Multicore aware Lyapunov based particle swarm optimization framework for efficient edge task offloading

Effective task offloading in multi-access edge computing (MEC) is challenged by high processing delay, energy consumption, system cost, and system instability. This paper proposes MC-LyPSO, a multi-core-aware Lyapunov-based PSO method that integrates Lyapunov drift-plus-penalty (LDPP), Dynamic voltage and frequency scaling (DVFS)-based adaptive CPU control, and multi-core delay modeling with prediction-driven scheduling and queue-aware cost optimization for robust edge task offloading. We formulate the UAV-assisted MEC task offloading problem as a mixed-integer nonlinear programming (MINLP) model with binary offloading decisions, continuous allocation such as CPU frequency, transmit power, and coupled nonlinear delay–energy constraints. The proposed (MC-LyPSO) framework is evaluated against existing methods, including Hybrid Lyapunov-PSO, Lyapunov-only, PSO-only, and Random offloading strategies in a dynamic MEC environment with multiple UAV-mounted edge servers and mobile devices. Results show that MC-LyPSO achieves 20% lower system cost, 60–65% lower energy consumption and delay, with average energy of 0.03 J, delay of 65 ms, a system cost of 0.12, and stability index of 0.25, outperforming baseline methods in dynamic edge conditions.