Rapid, artificial intelligence estimates of Los Angeles highway flows: Applications to seismic risk analysis and congestion pricing

Vehicles competing for a limited supply of road space impose time costs on movement. Economists propose congestion pricing to obtain the efficient use of roadways under a price system. System-wide traffic flows under pricing systems are difficult to predict because of the complexity of drivers' behavioral changes, the lack of empirical data, and the limitations of current traffic flow modeling approaches.; Earthquakes damage freeway bridges and structures, resulting in significant impacts on urban and transportation systems. Seismic risk analysis (SRA) procedures establish retrofit priorities for vulnerable highway bridges. Existing SRA procedures use average daily traffic volumes to determine the relative importance of a bridge. This is not adequate. The importance of network links should be evaluated in terms of additional system costs due to failure.; The objectives of this research are: (1) to develop efficient transportation network analysis procedures for traffic flow analyses, and (2) to evaluate the applicability of the procedures to a large-scale transportation network. An important feature of the TNA procedure is the use of an associative memory approach. Two TNA procedures are developed based on the quality of transportation system data sets: the general TNA procedure and the simplified TNA procedure.; A simple synthetic transportation network and an aggregated Los Angeles highway network are developed and used to evaluate the TNA procedures. We combine synthetic, network equilibrium transportation flows with empirical data associated with the opening of the Glen Anderson Freeway (I-105) and the 1994 Northridge earthquake. Synthetic system-optimal traffic flows are also simulated. Associative memory models are applied to estimate empirical and/or synthetic traffic flows.; Results from traffic flow analyses demonstrate the applicability of the TNA procedures to network flow estimation. Associative memory models provide reliable flow estimates compared to conventional network equilibrium models. The performance of associative memory models improves if empirical and synthetic system states are combined. Results from travel demand analyses indicate that a small change in network links may cause a significant impact on travel demand.