||How to transfer dexterously operational skills of humans to robots has been a key issue to establish human-like intelligence of robots to deal with uncertain and variable system parameters more effectively. Therefore, typical motion behaviors such as walking, taking the stairs, jumping, performing gymnastics, and even mimicking human emotion are always chosen for research targets. This thesis aims at realization of high-bar gymnastics through observation, analysis, design of fuzzy controllers, simulation and experiments.|
Based on observation on motion behavior of high-bar gymnastics, four different motion stages including swag, swing-up, balance, and swing-down can be concluded. Analysis using the work-energy theorem and mechanics principles fully resolves the purpose of body posture in those stages. Four independent fuzzy controllers are designed for those four motion stages. Besides, a switching mechanism is employed to determine which fuzzy controller should be selected.
In order to implement the high-bar gymnastics, a two-link robot is applied for investigation and demonstration. The control goal is that the robot needs to successfully complete a cycle of those four stages and the total number of swing in the stages of swag and swing-up before balance has to be less than twenty. Furthermore, uncertainties in the link mass and the link length are also considered in the high-bar gymnastics experiments. Experimental results consistently demonstrate satisfactory robustness and control performance. The proposed control strategies successfully exhibit human-like intelligence on a two-link robot conducting high-bar gymnastics.