Optimal charging scheduling for Indoor Autonomous Vehicles in manufacturing operations

Abstract

Indoor Autonomous Vehicles (IAVs) have become instrumental in modern logistics, particularly in dynamic operational environments. Their consistent availability is crucial for both timely task execution and energy efficiency in smart factories. Therefore, an inefficient charging schedule can waste energy, compromising production and warehousing efficiency. In addition, formulating an effective charging schedule is challenging due to the nonlinearity of its design and the uncertainties inherent in smart factory settings. In contrast to previous studies that either simplify the problem using linearization-based methods or approach it with computationally demanding algorithms, this paper efficiently addresses the nonlinear challenge, ensuring a closer alignment with real-world conditions. Hence, a simulation environment is first designed to replicate a smart factory, including validated models of IAVs, Charging Stations (CSs), and a variety of unpredictable static and dynamic obstacles. A comprehensive dataset is then provided from this simulation setup. Utilizing this dataset, a model tailored to ascertain the travel time of IAVs, accounting for inherent uncertainties, is trained using Deep Neural Networks (DNNs). Subsequently, a Mixed-Integer Nonlinear Programming (MINLP) problem is formulated to design the optimization task. Finally, integration of the DNNs’ model with the Branch-and-Bound (BnB) approach, forming the BnBD method, streamlines the determination of the optimal charging schedule. Experimental results highlight significant improvements in charging scheduling, establishing this approach as a viable and promising solution for manufacturing operations.