The surge in Internet of Things (IoT) devices and their diverse applications generates massive data volumes requiring substantial processing power, demanding efficient scientific workflow execution across resource-constrained edge devices, fog nodes, and the cloud. Efficiently matching workflow tasks with resources is crucial for minimizing total completion time (makespan), energy consumption, and cost, particularly in delay-sensitive applications. However, achieving this optimal allocation remains a challenge. To tackle this challenge, we introduce a novel multi-objective Improved Particle Swarm Optimization (IPSO) algorithm. We evaluate the performance of the IPSO algorithm against standard PSO. IPSO's effectiveness is assessed through simulations employing the Montage scientific workflow and a varying number of tasks, scaling up to 500. Simulations demonstrate that IPSO outperforms PSO in minimizing completion time (makespan), energy consumption, and total cost. This advantage becomes more pronounced as the number of tasks in the workflow increases, suggesting IPSO's efficacy in handling larger and more complex scientific workflows. For instance, with 500 tasks, IPSO demonstrably reduced makespan, energy consumption, and total cost compared to PSO by 15.11%, 24.02%, and 1.42%, respectively.