A matheuristic for AGV scheduling with battery constraints
Nitish Singh, Quang-Vinh Dang, Alp Akçay, Ivo Adan, Tugce Martagan
Abstract
This paper considers the problem of scheduling automated guided vehicles (AGVs) with battery constraints. Each transport request involves a soft time window, and the AGV fleet used to service those requests is heterogeneous with a diverse set of capabilities and travel costs. In contrast to the existing literature, each transport request may require different AGV material handling capabilities (such as lift loads, tow loads, or handle loads with a mounted robot arm), and the AGV batteries can be recharged partially under consideration of a critical battery threshold. The problem is to assign the transport and charging requests to AGVs, sequence the requests, and determine their starting times and the recharging durations of the AGVs with the objective of minimizing a weighted sum of the tardiness costs of transport requests and travel costs of AGVs. A mixed-integer linear programming model is formulated. We also propose a new matheuristic that makes use of an adaptive large neighborhood search algorithm and a linear program to solve industry-size instances. We illustrate the efficacy of our approach with an industry case study using real-world data.