When a critical amount of pedestrians herd together, pedestrians are showed several series of group behaviors. Existing dynamical model can duplicate pedestrian’s group behavior. But we create dynamical model in each case. Why do we need to create d...
When a critical amount of pedestrians herd together, pedestrians are showed several series of group behaviors. Existing dynamical model can duplicate pedestrian’s group behavior. But we create dynamical model in each case. Why do we need to create dynamical model in each case? We believe that we don’t understand what are the assumed forces and interactions. Our challenge is to explain forces and interactions between pedestrians in a more detailed, primitive level. We are using Direct-Vision-Based Reinforcement Learning to construct the primitive model. This learning system needs not a priori assumption ? only a reward.