Improvements of Traffic Flow Simulation Models Using Some Artificial Intelligence Techniques

点击翻译按钮获取中文摘要

This project aims to improve traffic flow simulation models for freeways and arterials with the aid of someartificial intelligent techniques. Itis diviede into three parts :1)Description of Macroscopic Relationships Among Traffic Flow Variables Using Neural Network Model.The relationships among traffic flow variables play important roles in traffic flow simulation models. A procedure was presented to describe the macroscopic relationships between traffic flow variables using some neuralnetwork models. First, a Kohonen Feature Map model was introduced to convert original observed data points into fewer, more uniformly distributed ones. This conversion improved regression precision and computational efficiency . Next, a multilayr neural network model was introduced to describe the two-andthree-dimensional relationships. The model was effective in describing the non-linear and discontinuous characteristics between traffic flow variables.2)A Neural-Kalman Filtering Method for Estimating Tr … More affic StatesBy integrating multilayr neural network models into a Kalman filtering technique, a procedure for estimating traffic ststes was proposed . That is, The Cremer model, which is a macroscopic traffic flow model combined with a Kalman filter, is revised using a neural network model. The observation equations that relate the state variables, such as density and space mean speed, to the observation variables, such as flow rate and time mean speed, were described accurately using a neural network model. The derivatives of both state and observation equations were easily obtained, too. This neural-kalman method was applied to a road section on the Metropolitan Expressway in Tokyo and it was examined how precisely the method could work as compared with the original Cremer model.3)Artificial Intelligence Approach for Optimizing Traffic Signal Timing on Urban Road NetworkUsing artificial intelligence techniques, a stepwise method was developed to optimize signal timing parameters, such as splits and offsets, on an urban street. The method is separated into two processes, a training process and an optimization process. In the training process, we used two neural network models, a multilayr model and Kohonen Feature Map model. The former modelbuilds an input-output relationship between the signal timing parameters and the objective variable. The latter model improves the computational efficiency and the estimation precision. In the optimization process, to avoid the entrapment into a local minimum, two artificial intelligence methods were used ; the Cauchy machine and a genetic algorithm . The timing parameters were adjusted so as to minimize the total weighted sum of delay time and stop frequencies . The solutions by both artificialintelligence methods were compared with those by a conventional method and confirmed that they were useful for establishing advanced traffic control systems in the future . Less