Knowledge discovery, learning, and transfer from spatiotemporal big data represent a cutting-edge interdisciplinary challenge across geography, transportation, urban planning, computer science, and artificial intelligence. Spatiotemporal trajectory analysis, as a core component of such research, demands innovative methodologies to address its technical complexities. This project introduces advanced techniques from complex network science and reinforcement learning (RL) to achieve precise behavioral pattern mining and sequential decision-making simulation for human/vehicle trajectory data.
Key Innovations:
Unified Analytical Framework:
Addresses the heterogeneity of multi-source spatiotemporal data by developing a complex network-based framework to systematically convert trajectory data into network-structured representations.
Designs efficient algorithms for network analysis to extract latent behavioral patterns.
Inverse Reinforcement Learning (IRL) Integration:
Simulates sequential decision-making processes by coupling agent-environment interactions while preserving spatiotemporal continuity and minimizing data information loss.
Applications include travel demand forecasting, traffic congestion prediction, and policy evaluation for urban-transport systems.
Significance:
Methodological Breakthrough: Bridges gaps between data-driven modeling and theoretical frameworks in trajectory analytics.
Policy Impact: Enhances predictive accuracy for urban mobility systems, supporting data-informed governance.
Principal Investigator:
Dr. Wenjia Zhang
Duration: January 2019 – December 2020
