Research on the image-based visual servoing framework considering the aerial manipulation

Abstract

In this thesis, a vision-enabled autonomous aerial parallel manipulator and the novel image-based visual servoing (IBVS) framework for the precise relative motion control are proposed. The main research purpose of existing aerial manipulators which has the serial manipulator is to maximize the hovering stabilization performance of the host vehicle in order to utilize the large workspace of its manipulator. On the other hands, the main research purpose of the proposed platform, as called the aerial parallel manipulator which has the parallel manipulator, is to perform the relative motion control of the host vehicle more actively by calculating the guidance command throughout the movement of the end-effector. In this thesis, the novel image-based visual servoing (IBVS) framework are proposed in order to perform the relative motion control more actively in whole mission region including the approaching and contacting phase. Furthermore, it is validated and verified in the MATLAB/Simulink six-DOF simulation and the indoor/outdoor flight test. All systems of the proposed platform are onboard unlike previous studies. Also, the proposed framework provides a Gaussian process-based reference feature path consisting of a set of reference point-like features between the initial and desired features. In order to overcome the vulnerabilities of standard IBVS with respect to large disparities, both the interpolated and extrapolated feature paths are fused to generate relative body velocity commands fed into the under-actuated aerial parallel manipulator. The proposed framework can generate not only stable but also feasible control input. In addition, the proposed framework not only achieves low computational complexity and high real-time performance but also high performance under more difficult initial conditions, unlike general verified environment.