AUTHOR=Zhang Jiankang
TITLE=Optimization design of highway route based on deep learning
JOURNAL=Frontiers in Future Transportation
VOLUME=5
YEAR=2024
URL=https://www.frontiersin.org/journals/future-transportation/articles/10.3389/ffutr.2024.1430509
DOI=10.3389/ffutr.2024.1430509
ISSN=2673-5210
ABSTRACT=Introduction: In recent years, the total mileage and line density of China’s highways have increased year by year. It is estimated that by 2026, the total mileage of national highways will exceed 5.74 million kilometers. An efficient highway network is crucial for a country’s move towards traffic modernization, economic development, and improvement of people's livelihoods. The highway route is the basic structure of the highway network, determining whether the highway can maximize its economic and traffic effects. Therefore, research on highway route design holds significant engineering value. Highway planning is a complex issue involving a wide range of factors. Especially with the increasing awareness of environmental protection, it is necessary to consider natural problems in addition to technical and economic costs.
Methods: This paper first points out the important position of highway route research in highway rules, summarizes the research status at home and abroad, and lists conventional highway planning measures. It then discusses the optimization design based on vehicle running speed and driver comfort, and introduces related theories of deep learning and their applicability to multi-objective optimization problems. Finally, aiming at the problem of highway route planning influenced by many factors, a deep learning strategy based on a multi-objective genetic algorithm is adopted, and its multi-objective optimization model and optimization objective function are presented.
Results: The proposed deep learning strategy based on a multi-objective genetic algorithm is a new attempt to combine genetic algorithms with deep learning in highway route planning to solve its multi-objective comprehensive optimization problem. The results indicate that this strategy can determine the best route scheme by optimizing technology while satisfying external constraints, thereby achieving the optimal solution in terms of technology and economy, and improving the overall efficiency and sustainability of the highway.
Discussion: This study provides a reference for the application of deep learning and other nonlinear multi-objective optimization research, aiding the research on highway route optimization design. By combining multi-objective genetic algorithms with deep learning, it effectively solves various multi-objective nonlinear problems, providing new methods and tools for highway route planning.