Time-varying Pedestrian Flow Models for Service Robots

Abstract

We present a human-centric spatio-temporal model for service robots operating in densely populated environments for long time periods. The method integrates observations of pedestrians performed by a mobile robot at different locations and times into a memory efficient model, that represents the spatial layout of natural pedestrian flows and how they change over time. To represent temporal variations of the observed flows, our method does not model the time in a linear fashion, but by several dimensions wrapped into themselves. This representation of time can capture long-term (i.e. days to weeks) periodic patterns of peoples' routines and habits. Knowledge of these patterns allows making long-term predictions of future human presence and walking directions, which can support mobile robot navigation in human-populated environments. Using datasets gathered by a robot for several weeks, we compare the model to state-of-the-art methods for pedestrian flow modelling.

Human motion trajectory prediction: a survey

“…In safety-critical applications, reachability-based methods provide a guarantee on local collision avoidance (Bansal et al, 2019). Furthermore, understanding local motion patterns is useful for compliant and unobstructive navigation (Palmieri et al, 2017; Vintr et al, 2019).…”

Robotic Exploration for Learning Human Motion Patterns

“…The spatio-temporal exploration strategies compared in the experimental part of this paper aim to build the STeF-Map model introduced in [8], which creates a time-dependent probabilistic map able to model and predict patterns of people in indoor environments from sparse data. This representation is compared with other similar maps of human dynamics in the literature, where it achieves favourable performance not only in terms of prediction accuracy using the Chi-squared metric [33], but also, it allows a more human compliant path planning and scheduling of robot activities in human-populated environments [34]. However, introducing the temporal dimension to robot mapping requires novel exploration strategies to build and maintain the spatio-temporal models during the robot's deployment in a given environment [35].…”

3D ToF LiDAR in Mobile Robotics: A Review

“…In particular [51] shows that long-term 3D LiDAR observations can be used not only to update the scene geometry, but also to segment dynamic objects like pedestrians or cars. The information about dynamic objects can be further processed to identify dominant motion flows [50], [52], [54], which can be used to improve robot navigation [56], [91] in crowded environments. Apart from detecting the change [92], long-term, 3D observations allow to learn the temporal patterns of the changes and use this knowledge to forecast the environmental states [53], [93], which improves the performance of mobile robots in the long-term.…”

“…The success of Junior also directly affects and promotes the development of 3D LiDAR in autonomous driving [9], [6], [16], [17], [18], [19], [20], [21], [106], [59], [119], [108]. On the other hand, the demand for unconventional sensors typified by 3D LiDAR is also present in the field of service robots [14], [15], [36], [37], [87], [57], [56], [4], [54], [55], [52], [53]. Furthermore, in the face of the recent severe worldwide public health crisis, namely COVID-19, 3D LiDAR has also shown encouraging results and broad application prospects.…”