Using EM to learn motion behaviors of persons with mobile robots

Abstract

We propose a method for learning models of people's motion behaviors in indoor environments. As people move through their environments, they do not move randomly. Instead, they often engage in typical motion patterns, related to specific locations that they might be interested in approaching and specific trajectories that they might follow in doing so. Knowledge about such patterns may enable a mobile robot to develop improved people following and obstacle avoidance skills. This paper proposes an algorithm that learns collections of typical trajectories that characterize a person's motion patterns. Data, recorded by mobile robots equipped with laser-range finders, is clustered into different types of motion using the popular expectation maximization algorithm, while simultaneously learning multiple motion patterns. Experimental results, obtained using data collected in a domestic residence and in an office building, illustrate that highly predictive models of human motion patterns can be learned.

Mobile robot mapping in populated environments

“…Furthermore, there has been developed a variety of techniques for tracking persons, for predicting future poses of persons, or for adopting the behavior of the robot according to the information obtained about the persons in its vicinity [29,19,34,20,5,17,18,16,32,4,2,28]. All these approaches, however, do not filter the measurements corresponding to persons in order to improve the model of the environment.…”

Acquiring After-Sales Knowledge from Human Motions

“…Bennewitz [8] also proposes a learning method of human motion for robot teaching. Lee and Mase [9] developed a system of activity and location recognition using wearable sensors, which includes IC accelerometers.…”

Learning hierarchical models of activity

“…We also ran experiments on data collected by Bennewitz et al [4]. The data consist of sequences of a person moving around a house.…”

“…However, in this work the person is tracked based on a motion model of typical human movement: the intentions or the higher level task of the person is not modeled. Other work has been done that does model intention [3], [4]. This work clusters similar motion trajectories but does not model activity in a hierarchical manner.…”