2022 Fiscal Year Annual Research Report
Development of a socially aware crowd flow model and control of crowd flow based on reinforcement learning
| Project/Area Number |
22J20686
|
|---|
| Allocation Type | Single-year Grants |
|---|
| Research Institution | Okayama University |
|---|
Principal Investigator |
GREGORJ Adrien 岡山大学, 自然科学研究科, 特別研究員(DC2)
|
|---|
| Project Period (FY) |
2022-04-22 – 2025-03-31
|
| Keywords | pedestrian dynamics / collision avoidance / social interactions / crowd steering / gaze tracking |
|---|
| Outline of Annual Research Achievements |
Analyses of the impact of social relations and social interactions on pedestrian motion patterns have been pursued. In particular, the dynamics of collision avoidance between pedestrian two-person groups and individual pedestrians in an ecological environment have been studied. Our findings revealed that, as social bonding and interaction intensity increase, collision avoidance becomes more pronounced. Additionally, we observed that the likelihood of intrusion decreases as the strength of social bonding and interaction intensity increases.
Furthermore, to understand how visual stimuli could be used to influence pedestrian motion (steering, etc.), a preliminary study to evaluate differences in tracking performances when viewing natural motion against artificial motion has been designed.
|
| Current Status of Research Progress |
Current Status of Research Progress
2: Research has progressed on the whole more than it was originally planned.
Reason
The carried out studies aligned with the planned research and provided insights into the influence of social bonding within groups on the movement of individuals in crowds. This represents a significant advancement in the comprehension of how macro-level social behaviors can impact the dynamics of pedestrians. The potential application of these findings to develop more authentic crowd models and simulations is being evaluated.
Furthermore, initial investigations have been conducted on the impact of visual cues in regulating or directing crowd velocity. Notably, particular attention has been paid to selecting the appropriate type of stimuli, as previous literature has shown insufficient attention to this subject matter.
|
| Strategy for Future Research Activity |
To study how visual stimuli can be used to steer pedestrian individuals, a better understanding of human gaze mechanics is necessary. The mechanism of smooth pursuit will be examined. A study designed to evaluate differences in tracking performances when viewing natural motion against artificial motion will be conducted. Some preliminary studies are also necessary to validate the use of some of the experiment material (stimuli target, etc.). Based on the results of these studies, the use of visual stimuli to affect pedestrian motions will be evaluated. The type of stimuli to select for this purpose will be guided by these results (more natural, easier to track, and less intrusive).
The use of reinforcement learning algorithms for crowd modeling/steering algorithms will also be investigated.
|
