# Optimization of multi-AGV task allocation based on an improved PSO algorithm

## TL;DR

This paper improves task allocation for automated guided vehicles in factories using an enhanced particle swarm optimization algorithm, leading to better efficiency and reduced idle distances.

## Contribution

An improved particle swarm optimization algorithm is proposed to enhance task allocation fairness, priority, and AGV utilization in real factory environments.

## Key findings

- The proposed algorithm outperforms MGA and MBTA in reducing no-load distances and improving AGV utilization.
- Simulation experiments show the algorithm achieves shorter total task completion times compared to existing methods.
- The algorithm's performance is stable and applicable in real-world production settings.

## Abstract

Research on task allocation for multiple automated guided vehicles (AGVs) in factory environments is a key topic in intelligent manufacturing. Existing studies often struggle to balance fairness and priority in task allocation, leading to low AGV utilization and high no-load distances. Moreover, the stability and applicability of task allocation algorithms in real-world production environments face significant challenges. To address these issues, a mathematical model is formulated with the objective of minimizing the no-load distances of all AGVs in material delivery tasks. The model is subsequently enhanced by incorporating task allocation balance and priority. To solve the optimization model, an improved particle swarm optimization algorithm is proposed, and extensive simulation experiments are conducted based on a real factory environment. By comparing the optimization results of the proposed algorithm with those of the latest multi-population genetic algorithm (MGA) and the market-based bundle task allocation method (MBTA), it is evident that both the proposed algorithm and MGA achieve higher AGV utilization and shorter total task completion times than MBTA, while also optimizing no-load distances. Although the running time of the proposed algorithm is slightly higher than that of MBTA, it is significantly lower than that of MGA, and its overall performance in reducing no-load distances and enhancing AGV utilization is superior to that of MGA. The proposed method can be applied to guide multiple AGVs in multi-material delivery tasks in real factory environments.