Showing papers by "Xiaolei Ma published in 2015"

Long short-term memory neural network for traffic speed prediction using remote microwave sensor data

Abstract: Neural networks have been extensively applied to short-term traffic prediction in the past years. This study proposes a novel architecture of neural networks, Long Short-Term Neural Network (LSTM NN), to capture nonlinear traffic dynamic in an effective manner. The LSTM NN can overcome the issue of back-propagated error decay through memory blocks, and thus exhibits the superior capability for time series prediction with long temporal dependency. In addition, the LSTM NN can automatically determine the optimal time lags. To validate the effectiveness of LSTM NN, travel speed data from traffic microwave detectors in Beijing are used for model training and testing. A comparison with different topologies of dynamic neural networks as well as other prevailing parametric and nonparametric algorithms suggests that LSTM NN can achieve the best prediction performance in terms of both accuracy and stability.

Large-scale transportation network congestion evolution prediction using deep learning theory.

Abstract: Understanding how congestion at one location can cause ripples throughout large-scale transportation network is vital for transportation researchers and practitioners to pinpoint traffic bottlenecks for congestion mitigation. Traditional studies rely on either mathematical equations or simulation techniques to model traffic congestion dynamics. However, most of the approaches have limitations, largely due to unrealistic assumptions and cumbersome parameter calibration process. With the development of Intelligent Transportation Systems (ITS) and Internet of Things (IoT), transportation data become more and more ubiquitous. This triggers a series of data-driven research to investigate transportation phenomena. Among them, deep learning theory is considered one of the most promising techniques to tackle tremendous high-dimensional data. This study attempts to extend deep learning theory into large-scale transportation network analysis. A deep Restricted Boltzmann Machine and Recurrent Neural Network architecture is utilized to model and predict traffic congestion evolution based on Global Positioning System (GPS) data from taxi. A numerical study in Ningbo, China is conducted to validate the effectiveness and efficiency of the proposed method. Results show that the prediction accuracy can achieve as high as 88% within less than 6 minutes when the model is implemented in a Graphic Processing Unit (GPU)-based parallel computing environment. The predicted congestion evolution patterns can be visualized temporally and spatially through a map-based platform to identify the vulnerable links for proactive congestion mitigation.

Two-echelon logistics distribution region partitioning problem based on a hybrid particle swarm optimization-genetic algorithm

Abstract: A logistics distribution region partitioning model is developed.This model is to minimize the cost of two-echelon logistics distribution network.A hybrid algorithm with PSO and GA is proposed.The empirical results reveal that EPSO-GA algorithm outperforms other algorithms. Two-echelon logistics distribution region partitioning is a critical step to optimize two or multi-echelon logistics distribution network, and it aims to assign distribution unit to a certain logistics facility (i.e. logistic center and distribution center). Given the partitioned regions, vehicle routing problem can be further developed and solved. This paper established a model to minimize the total cost of the two-echelon logistics distribution network. A hybrid algorithm named as the Extended Particle Swarm Optimization and Genetic Algorithm (EPSO-GA) is proposed to tackle the model formulation. A two-dimensional particle encoding method is adopted to generate the initial population of particles. EPSO-GA combines the merits of Particle Swarm Optimization (PSO) algorithm and Genetic Algorithm (GA) with both global and local search capability. By updating the inertia weight and exchanging best-fit solutions and worst-fit solutions between PSO and GA, EPSO-GA algorithm is able to converge to an optimal solution with a reasonable design of termination and iteration rules. The computation results from a case study in Guiyang city, China, reveal that EPSO-GA algorithm is superior to the other three algorithms, Hybrid Particle Swarm Optimization (HPSO), GA, and Ant Colony Optimization (ACO), in terms of the partitioning schemes, the total cost and number of iterations. By comparing with the exact method, the proposed approach demonstrates its capability to optimize a small scale two-echelon logistics distribution network. The proposed approach can be readily implemented in practice to assist the logistics operators reduce operational costs and improve customer service. In addition, the proposed approach is of great potential to apply in other research domains.

A Methodology to Exploit Profit Allocation in Logistics Joint Distribution Network Optimization

Abstract: Logistics joint distribution network (LJDN) optimization involves vehicle routes scheduling and profit allocation for multiple distribution centers. This is essentially a combinational and cooperative game optimization problem seeking to serve a number of customers with a fleet of vehicles and allocate profit among multiple centers. LJDN routing optimization based on customer clustering units can alleviate the computational complexity and improve the calculation accuracy. In addition, the profit allocation mechanism can be realized based on cooperative game theory through a negotiation procedure by the Logistics Service Provider (LSP). This paper establishes a model to minimize the total cost of the multiple centers joint distribution network when each distribution center is assigned to serve a series of distribution units. An improved particle swarm optimization (PSO) algorithm is presented to tackle the model formulation by assigning distribution centers (DCs) to distribution units. Improved PSO algorithm combines merits of PSO algorithm and genetic algorithm (GA) with global and local search capabilities. Finally, a Shapley value model based on cooperative game theory is proposed to obtain the optimal profit allocation strategy among distribution centers from nonempty coalitions. The computational results from a case study in Guiyang city, China, suggest the optimal sequential coalition of distribution centers can be achieved according to Strictly Monotonic Path (SMP).

Vehicle routing problem based on a fuzzy customer clustering approach for logistics network optimization

Abstract: 7 Abstract. Vehicle routing problem (VRP) is a combinatorial optimization and integer programming problem seeking to service a number of customers with a fleet of vehicles. Customer characteristics are neglected in traditional VRPs in the past due to the heterogeneity and ambiguousness. This study presents a vehicle route optimization model in consideration of customer characteristics with three major components: (1) A hierarchical analysis structure is developed to convert customers' characteristics into linguistic variables, and fuzzy integration method is used to map the sub-criteria into higher hierarchical criteria based on the trapezoidal fuzzy numbers; (2) A fuzzy clustering algorithm based on Axiomatic Fuzzy Set is proposed to group the customers into multiple clusters; (3) The fuzzy technique for order preference by similarity to ideal solution (TOPSIS) approach is integrated into the dynamic programming approach to optimize vehicle routes in each cluster. A numerical case study in Anshun, China demonstrates the advantages of the proposed method by comparing with the other two prevailing algorithms. In addition, a sensitivity analysis is conducted to capture the impacts of various evaluation criteria weights. The results indicate our approach performs very well to identify similar customer groups and incorporate individual customer's service priority into VRP. 8 9 10 11 12 13 14 15 16 17 18