Search

Supply chains are complex, stochastic systems. Nowadays, logistics managers face two main problems: increasingly diverse and variable customer demand that is difficult to predict. Classical forecasting methods implemented in many business units have limitations with the fluctuating demand and the complexity of fully connected supply chains. Machine Learning methods have been proposed to improve prediction. In this paper, a Long Short-Term Memory (LSTM) is proposed for demand forecasting in a physical internet supply chain network. A hybrid genetic algorithm and scatter search are proposed to automate tuning of the LSTM hyperparameters. To assess the performance of the proposed method, a real-case study on agricultural products in a supply chain in Thailand was considered. Accuracy and coefficient of determination were the key performance indicators used to compare the performance of the proposed method with other supervised learnings: ARIMAX, Support Vector Regression, and Multiple Linear Regression. The results prove the better forecasting efficiency of the LSTM method with continuous fluctuating demand, whereas the others offer greater performance with less varied demand. The performance of hybrid metaheuristics is higher than with trial-and-error. Finally, the results of forecasting model are effective in transportation and holding costs in the distribution process of the Physical Internet. [ABSTRACT FROM AUTHOR]

Job-shop scheduling problems are among most studied problems in last years because of their importance for industries and manufacturing processes. They are classified as NP-hard problems in the strong sense. In order to tackle these problems several models and methods have been used. In this paper, we propose a hybrid metaheuristic composed of a genetic algorithm and a tabu search algorithm to solve the stochastic job-shop scheduling problem. Our contribution is based on a study of the perturbations that affect the processing times of the jobs. These perturbations, due to machine failures, occur according to a Poisson process; the results of our approach are validated on a set of instances originating from the OR-Library (Beasley, J. Oper. Res. Soc.41 (1990) 1069–1072). On the basis of these instances, the hybrid metaheuristic is used to solve the stochastic job-shop scheduling problem with the objective of minimizing the makespan as first objective and the number of critical operations as second objective during the robustness analysis. Indeed, the results show that a high value of the number of critical operations is linked to high variations of the makespan of the perturbed schedules, or in other words to a weak robustness of the relating GA's best schedule. Consequently, critical operations are not only good targets for optimizing a schedule, but also a clue of its goodness when considering stochastic and robustness aspects: the less critical operations it contains, the better it is. [ABSTRACT FROM AUTHOR]

This study aims to address the challenge of developing accurate and efficient parking occupancy forecasting models at the city level for autonomous vehicles. Although deep learning techniques have been successfully employed to develop such models for individual parking lots, it is a resource-intensive process that requires significant amounts of time and data for each parking lot. To overcome this challenge, we propose a novel two-step clustering technique that groups parking lots based on their spatiotemporal patterns. By identifying the relevant spatial and temporal characteristics of each parking lot (parking profile) and grouping them accordingly, our approach allows for the development of accurate occupancy forecasting models for a set of parking lots, thereby reducing computational costs and improving model transferability. Our models were built and evaluated using real-time parking data. The obtained correlation rates of 86% for the spatial dimension, 96% for the temporal one, and 92% for both demonstrate the effectiveness of the proposed strategy in reducing model deployment costs while improving model applicability and transfer learning across parking lots. [ABSTRACT FROM AUTHOR]

In the era of industry 5.0, digital twins (DTs) play an increasingly pivotal role in contemporary society. Despite the literature's lack of a consistent definition, DTs have been applied to numerous areas as virtual replicas of physical objects, machines, or systems, particularly in manufacturing, production, and operations. One of the major advantages of digital twins is their ability to supervise the system's evolution and run simulations, making them connected and capable of supporting decision-making. Additionally, they are highly compatible with artificial intelligence (AI) as they can be mapped to all data types and intelligence associated with the physical system. Given their potential benefits, it is surprising that the utilization of DTs for warehouse management has been relatively neglected over the years, despite its importance in ensuring supply chain and production uptime. Effective warehouse management is crucial for ensuring supply chain and production continuity in both manufacturing and retail operations. It also involves uncertain material handling operations, making it challenging to control the activity. This paper aims to evaluate the synergies between AI and digital twins as state-of-the-art technologies and examines warehouse digital twins' (WDT) use cases to assess the maturity of AI applications within WDT, including techniques, objectives, and challenges. We also identify inconsistencies and research gaps, which pave the way for future development and innovation. Ultimately, this research work's findings can contribute to improving warehouse management, supply chain optimization, and operational efficiency in various industries. [ABSTRACT FROM AUTHOR]