Recurrent Convolutional Neural Network に基づく一人称視点画像から のロボッ ト の自己位置推定

抄録

Many robots are pervading environments of human daily life. It is crucial for those robots to estimate the current self-positions. This is called robot localization. In this paper, we propose a method using a recurrent convolutional neural network (RCNN), which is known as one of deep learning, to achieve robot localization. RCNN is a neural network model that has a convolutional architecture known as CNN with recurrent nodes. We train RCNN to estimate the current position of a robot from the view images of the first person perspectives. Our experimental results show that the estimation error decrease when the successive view images are given and it can estimate the current position accurately.