Connections between classical car following models and artificial neural networks

This article considers the problem of traffic modeling via modeling at the microscopic (i.e., vehicle) scale. It provides a connection between classical ordinary differential equation based models and data driven artificial neural network (ANN) based models by showing an example of a car following model which can be exactly expressed as an ANN. In a set of numerical experiments, four ANN models (ranging in structure from a model that is able to exactly capture a classical car following model, to a generic neural network model) are proposed and then trained from data and their resulting accuracy is assessed. It is shown that by adding structure into the neural network (i.e., via the architecture and the activation functions), it is possible to outperform generic ANN models to emergent phenomena such as stop and go waves.

[1]  Helbing,et al.  Congested traffic states in empirical observations and microscopic simulations , 2000, Physical review. E, Statistical physics, plasmas, fluids, and related interdisciplinary topics.

[2]  Kunihiko Fukushima,et al.  Neocognitron: A self-organizing neural network model for a mechanism of pattern recognition unaffected by shift in position , 1980, Biological Cybernetics.

[3]  Yoshua Bengio,et al.  Gradient-based learning applied to document recognition , 1998, Proc. IEEE.

[4]  J. Dormand,et al.  A family of embedded Runge-Kutta formulae , 1980 .

[5]  Maria Laura Delle Monache,et al.  Dissipation of stop-and-go waves via control of autonomous vehicles: Field experiments , 2017, ArXiv.

[6]  Jimmy Ba,et al.  Adam: A Method for Stochastic Optimization , 2014, ICLR.

[7]  Nakayama,et al.  Dynamical model of traffic congestion and numerical simulation. , 1995, Physical review. E, Statistical physics, plasmas, fluids, and related interdisciplinary topics.

[8]  G. F. Newell.  Nonlinear Effects in the Dynamics of Car Following , 1961 .

[9]  R. Jiang,et al.  Full velocity difference model for a car-following theory. , 2001, Physical review. E, Statistical, nonlinear, and soft matter physics.

[10]  Yoshua Bengio,et al.  Understanding the difficulty of training deep feedforward neural networks , 2010, AISTATS.

[11]  P. G. Gipps,et al.  A behavioural car-following model for computer simulation , 1981 .

[12]  Alexiadis,et al.  The next generation simulation program , 2004 .

[13]  Alexandre M. Bayen,et al.  Flow: Architecture and Benchmarking for Reinforcement Learning in Traffic Control , 2017, ArXiv.

[14]  Ali Ghaffari,et al.  A Modified Car-Following Model Based on a Neural Network Model of the Human Driver Effects , 2012, IEEE Transactions on Systems, Man, and Cybernetics - Part A: Systems and Humans.

[15]  Juan Zhicai,et al.  Develop a car-following model using data collected by "five-wheel system" , 2003, Proceedings of the 2003 IEEE International Conference on Intelligent Transportation Systems.

[16]  L. Shampine,et al.  Some practical Runge-Kutta formulas , 1986 .

[17]  Nicola Bellomo,et al.  On the Modeling of Traffic and Crowds: A Survey of Models, Speculations, and Perspectives , 2011, SIAM Rev..

[18]  Jürgen Schmidhuber,et al.  Long Short-Term Memory , 1997, Neural Computation.

[19]  Martin Treiber,et al.  Microscopic Calibration and Validation of Car-Following Models – A Systematic Approach , 2013, 1403.4990.

[20]  Bin Jia,et al.  A New Car Following Model: Comprehensive Optimal Velocity Model , 2011 .

[21]  L. A. Pipes.  An Operational Analysis of Traffic Dynamics , 1953 .

[22]  Xiaobo Qu,et al.  A recurrent neural network based microscopic car following model to predict traffic oscillation , 2017 .

[23]  Linsen Chong,et al.  Simulation of Driver Behavior with Agent-Based Back-Propagation Neural Network , 2011 .