Online Optimization in UAV-Enabled MEC System: Minimizing Long-Term Energy Consumption Under Adapting to Heterogeneous Demands

Unmanned aerial vehicle (UAV) can work as a flying computing platform to supply computation services to users when the terrestrial infrastructure is insufficient or damaged, due to its high mobility, flexibility and controllability. However, there remain many challenges in practical UAV-assisted mobile edge computing (MEC) system. Among them, a unique challenge is how to coordinate communication and computing resources to adapt the diverse heterogeneous demands of users in dynamic network environments. Accordingly, this paper investigates a more practical UAV-enabled MEC network, which considers the task backlog queues and the heterogeneous demands of users. With joint optimization transmit power, bandwidth ratio and UAV trajectory, we minimize the long-term energy consumption while ensuring the controllable task backlog queues. As the proposed problem is a long-term stochastic optimization problem, we utilize the Lyapunov method to transform it into two deterministic online optimization sub-problems and iteratively solve them. Moreover, we design personalized Lyapunov control factors to meet the tradeoff between energy consumption and queue stability for different users with heterogeneous requirements. In terms of solving sub-problems, for the first sub-problem, we prove its convexity by using the convexity-preserving property of composite perspective function, and then obtain the closed-form optimal solution. For the second sub-problem, we skillfully design a low-complexity trajectory scheduling algorithm by using successive convex approximation (SCA), penalty function, and convex function properties. The simulation results show that the proposed algorithm with a lower complexity effectively reduces the long-term energy consumption of the system while meeting the heterogeneous requirements of users.