To efficiently operate in human-populated environments, mobile robots benefit from the use of models that represent not only the environment structure but also human behaviour. One class of these models aims to capture the observed spatial and temporal layout of pedestrian flows to make this knowledge available to improve the way the robots move around humans. We are experimentally verifying a new benchmarking methodology of spatio-temporal models. We argue that these should be evaluated in terms of the afforded socially-compliant navigation of mobile robots in human-populated environments and not just by the quality of their predictions. To learn more about the research, see our paper in pedestrian flow modeling [1].
- Vintr, Molina et al: Time-varying pedestrian flow models for service robots In proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2016. [bibtex]