Real-Time Implementation of an Online Controller for the Walking of the NAO Robot

Abstract

Practical problems for the implementation of a real-time controller for gaits subject to disturbances on the NAO robot are shown in this paper. Moreover, how to deal with them and carry out real-time practical stable walkings is explained. The robot is modeled with the Linear Inverted Pendulum (LIP) model, which is a simplified model that concentrates the mass of the robot in a single point and constrains the height of the Center of Mass to a straight line. The controller in charge of keeping a stable gait is the Model Predictive Controller (MPC). The MPC is perhaps the most used controller in legged robots due to its optimization over a time horizon and to the possibility of including constraints in the optimization problem. Furthermore, important notes about the implementation of real-time experiments with the NAO robot are discussed in this work. This work gives the first steps to achieve a real-time control using the Essential Model.