Whilst the necessity of finding an intelligent-based controlling method for a two-leg walking robot increases, the balance of the under-actuated leg consisting of two links is emphasized in this study. This is not only a nonlinear structure, but also a single-input double-output system. However, the problem becomes concrete through the proposed diagonal recurrent neural networks (DRNN) method. In this paper, two kinds of DRNN are introduced into the control system. The diagonal recurrent

more »

... entifier (DRNI) is selected as an identifier, and the diagonal recurrent neurocontroller (DRNC) is determined as a controller. Additionally, a generalized dynamic backpropagation algorithm (DBP) is also applied to train both DRNC and DRNI. With the simulated results, it is shown that the under-actuated leg is balanced and stabilized by DRNN. This study definitely contributes the intelligent-based as well as the real-time controlled method for a two-leg walking robot with profound insight Keywords: recurrent neural networks; walking robot; dynamic backpropagation algorithm The interest of a two-leg (two-arm) robot coupled to a human and used to augment this strength is arising [1] . The walking machines with the actuation at the knee but without the actuation at the ankle surely avoid the large feet's problem [2, 3] . With the reason above, this paper contemplates to control and stabilize the under-actuated leg, which is a both nonlinear and single-input double output system. The general control method and PI or PID control are far from giving the satisfactory solution. The computed toque method is one of the most intuitive schemes for robot control. Nevertheless, this method is only applicable for SISO or MIMO system. In order to control the under-actuated leg, an intelligent-based solution is requested. Therefore, the diagonal recurrent neural network (DRNN) [3] is introduced to control the walking robot in this study. Ku [4] has indicated that the traditional neural networks, feedforward neural network (FNN), is a static mapping without the aid of tapped delays and is unable to represent a dynamic system mapping. Although many researches have utilized the FNN with tapped delays to handle the dynamic problems, the FNN requires a larger number of neurons to describe the dynamical responses in the time domain. The recurrent neural networks [4, 5, 6] have numerous important capabilities in feedforward network, and it is a much appropriate dynamic mapping than the FNN when applied to dynamic systems. In this paper, two DRNI are used to provide the sensitivity information of the links to the DRNC, and two DRNC are used to control the under-actuated leg. Between the controller and plant, a regulator to determine the input of the plant is enhanced.