Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Task Delay and Energy Consumption Minimization for Low-altitude MEC via Evolutionary Multi-objective Deep Reinforcement Learning

Geng Sun, Weilong Ma, Jiahui Li, Zemin Sun, Jiacheng Wang, Dusit Niyato, Shiwen Mao

TL;DR

The paper tackles the dual objective of minimizing total task delay $f_1$ and UAV energy consumption $f_2$ in a UAV-assisted MEC system tailored for the low-altitude economy. It casts the problem as a multi-objective Markov decision process and introduces an evolutionary multi-objective DRL approach (EMODRL) with a multi-objective target distribution learning (TDL) component, plus a simulated-annealing–based scheduling (SA) to reduce action space. The proposed EMO-TDL-SA framework yields non-dominated Pareto policies and demonstrates superior convergence and trade-off performance against strong baselines in simulations. This approach enables dynamic, Pareto-aware control of UAV trajectory and offloading decisions, offering practical gains for LAE deployments with varying requirements and conditions.

Abstract

The low-altitude economy (LAE), driven by unmanned aerial vehicles (UAVs) and other aircraft, has revolutionized fields such as transportation, agriculture, and environmental monitoring. In the upcoming six-generation (6G) era, UAV-assisted mobile edge computing (MEC) is particularly crucial in challenging environments such as mountainous or disaster-stricken areas. The computation task offloading problem is one of the key issues in UAV-assisted MEC, primarily addressing the trade-off between minimizing the task delay and the energy consumption of the UAV. In this paper, we consider a UAV-assisted MEC system where the UAV carries the edge servers to facilitate task offloading for ground devices (GDs), and formulate a calculation delay and energy consumption multi-objective optimization problem (CDECMOP) to simultaneously improve the performance and reduce the cost of the system. Then, by modeling the formulated problem as a multi-objective Markov decision process (MOMDP), we propose a multi-objective deep reinforcement learning (DRL) algorithm within an evolutionary framework to dynamically adjust the weights and obtain non-dominated policies. Moreover, to ensure stable convergence and improve performance, we incorporate a target distribution learning (TDL) algorithm. Simulation results demonstrate that the proposed algorithm can better balance multiple optimization objectives and obtain superior non-dominated solutions compared to other methods.

Task Delay and Energy Consumption Minimization for Low-altitude MEC via Evolutionary Multi-objective Deep Reinforcement Learning

TL;DR

The paper tackles the dual objective of minimizing total task delay and UAV energy consumption in a UAV-assisted MEC system tailored for the low-altitude economy. It casts the problem as a multi-objective Markov decision process and introduces an evolutionary multi-objective DRL approach (EMODRL) with a multi-objective target distribution learning (TDL) component, plus a simulated-annealing–based scheduling (SA) to reduce action space. The proposed EMO-TDL-SA framework yields non-dominated Pareto policies and demonstrates superior convergence and trade-off performance against strong baselines in simulations. This approach enables dynamic, Pareto-aware control of UAV trajectory and offloading decisions, offering practical gains for LAE deployments with varying requirements and conditions.

Abstract

The low-altitude economy (LAE), driven by unmanned aerial vehicles (UAVs) and other aircraft, has revolutionized fields such as transportation, agriculture, and environmental monitoring. In the upcoming six-generation (6G) era, UAV-assisted mobile edge computing (MEC) is particularly crucial in challenging environments such as mountainous or disaster-stricken areas. The computation task offloading problem is one of the key issues in UAV-assisted MEC, primarily addressing the trade-off between minimizing the task delay and the energy consumption of the UAV. In this paper, we consider a UAV-assisted MEC system where the UAV carries the edge servers to facilitate task offloading for ground devices (GDs), and formulate a calculation delay and energy consumption multi-objective optimization problem (CDECMOP) to simultaneously improve the performance and reduce the cost of the system. Then, by modeling the formulated problem as a multi-objective Markov decision process (MOMDP), we propose a multi-objective deep reinforcement learning (DRL) algorithm within an evolutionary framework to dynamically adjust the weights and obtain non-dominated policies. Moreover, to ensure stable convergence and improve performance, we incorporate a target distribution learning (TDL) algorithm. Simulation results demonstrate that the proposed algorithm can better balance multiple optimization objectives and obtain superior non-dominated solutions compared to other methods.
Paper Structure (25 sections, 29 equations, 8 figures, 2 tables, 5 algorithms)

This paper contains 25 sections, 29 equations, 8 figures, 2 tables, 5 algorithms.

Table of Contents

  1. Introduction
  2. Related Work
  3. Models and Preliminaries
  4. System Overview
  5. Task Model
  6. Communication Model
  7. Computational Model
  8. UAV Movement Model
  9. Problem Formulation
  10. EMODRL-based Approach
  11. MOMDP Simplification and Formulation
  12. Action Simplification
  13. MOMDP Formulation
  14. The Proposed EMO-TDL-SA
  15. Evolutionary Multi-objective Optimization Framework Overview
  16. ...and 10 more sections

Figures (8)

  • Figure 1: A schematic diagram of the system for task computation assisted by a UAV equipped with an edge server.
  • Figure 2: EMO-TDL-SA framework
  • Figure 3: TDL Framework
  • Figure 4: Convergence curves of various DRL algorithms. (a) Comparative convergence results of EMO-PPO-SA, EMO-TDL-SA, ETD3, and ESAC algorithms (highlighting the general convergence trends and performance differences). (b) Comparative convergence results of EMO-PPO-SA, EMO-TDL-SA, ETD3, and ESAC algorithms (highlighting the initial convergence in early training episodes).
  • Figure 5: Convergence trend for EMO-TDL with FCFS, SJF, SA, and PS scheduling strategies.
  • ...and 3 more figures