An Approach to Determining Internal Forces of a Walking Mechanism Using Multilayer Perceptron Neural Network

In order to determine the maximum internal forces acting on certain joint of a walking mechanism for the structural strength analysis, a mechanical model for the chassis including all position freedom parameters of the mechanism in operation is established, and an optimization model for determining the internal force of the walking mechanism is presented as well. A method using the bi-level optimization model based on multilayer perceptron (MLP) for solving the optimization problem is proposed. In the first level, the maximum forces acting on certain joint of the chassis are optimized by the sequential quadratic programming method at a set of positional parameters. The MLP, as a nonlinear mapping relationship between the maximum force state at certain joint and the positional parameters, is trained by using orthogonal design. In the second level, the genetic algorithm is used to optimize the MLP. A high precision MLP is obtained through an iterative procedure, in which the searching volume is reduced by using the bisection method. The results indicate that the proposed method is effective in solving the multivariable optimization of the complex structural system. It is also concluded that the method provides a theoretical basis for designing the walking mechanism.