Fabrication and Experiment of an Automatic Continuum Robot System Using Image Recognition

Abstract

Abstract This paper proposes a continuum robot that can be controlled automatically using image recognition. The proposed robot can operate in narrower spaces than the existing robots composed of links and joints. In addition, because it is automatically controlled through image recognition, the robot can be operated irrespective of the human controller's skill level. The manipulator is divided into two stages, with three wires connected to each stage to minimize the energy used to control the manipulator posture. The manipulator's posture is controlled by adjusting the length of the wire, similar to the relaxation and contraction of the muscles. Denavit–Hartenberg transformation and the Monte Carlo method were used to analyze the robot's kinematics and workspace. In a performance test, an experimental plate with nine targets was fabricated and the manipulator speed was adjusted to 5, 10, and 20 mm/s. Experimental results show that the manipulator was automatically controlled and reached all targets, with errors of 2.58, 3.28, and 9.18 mm.