Travel mode detection is a crucial task for transport demand management and understanding mobility patterns. Lack of data makes the mode detection task challenging to detect new transport modes and track trip chains. Thus, researchers and practitioners try to identify new data sources and plan optimal data collection infrastructure to develop advanced and robust models. In the MODA project, we will tackle travel mode detection from both the data as well as the modeling side. We aim to design and implement a simulation test bed to collect (generate) the mobility data covering new and traditional transport modes. We will design and test the opportunistic data collection framework to reflect the real-world challenges of data scarcity, sparsity, quality, and technical feasibility. Use of wavelet transformation and signal processing methods will be used to process and filter the data. We will propose and test state-of-the-art deep learning models to detect travel modes with high accuracy and robustness while addressing data quality issues. These models will be validated with respect to the simulation ground truth, as well as real-world data. Finally, models will be transferred to new contexts and study areas to tackle the lack of data. Thus, our proposal tackles the travel model detection holistically and has potential practical benefits to the transport research community. It provides a wider framework to apply mode detection using free and opportunistic data, assesses if these data are sufficient for the task, and investigate the limits of the data.

Yang, N., Lu, Q. L., Lyu, C., & Antoniou, C. (2024). Transfer Learning for Transportation Demand Resilience Pattern Prediction Using Floating Car Data. Transportation Research Record. https://doi.org/10.1177/03611981241245681