Abstract
Safety is the most important to the mobile robots that coexist with human. In order to put those into practice, mobile robots need to plan trajectory considering human trajectory. In the field of mobile robots, there are a lot of trajectory planning methods that take account of human trajectory. However, it is still difficult to avoid human in the corner because there are blind areas in mounted sensors. Moreover it is difficult to predict human trajectory in the corner. This study proposes a predictive method of human trajectory by environmental sensors. First, we gathered human trajectory data at a crossroads using environmental sensors and estimated position of gravity point of human using Extended Kalman Filter. Second, we constructed human transfer models using Vector Auto Regressive Models. Third, we predicted human trajectory using the constructed models. Finally, we validated the constructed models comparing with predicted trajectory and experimental data.
