Optimizing Total Service Time for Incoming Parallel Requests Using a Hybrid Deep Q-Learning Algorithm in Mobile Edge Computing

Cellular networks’ development drives traffic growth, enhances device capabilities, and introduces complex applications. This trend demands advanced processing, met by Mobile Edge Computing (MEC). MEC, adding edge computing, addresses limitations and boosts network performance, notably reducing latency for critical applications. In this work, the proposed hybrid Deep Q-Learning (DQL) algorithm which adaptively assigns computing and network resources is used. This algorithm is examined for different hyper parameters by using cumulative reward. We also conducted extensive simulations to observe total service time by using different reward functions. In addition, in case of parallel incoming application requests from the cellular network, the time taken respond to the requests has been examined and presented comparatively.