Neural network optimization of routes and logistics of construction equipment at large facilities: from planning to autonomous management

Abstract

This study is devoted to the study of innovative approaches to the use of neural network technologies to optimize the routing and logistics of construction equipment at large-scale construction sites. The relevance of the topic is due to the growing shortage of qualified personnel in the construction industry, the increasing complexity of infrastructure projects and the need to improve the efficiency of using construction resources. The paper presents a comprehensive analysis of existing neural network architectures for route optimization tasks, including convolutional neural networks (CNNs), graph neural networks (GNNs), and reinforcement learning systems. The methodology is based on a comparative analysis of the application of various machine learning algorithms on a sample of 15 large construction sites with a total of 347 units of construction equipment. The empirical database includes telemetry, GPS tracking, and production data collected between January and October 2024. The results demonstrate that the integration of neural network optimization reduces equipment downtime by 28.3%, reduces fuel consumption by 17.6%, and increases the overall efficiency of logistics operations by 22.5%. The proposed multimodal optimization system, combining predictive analytics and autonomous management, has shown superiority over traditional methods in various scenarios of construction work. The research makes a significant contribution to the development of the theory and practice of the use of artificial intelligence in the management of construction processes and forms the basis for further improvement of autonomous control systems for construction machinery.