Le Van Hoa^{1[0000-0001-6621-8617]}, Pham Trung Duc^{2[0000-0003-2303-9920]},
Pham Trong Nguyen^{3,4[0009-0007-2461-5957]}, Vo Viet Minh Nhat^{5[0000-0002-0686-5529]}

Mobile Edge Computing (MEC) is a promising framework to support latency-sensitive and computationally intensive applications for the Internet of Things (IoT) in edge computing systems. The joint task offloading and resource allo-cation (TORA) policy is important in multi-user, multi-server MEC systems to minimize execution delay, energy consumption, and resource usage. This pa-per investigates the TORA problem with a comparative study of metaheuristic optimization algorithms. Specifically, the TORA problem is modeled as a con-strained mixed-integer nonlinear programming model that minimises the joint consideration of the execution delay, the energy consumption, the connectivity availability and the resource utilisation on an edge server while meeting laten-cy and energy constraints. In this paper, five popular representative metaheu-ristic algorithms: genetic algorithm (GA), particle swarm optimization (PSO), artificial bee colony (ABC), cuckoo search (CS) and firefly algorithm (FA), are integrated into a unified optimization framework with a shared representa-tion, discretisation scheme, constraint-handling technique and fitness evalua-tion method. The simulations reveal that GA achieves the lowest cost (1.72×10⁻²) and the lowest failure rate (6%), with an offloading rate of 65%, while PSO achieves the lowest execution latency, energy consumption and op-timisation time due to fast convergence, with a cost of 1.79×10⁻² and a failure rate of 18%. ABC and CS provide competitive results in between the best and the worst of algorithms whereas FA shows the highest cost (2.26×10⁻²) and failure rate (27%) among the five evaluated algorithms