Elder-oriented Active Learning for Adaptation of Perception Intelligence in Home Service Robots

Abstract

Active learning is a special case of machine learning in which a learning algorithm can interactively query a user to label new data points with the desired outputs. In robotics, active learning allows a robot to adapt its perception intelligence to a new environment with users’ help. This paper presents a new active learning method for elderly care robots to select data that is not only useful for learning but also easy for the elderly user to label. First, a series of image properties related to annotation difficulty are determined based on existing medical researches in human vision in elderly population. Based on that, a user study is conducted to determine the ground truth of annotation difficulty of images for the older adults. Second, a robust annotation difficulty predictor is developed using the results of the user study, and the difficulty prediction of an image is combined with three other active learning criteria to form an annotation difficulty-aware active learning metric, which facilitates the query data selection as the robot adapts its perception intelligence in a home environment. Third, we present an ablation study of the proposed active learning method through a simulation experiment. The experimental results validate the advantages of the proposed method.