Handrail detection and pose estimation for a free-flying robot. (16th January 2018)

• Record Type: Journal Article
• Authors: Lee, Dong-Hyun; Coltin, Brian; Morse, Theodore; Park, In-Won; Flückiger, Lorenzo; Smith, Trey
• Is Part Of: International journal of advanced robotic systems. Volume 15:Number 1(2018:Jan./Feb.)
• Publication Date: 2018-01-16
• Subjects: Space robots -- free-flyers -- object detection -- pose estimation; Robotics -- Periodicals; Robotics; Periodicals; 629.892
• DOI: 10.1177/1729881417753691
• Languages: English
• ISSNs: 1729-8806

Abstract

We present a handrail detection and pose estimation algorithm for the free-flying Astrobee robots that will operate inside the International Space Station. The Astrobee will be equipped with a single time-of-flight depth sensor and a compliant perching arm to grab the International Space Station handrails. Autonomous perching enables a free-flying robot to minimize power consumption by holding its position without using propulsion. Astrobee is a small robot with many competing demands on its computing, power, and volume resources. Therefore, for perching, we were limited to using a single compact sensor and a lightweight detection algorithm. Moreover, the handrails on the International Space Station are surrounded by various instruments and cables, and the lighting conditions change significantly depending on the light sources, time, and robot location. The proposed algorithm uses a time-of-flight depth sensor for handrail perception under varying lighting conditions and utilizes the geometric characteristics of the handrails for robust detection and pose estimation. We demonstrate the robustness and accuracy of the algorithm in various environment scenarios.