Research Summary


Online multi-modal-variate learning – non-i.i.d. – multiagent – sampling

I make proactive dynamic distributed decisions by integrating different types of resources (aerial + ground + CAV, electric vehicles) and modality (on-demand + walk), sensing (mobile + fixed) under uncertainties caused by heterogenous users (informed + uninformed) and mutually dependent events (assuming non i.i.d).

The new foundational approach analyzes unobserved heterogeneity in transportation data as mixture of multimodal multivariate traffic parameters simultaneously sampled from more than one peak in probability distribution and more than one variable among travel time, flow, density, jam density, headway data, etc. For example, instead of a realization of travel time of nearby links for online updating route choice, I analyze each cell as multivariate multimodal probability distributions, aggregate all cells in a network, group cells with similar type mixture distribution across different far distance, non-contagious locations, and approximate the posterior when I have new observations.

Online multimodal learning can improve the mobility prediction in the backend cloud source (passive) or we can optimize the route or mobile sensing (active) to maximize the improvement in the uncertainty in prediction have been applied for smart technologies in resilient, equitable, and sustainable cities issued for two patents (US Patent 10,743,198 for mobile sensing and US Patent 11,046,247 for in-vehicle monitoring).

This new approach can understand the dynamics of network behavior and monitor the resilience of infrastructure to enable intelligent decision-making. Deep semantic learning could extract spatiotemporal interactions between heterogeneous agents, predict the future network behaviors considering dynamic changes from disruptions, and enhance city resilience by systematically responding to events, optimize energy-aware mobility, and minimize social inequity’s disproportionate impact on underserved communities in a sustainable manner. The city accessibility was significantly improved by integrating mobility preference of vulnerable road users on mixed sidewalks and on-transit reliability. Safe and effective evacuation plans were developed by considering the motion of pedestrian dynamics subject to social forces.


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: