Deep Trail Following Robotic Guide Dog in Pedestrian Environments for People Who Are Blind and Visually Impaired

• Main Authors: Chen, Jih-Shi, 陳季希
• Other Authors: Wang, Hsueh-Cheng
• Format: Others
• Language: zh-TW
• Published: 2018
• Online Access: http://ndltd.ncl.edu.tw/handle/djg74m

碩士 === 國立交通大學 === 電控工程研究所 === 106 === Navigation in pedestrian environments is critical to enabling independent mobility for the blind and visually impaired (BVI) in their daily lives. White canes have been commonly used to obtain contact feedback for following walls, curbs, or man-made trails, whereas guide dogs can assist in avoiding physical contact with obstacles or other pedestrians. However, the infras- tructures of tactile trails or guide dogs are expensive to maintain. To solve these problems, we develop a robotic guide dog, use yellow-blue line and Boston Freedom trails as following trails. In the previous works, camera observations may vary when man-made trails are deployed on various background textures under different illuminations or shadows. We proposed an au- tonomous, trail-following robotic guide dog that would be robust to variances of background textures, illuminations, and interclass trail variations. A deep convolutional neural network (CNN) is trained from the real world environments. The results show that our robotic guide dog may run in different background textures and illuminations. This thesis conducted user studies with 10 BVI users, who did not have prior experience with our system. Each participant was introduced to the tasks and signed a consent form. We also used a questionnaire to capture the users’ subjective experience. All 10 participants completed the questionnaires after the experiments, and all of then are able to build and draw a mental map.