Automatic Vehicle Identification (AVI) data is used to identify the location of a particular vehicle in and can also be used for route choice behaviour modelling. But the use of AVI doesn’t provide accurate information on OD pair and the particular route that is chosen. This problem is addressed in this paper using a semi-supervised learning method which can be used to identify the route on prior training. As the first step, the AVI trace is segregated into observation pairs using the Maximum Likelihood Estimation and then it is further joined with GPS co-ordinates to tackle the sparse issues. As the next step, the heterogeneity and correlation between the various pairs are determined using Mixed Logit model. As the final step, a relationship between the likelihood function and route choice model is established using Maximum to log-likelihood function. Based on the observations, the results are recorded and the proposed work shows significant improvement in the accuracy in route determination. The evaluation scenario shows that the proposed work could be expanded to a larger area. Moreover, the robustness of the system is illustrated using sensitivity analysis. This work uses AVI data with respect to its behaviour in routes through high penetration.
Route Choice Behaviour Modeling using IoT Integrated Artificial Intelligence
