service migration model for edge cloud systems is proposed. This model constructs a three-tier architecture consisting of resource allocation, data management, and intelligent collaboration. It employs a Markov decision process optimization algorithm to solve resource allocation challenges and establishes an experimental environment involving multiple edge devices and cloud servers. Experiments demonstrate that the model dynamically adjusts task priorities through three-dimensional evaluation, significantly reducing task completion time and enhancing resource utilization. Additionally, it utilizes a decision tree to optimize migration decisions, minimizing ineffective migrations, effectively compensating for the deficiencies of traditional scheduling models.