Active learning for multi-objective optimal road congestion pricing considering negative land use effect. (April 2021)

• Record Type: Journal Article
• Authors: Zhong, Shaopeng; Gong, Yunhai; Zhou, Zhijian; Cheng, Rong; Xiao, Feng
• Is Part Of: Transportation research. Volume 125(2021)
• Publication Date: 2021-04
• Subjects: Road congestion pricing -- Negative land use effect -- Integrated land use and transportation model -- Probabilistic surrogate model -- Multi-objective Bayesian optimization -- Simulation-based optimization; Transportation -- Periodicals; Transportation -- Technological innovations -- Periodicals; 388.011
• DOI: 10.1016/j.trc.2021.103002
• Languages: English
• ISSNs: 0968-090X

Abstract

Highlights: A multi-objective bi-level programming road congestion pricing model is established to consider the negative land use effects. A set of objectives are proposed to measure the effects of road congestion pricing on land use and transportation. A novel active learning algorithm is developed to solve the multi-objective bi-level programming road congestion pricing problem. An empirical analysis is conducted to verify the proposed model and algorithm. The proposed active learning optimization framework can be applied to solve other black box problems with large computations. Abstract: The road congestion pricing policy is implemented to alleviate traffic congestion and improve the efficiency of the transportation system during peak hours. However, the negative land use effect caused by this policy could not be ignored. How to design the optimal congestion toll that can not only ensure its positive effect on the transportation system but also reduce its negative effect on land use is an urgent problem to be solved. Given this, this paper proposed a multi-objective bi-level programming road congestion pricing model based on the integrated land use and transportation model to optimize the regional average accessibility, regional average land use diversity, and regional total flow time. Since the proposed problem is NP-hard, this paper innovatively proposed an active learning optimization algorithm based on multi-objective Bayesian optimization, which improves the computation efficiency of the bi-level programming model by automatically finding the next sampling point (candidate solution) according to the probability information. An empirical analysis of Jiangyin City demonstrated the effectiveness of the proposed approach in coordinating the relationship between land use and transportation and alleviating the negative land use effect caused by road congestion pricing. Moreover, the algorithm proposed in this paper can also be used to solve other transportation-related black box problems with high computation complexity.