Two dimensional estimation of speed flow relationships with backpropagation neural networks

This chapter presents a method of estimating the speed flow density relationships from locally measured data sets using an analogy of backpropagation neural networks. The problem of steady state data is not discussed further. Here, an off-line least squares estimation of fundamental traffic flow relationships from two local data sets is made with three different methods by using an analogy of backpropagation neural networks. The examples given are based on traditional local data with dynamic fluctuations, and the relationships obtained should not be regarded as steady state estimates but only as examples of the estimation procedure. The backpropagation neural networks proved to be a useful and functioning tool in the estimation of the parameters of the fundamental relationships. A separate network with varying complexity is needed for each mathematical form of the fundamental diagram included in the analysis.