D:aisy-T
D:aisy-T
Decision-making: automated & integrated system of Transportation
D:aisy-T 2023Transportation systems have witnessed disruptive changes in the past decades due to the Internet-of-Things (IoT) technology development and business model innovations. Those changes generated unprecedented amounts of data which provides great opportunities to better understand the evolving system and inform smart decision-making in transportation system management and planning.
An innovative, pragmatic, and forward-thinking paradigm is needed more than ever to tackle the desire for smart urban transportation systems. The Decision-making: automated & integrated system of Transportation (D:aisy-T) is a research framework to address emerging challenges in smart cities.
D:aisy-T, enabled by data science, economic theory, computational simulation, and advanced transportation modeling techniques, provides opportunities for decision-makers to understand the complex interconnections between people and transportation systems. D:aisy-T integrates ML/AI methods with classic transportation models/theories and develops a transportation system digital twin that automatically learns from multi-source big data and adaptively simulates the dynamic system with a high fidelity.
D:aisy-T tackles future urban challenges arising from New Mobility Options Evaluation and Regulation, Built Environment & Public Policy Assessment, Energy Consumption Estimation, Equity Analysis in Transportation, Resilience in Urban Transportation Systems, and Sustainability and Environmental Analyses.
About
Research Framework
D:aisy-T enables the interplay between the physical system and digital twin of urban transportation systems. It is a highly interactive, automated, and integrated decision-making support framework. It encompasses four major components and two ancillary modules.
Major Components
Learning-based Data Mining: ML/AI techniques are adopted to learn from multiple data sources (e.g., road detectors, CAVs, weather sensors, and third-party providers) and support the development, calibration, and adaption of integrated transportation models. Moreover, data-driven decision-making is also incorporated for some purposes, such as time-variant traffic speed/volume prediction, etc.
Integrated Models: Integrated models simulate the dynamic interactions between demand and supply and produces outputs at both system and individual levels by integrating activity-based behavior models and dynamic network models. The integrated models are the core of the transportation system digital twin.
High-Fidelity Analytics: Based on the high-fidelity outputs from integrated models, the system status is evaluated with multiple performance metrics to support the decision-making with different purposes. Built upon the digital twin, the analytics layer provides an efficient alternative to traditional assessment approaches (e.g., filed pilot and market deployment) and saves massive time and financial expenses.
Decision-making in Digital Twin: Using the digital twin as a virtual environment, the decision-making layer adopts simulation-based optimization and reinforcement learning methods to search for the optimal/near-optimal planning policies and management strategies for different purposes, from traditional purpose like Travel Demand Management (TDM) to emerging purpose as new mobility operation and regulation.
Ancillary Modules
Scenario Manager: The scenario manager adapts the digital twin to different decision-making application scenarios by controlling the configuration and adding necessary modules. Regarding the digital twin as a living laboratory, the scenario manager can also launch some synthetic scenarios to produce data and inform decision-making in the future.
Optimal Learning: To improve the performance and accuracy of the analytics layer and the model layer, the optimal learning module estimates the data needs and guides the data layer to automatically acquire or synthesize specific data for the digital twin or data-driven decision-making.
Example Projects
Hidden Activity Signal and Trajectory Anomaly Characterization
The advancement of the Internet of Things and Smart City infrastructures has resulted in an abundance of data on human mobility. This data offers the potential to develop new models for understanding human dynamics at an unprecedented level of detail. However, this also raises the important issue of privacy concerns within a sensor-rich environment. Current modeling capabilities primarily focus on aggregated aspects such as population migration and disease spread, lacking the ability to capture the intricacies of daily human movement. The main challenge lies in the absence of comprehensive datasets to support the development of artificial intelligence in understanding detailed trajectories. The project seeks to overcome this limitation by creating a large-scale microsimulation of background activity and associated trajectories, integrating specific movement activities, and endeavoring to distinguish inserted activities from the background.