Trajectory data collecting and sharing:Where are we? Where shall we go?

Purpose: Many trajectories data sets have been openly provided by different teams worldwide during the last decade. The experimental design, data acquisition setup, sensor placement, data analysis and presentation is biased to the ultimate scope of research, financial support, and experimental constraints. While this is understandable, there is a clear need for the adoption of a common and generic template that will act as an umbrella for the presentation of any new dataset. This is important to ensure transparency regarding data acquisition and limitations, applicability in terms of research investigations, complementarity with existing available resources and the realization of benchmarks for different research topics. This paper aims to contribute towards this cause. Method: A common template is used to describe each of recent nine datasets, representative of different data collection methods and research purposes. For each dataset, a dedicated section identifies the primary sources of data errors or incompleteness. In addition, a methodological framework that systematically retraces the full process of trajectory dataset collection, from preparation to sharing, has been identified. This comprehensive approach made it possible to draw lessons and formulate proposals at each stage of the dataset life cycle. Finally, a collective discussion process among all the authors was conducted to define future data collection methods and how to improve data sharing. Results: This paper produces a structured synthesis of lessons and key observations drawn from open-source trajectory datasets. It develops a taxonomy that retraces the full lifecycle of trajectory dataset collection, from preparation to sharing, supported by a critical discussion of current practices. Building on this framework, the paper delivers recommendations and discussion at two levels: (i) general recommendations on data collection and sharing, and (ii) recommendations for new types of trajectory datasets. Discussion: Although the datasets presented and discussed here are only a subpart of the numerous undergoing efforts for trajectory data collection and sharing across the globe, their representativeness is sufficient for a clear definition of guidelines for future initiatives, helping the community to build richer, more interoperable datasets without repeating past limitations.