Joint Optimization of Latency and Energy Consumption for Computing Task Offloading Based on Cooperative Multi-UAV and HAP Networks
The rapid expansion of computation-intensive applications renders current mobile edge computing (MEC) frameworks inadequate for delivering high-quality computing services in environments with sparse network infrastructure. Air-based stations, represented by low-altitude unmanned aerial vehicles (UAVs) and high-altitude platforms (HAPs), are considered a promising solution to this problem due to their flexible deployment, unconstrained by geographical conditions, and relatively low cost. However, UAV-based communication systems are highly sensitive to energy consumption, and HAPs face challenges in establishing stable connections with power-constrained devices while meeting their requirements for computation. To solve these limitations, we design a multi-UAV and HAP collaborative offloading framework innovatively, which takes both system energy cost and task processing delay into consideration, with their weighted consumption defined as the optimization objective. Since it is a mixed integer nonlinear programming (MINLP) problem, which is complicated to address by mathematical approaches, we reformulate it as a Markov decision process (MDP) and use a joint approach based on double deep Q network (DDQN) - proximal policy optimization (PPO) to assign task offloading methods and ratios, respectively. According to simulation results, the suggested approach ensures the timeliness of task processing while efficiently reducing the weighted consumption under varying number of users, task arrival densities, and task complexities.