Optimizations of a Multi-Agent System for a Real-World Warehouse Problem

Abstract

Abstract In recent years, many warehouses applied mobile robots to move products from one location to another. We focus on a traditional warehouse where agents are humans, and they are engaged with tasks to navigate to the next destination one after the other. The possible destinations are determined at the beginning of the daily shift. Our real-world warehouse client asked us to minimize the total wage cost, and to minimize the irritation of the workers because of conflicts in their tasks. We define a heuristic for the optimizations for splitting the orders into warehouse carts, defining the sequence of the products within the carts, and the assignment of the carts to workers. We extend Multi-Agent Path Finding (MAPF) solution techniques. Furthermore, we have implemented our proposal in a simulation software, and we have run several experiments. According to the experiments, the make-span and the wage cost cannot be reduced with the heuristic optimization, however the heuristic optimization considerably reduces the irritation of the workers. We conclude our work with a guideline for the warehouse.