TY - GEN
T1 - Two-Stage Constraint-Aware NSGA-III for Spatio-Temporal Task Allocation in Edge Computing
AU - Zhang, Weiyi
AU - Yu, Xiaodong
AU - Wang, Huijuan
AU - Huang, Wei
N1 - Publisher Copyright:
© 2026 IEEE.
PY - 2026
Y1 - 2026
N2 - Spatio-temporal task allocation in edge computing is challenging due to dynamic task arrivals, heterogeneous edge nodes, and strict deadline constraints. Under highly imbalanced workloads, traditional scheduling methods and standard evolutionary algorithms often fail to simultaneously ensure feasibility and high scheduling efficiency, resulting in excessive deadline violations or suboptimal makespan. This paper proposes a twostage constraint-aware NSGA-III (hereinafter referred to as EA-NSGA-III for brevity) for spatio-temporal task allocation in edge computing. The proposed algorithm explicitly separates constraint handling and performance optimization into two stages. In the first stage, feasibility is emphasized by progressively reducing deadline violations through a constraint-aware search process. In the second stage, the algorithm focuses on improving makespan, response time, and load balance across edge nodes. Experimental results under imbalanced workload conditions demonstrate that the proposed approach consistently outperforms FCFS, NSGA-II, and standard NSGA-III in terms of makespan, deadline satisfaction, and load balance.
AB - Spatio-temporal task allocation in edge computing is challenging due to dynamic task arrivals, heterogeneous edge nodes, and strict deadline constraints. Under highly imbalanced workloads, traditional scheduling methods and standard evolutionary algorithms often fail to simultaneously ensure feasibility and high scheduling efficiency, resulting in excessive deadline violations or suboptimal makespan. This paper proposes a twostage constraint-aware NSGA-III (hereinafter referred to as EA-NSGA-III for brevity) for spatio-temporal task allocation in edge computing. The proposed algorithm explicitly separates constraint handling and performance optimization into two stages. In the first stage, feasibility is emphasized by progressively reducing deadline violations through a constraint-aware search process. In the second stage, the algorithm focuses on improving makespan, response time, and load balance across edge nodes. Experimental results under imbalanced workload conditions demonstrate that the proposed approach consistently outperforms FCFS, NSGA-II, and standard NSGA-III in terms of makespan, deadline satisfaction, and load balance.
KW - edge computing
KW - multiobjective optimization
KW - spatio-temporal task allocation
KW - task scheduling
UR - https://www.scopus.com/pages/publications/105051501405
U2 - 10.1109/ICCCBDA69573.2026.11676433
DO - 10.1109/ICCCBDA69573.2026.11676433
M3 - Conference contribution
AN - SCOPUS:105051501405
T3 - 2026 11th International Conference on Cloud Computing and Big Data Analytics, ICCCBDA 2026
SP - 582
EP - 586
BT - 2026 11th International Conference on Cloud Computing and Big Data Analytics, ICCCBDA 2026
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 11th International Conference on Cloud Computing and Big Data Analytics, ICCCBDA 2026
Y2 - 24 April 2026 through 26 April 2026
ER -