The generalized regression neural network-one kind of RBF neural network, is chosen to construct the inverse-kinematics model for the shotcrete robot which has redundant degree-of-freedom. The inverse-kinematics model...The generalized regression neural network-one kind of RBF neural network, is chosen to construct the inverse-kinematics model for the shotcrete robot which has redundant degree-of-freedom. The inverse-kinematics model of the object is trained by the general learning method. In constructing model process, different partition methods is tried to divide the joint space and different diffusion coefficient value to train the neural network. The influence of the spread coefficient to the approach ability is also studied. The simulation method is adopted to test the performance of the neural network. The simulation result turns out to be satisfactory.展开更多
The design and accomplishment of shotcrete robot is presented in this paper. This shotcrete robot is the first robot for underground coal mine in our country. It is a specific machine used for spraying concrete in und...The design and accomplishment of shotcrete robot is presented in this paper. This shotcrete robot is the first robot for underground coal mine in our country. It is a specific machine used for spraying concrete in underground coal mine. It has many advantages such as simple structure convenient operation and high reliability. The on-site experiments indicate that the sprayed layer is even, at the same time, the quantity of rebound and dust content decrease largely. The robot can satisfy the requirement of shotcrete process completely.展开更多
基金Support by Education Innovation fund of Shandong Education Department(SDYY06052) Support by Special Fund of Shandong Science and Technology Department(2006GG1108097-23)
文摘The generalized regression neural network-one kind of RBF neural network, is chosen to construct the inverse-kinematics model for the shotcrete robot which has redundant degree-of-freedom. The inverse-kinematics model of the object is trained by the general learning method. In constructing model process, different partition methods is tried to divide the joint space and different diffusion coefficient value to train the neural network. The influence of the spread coefficient to the approach ability is also studied. The simulation method is adopted to test the performance of the neural network. The simulation result turns out to be satisfactory.
文摘The design and accomplishment of shotcrete robot is presented in this paper. This shotcrete robot is the first robot for underground coal mine in our country. It is a specific machine used for spraying concrete in underground coal mine. It has many advantages such as simple structure convenient operation and high reliability. The on-site experiments indicate that the sprayed layer is even, at the same time, the quantity of rebound and dust content decrease largely. The robot can satisfy the requirement of shotcrete process completely.
