Depth-Map-Based Shape Recognition of Soft Continuum Manipulator Body

Abstract

Soft robotics is young and dynamically evolving field of scientific research. Soft robots are difficult to control and much more complex for modeling than traditional robots. They are, however, gaining more and more researchers’ interest due to their high potential. Soft robot body is made of flexible materials and it contains no rigid or discrete joints, its movement is generated by smooth and continuous deformation of the body. The softness of the body enables wide range of possible robot configurations which, on the other hand, are difficult to measure or predict. There were many shape detection approaches proposed so far, but most of them provide only discrete manipulator points position. Such a solution is not sufficient for some applications, since the manipulator body configuration in between the measured points is uncertain. In this paper Authors propose algorithm for continuous detection of the flexible manipulator shape. The algorithm is based on depth image provided by sensor such as Kinect. The depth image is processed in steps: detection of the manipulator central axis, the axis 3D shape reconstruction and orientation approximation for each point that lies on it. Numerical representation of the manipulator surface is also generated. The data obtained can be used for manipulator internal state correction including its input values and external disturbances.