论文部分内容阅读
Precise welding of the T-joints between aircraft hyperbolic panels and stringers is required. Therefore, a method of solving in-verse kinematics equations for a cooperative welding robot with multiple manipulators based on neural networks was investigated. To build an effective Denavit-Hartenberg(DH)model for this robot, sample data was obtained considering the movement ranges of the robot joints. Based on back propagation (BP) and radial basis function (RBF) neural networks, 18 joint sub-spaces were mapped to the workspaces of three manipulators. The high-dimensional and nonlinear inverse kinematics problem was transformed into a multi-input and multi-output prediction model. The results revealed that the prediction model of solving the cooperative welding robot kinematics equations was quite ac-curate. Moreover, compared with the BP-based model, the calculation process of the RBF-based prediction model was slower, but yielded more accurate predictions.