大量の人移動計測データに基づく移動ロボットの人回避経路計画

概要

論文の詳細を見る
We propose a path planning method for mobile robots to avoid a person dynamically using movement prediction model from massive captured people trajectories. The captured trajectories are represented as a sequence of grid cells. The method learns these sequences using Variable Length Markov Model (VLMM), which predicts human location from the last human path. Based on the learnt model, the method plans mobile robot path on the XYT configuration space that includes predicted human position. In the experiment, we collected the trajectory data in the corridor space for over a month and simulated our planning method. The result showed that human trajectory model is more suitable than linear motion model for the safety robot navigation and the longer observation of trajectories improved the safety at mobile robot navigation.