AUT Spherical Robot: Dynamic Estimation and Trajectory Control

Document Type : Original Article


1 Mechanical Engineering Department, Amirkabir University of Technology, Tehran, Iran

2 Institute of Geodesy and Geoinformation University of Bonn, Bonn, Germany

3 Mechanical Engeeniring Department, Amirkabir University of Technology, Tehran, Iran


Spherical robots provide a mean for extensively mobile robot research and studies. Due to their special attributes such as compact structure, omnidirectional and flexible movement, They have been successfully exploited in robotic and control discipline rather than conventional robots which consume much more energy. Even though numerous methods have been adopted to design the Inside Drive Unit (IDU), pendulum structure has been widely employed, given that it might alleviate the complexity either in the process of implementing the mechanical structure and designing the control scheme. The mathematical equations of ball-shaped robots are extremely nonlinear, and they are regarded as highly under-actuated and non-holonomic systems so that there are some relative complexities for dynamic equations to be calculated by the conservative knock on methods. Hence, the main focus of this paper is to experimentally determine the dynamic model of the robot by adopting offline approximation approaches. In this paper, a spherical robot equipped with a pendulum - driven is designed and constructed. The robot is programmed in which it is able to be maneuvered around manually by a remote control device. In the automatic control process, two control schemes are designed to guarantee the trajectory tracking and stability of the robot. The discrepancy between the first and second strategies is that in the former scheme, the dynamics model of the robot based on the Lagrangian method is proposed but in the later scheme, the transform functions between the driving motors and angular velocities are derived experimentally. For both control systems, however, the PID controller provides a mean to ensure the trajectory tracking of the robot. Experimental results depict compared with the control scheme in which the dynamic model is calculated by the Lagrangian method, the trajectory tracking generated by the transform function is more effective and it performs sufficiently more accurate.