🛰️ mobility patterns and behaviors from GPS trajectory: smart-grab collaboration

[hyunjimoon]

Thanks to Yanling Liu's introduction and Dingyi's followup, we had a first official meeting and among the topics, i'm most interested in sequential trajectory calibration techniques based on GPS data. Exposure to this data would help me flesh out GPS analogy built in #159 (comment). The idea is for startups which is shared vehicle that improves social mobility of its riders (entrepreneurs), GPS recommends paths based on multiple information on need, solution, technology, customer, industry, product, market etc.

Problem 6: Understand human mobility patterns and driving behaviors
Measuring Grab GPS trajectory data quality and calibrating trajectory sequences for human mobility patterns and driving behaviors
The proposed research aims at high-resolution trajectory data quality assessment and sequential trajectory calibration, providing insights into the analysis of human mobility patterns and driving behaviors.
(i) Investigating the data quality issues of Grab GPS trajectory data that are caused by unclear accuracy levels and limited time resolution of GPS pings.
(ii) Calibrating sequential trajectory of large-amount of trips by leveraging both historical and incremental data. Monitor road
network traffic states and optimize the efficiency of ridesharing fleet operations.

The challenges include:
(i) How to measure the data quality of GPS trajectories on such big data?
(ii) How to reduce the systematic data biases and improve the robustness of spatiotemporal
sequence models?
(iii) How to support the network-wide traffic state estimation on the sparse
movements of Grab vehicles?

The main tasks are
(i) designing the metrics for measuring the data quality of GPS trajectories,
(ii) building a flexible machine learning framework for calibrating GPS trajectories in a real-time
fashion that can reduce both data and model biases, and
(iii) developing a sparse sequence learning model for efficiently estimating the network traffic states from sparse movements. These tasks will be implemented through state-of-the-art artificial intelligence (AI) algorithms, fully utilizing the
trajectory data of Grab vehicles.
Anticipated research outputs include offering the calibrated and accurate trajectory data, improving the service quality for customers (e.g., accurate information), and better matching the travel demand. As one has accurate traffic states (e.g., movement speed) that are estimated from calibrated trajectories, it is possible to optimize the efficiency of ridesharing fleets, e.g., "shortest" paths for drivers and cheapest fee for customers.