A Hybrid ACO-DQN Approach for Wireless Power Transfer-Based Task Offloading in Mobile Edge Computing

This paper proposes a hybrid Ant Colony Optimization– Deep Q-Network (ACO-DQN) approach for task offloading in wireless power transfer enabled mobile edge computing (WPT-MEC) systems. Internet of Things (IoT) devices harvest energy from edge servers and dynamically select local execution or edge offloading to minimize the weighted cost of task delay and energy consumption under energy and latency constraints. Existing heuristic and reinforcement learning methods suffer from cold-start inefficiency and limited adaptability in dynamic MEC environments. To address these challenges, ACO is used to generate high-quality demonstration data for behavioral cloning pre-training, followed by DQN fine-tuning for online adaptation. Simulation results demonstrate that the proposed method achieves more stable training and higher average rewards than baseline algorithms.