Gymnasium Library Interface for Multi Access Edge Computing

This paper addresses the problem of developing reinforcement learning algorithms for capsule placement decision making tasks in a geo-distributed Multi Access Edge Computing environment. A Python library called gym-Iwmecps has been developed to speed up and simplify the process of developing machine learning algorithms with reinforcement learning for decision-making system tasks in MEC networks. At the same time, the Gymnasium Capability API has not been used previously in the tasks of organizing and testing MEC platforms. We also tested basic algorithms Q-network and DQN in conjunction with gym-lwmecps and proved the performance of the developed Python library.