| Humans can effortlessly navigate through complex social environment.During this process,humans need to bypass physical obstacles,circumvent static or dynamic pedestrians,and finally reach the destination.Moreover,social cognition also plays an important role in human spatial locomotion.For example,people will try to avoid passing through talking groups,in order not to disturb their social interaction.Although the importance of social cognition in human spatial locomotion cannot be ignored,it receives little attention in previous studies.Neither has it been systematically investigated in laboratory environment,nor has it been quantified by computational models.The situation is mainly caused by both the lack of suitable theoretical framework to quantify the social impact of humans and the technical limitations on presenting close-to-real stimuli and meanwhile recording the spatial location of subjects.In our previous research,we proposed a social interaction field model which quantifies human social interaction behaviour.The model provides a feasible framework for studying human social impact during locomotion.The development of all-in-one virtual reality equipment has also made the study of human locomotion in controlled lab environment possible.Therefore,by combining the theoretical and technical basis,we conducted following experiments to study and quantify the impact of human social information on locomotion behaviour.In Experiment 1 to 5,we systematically investigated the human social impact on locomotion behaviour by manipulating the size of physical obstacles and the orientation of virtual humans in virtual environment.Through analyzing the trajectories and route selection behaviour of subjects,we found that subjects tend to bypass other pedestrians from their back sides[F（11,264）=37.090,p<.001]and take a larger detour if they have to bypass from the front side of them[F（11,264）=22.745,p<.001].We further demonstrated the effect cannot be explained by the motion prediction of other people[t（24）=-6.343,p<.001].By further analyzing the data from Experiment 1 to 5,we proposed the social energy model which borrows the concept of energy landscape in the field of ecology.The model analogized the social and physical impact of humans as a distribution of“energy”in the space around them.By doing so,the detour trajectory of human pedestrian can be generated by finding the route of the least energy cost with Fast Marching Method.The model performed well in fitting the trajectories of human subjects and succeeded in predicting their route selection behaviours.In Experiment 6 to 11,we investigated the robustness of the model by applying it to more complex social scenes.In Experiment 6 which subjects have to decide between costing extra physical energy and crossing from interacting social groups,we primarily demonstrated that the“social energy”and actual physical energy are comparable and transferable in the cognitive level.Then,we verified our model step by step in virtual scenes with one static virtual human,real-life scenes with one real human being,virtual scenes with two static virtual human,virtual scenes with one dynamic virtual human and real-life scenes with multiple dynamic real humans.In all challenging scenes,the model performed well with the exact same set of parameters,proving that the model is robust and can be extended to very complex social scenes.In Experiment 12,the model was transformed into computer algorithm and applied to a navigation robot platform.The robot was programmed to bypass subjects with different spatial locations and orientations.Results showed that the robot navigating with the social energy model outperformed the traditional SLAM navigation algorithms with both higher scores in comfortableness rating and higher probability of being judged as recovering the trajectories of real human beings.In conclusion,this study for the first time systematically investigated human social impact on locomotion behaviour and proposed the social energy model to compute the trajectory of human beings.The model not only revealed the coding algorithm of other people’s social impact during locomotion,but also provided valuable guides in designing socially-aware service robots. |
PDF Full Text Request