Caliper Corporation Selected by FHWA to Lead a Project on Incorporating Artificial Intelligence (AI) in Travel Forecasting

NEWTON, MA (USA) - Caliper Corporation has recently been selected by the Federal Highway Administration (FHWA) to lead a project designed to improve travel forecasting through the incorporation of Big Data and AI.

Artificial Intelligence (AI) and big data provide what are probably the two most promising opportunities to improve travel forecasts today. To explore these opportunities in a rigorous manner, FHWA has reached out to Caliper Corporation with the aim of fundamentally improving travel forecasting in U.S. practice. The project hopes not to simply make a marginal improvement, but to substantially improve the accuracy and usefulness of travel forecasts used in long range transportation planning through the incorporation of AI and big data, revolutionizing this field just as they have revolutionized computer vision and natural language processing.

Given the project's focus on practical improvements together with the understanding that models used for public policy analysis need to be behaviorally explainable and defensible (not total black boxes), the study is particularly focused on Artificial Intelligence – Discrete Choice Models (AI-DCMs). The integration of AI in DCMs and training these models with big data has particular promise for several reasons. First, in a variety of different ways, different types of hybrid AI-DCMs can combine the attractive features of these two classes of models. On the one hand, they can capitalize on the predictive power of AI, while at the same time, preserving the interpretability and behavioral realism of traditional DCMs such as logit models. Second, existing travel demand models – both trip-based and activity-based – are generally conceived and implemented as a series of discrete choice models. Therefore, the incorporation of AI in discrete choice models does not require changes to the overall structure or framework of existing travel demand models. Agencies could continue to use their trip-based or activity-based models, just with AI-enhanced components.

The project is beginning with a review and formal meta-analysis of the published literature. The most promising methods will be recommended for testing. Required data and computing requirements will be considered together with accuracy in evaluating the methods and recommending which should be carried forward for implementation. The results of this testing with real regional datasets will identify final methods to be implemented in pilot projects. These pilot projects will be written up as case studies and presented in a series of webinars. .