Real-time motion prediction based environment map

Abstract

In this dissertation, we propose graph based prediction map of human for robot navigation. The prediction of motions of human beings is very important for autonomous robots in dynamic environment. Previous researches of prediction of human motions summarized into 2 classes: grid based motion prediction and trajectory based motion prediction. Since the previous methods have a too many computation problem, we need to design more simplified model to predict model in real-time. Therefore, we assume that human beings have reason to move around the space as follows: 1. human beings stay to achieve specific goal, and 2. human beings move along the shortest path. By using these assumptions, we firstly design a graph model that includes information of human motion prediction. The time complexity of the human motion prediction map is faster than that of others in terms of finding future position of human beings. Besides, our method could guarantee real-time.