Active Force Control with Nonlinear Predictive Control based on Receding-Horizon Cost Function Optimization for Five-Link Biped Model in Sagittal Plane NURFARAHIN ONN, MOHAMED HUSSEIN, COLLIN HOWE HING TANG, MOHD ZARHAMDY MD ZAIN, MAZIAH MOHAMAD and LAI WEI YING Faculty of Mechanical Engineering Universiti Teknologi Malaysia 81310 Skudai, Johor MALAYSIA nurfarahin5@live.utm.my, mohamed@fkm.utm.my, tanghh@fkm.utm.my, zarhamdy@fkm.utm.my, maziah@fkm.utm.my, wylai2@live.utm.my Abstract: - This paper discusses about the incorporation of Nonlinear Predictive Controller (NPC) with Active Force Control (AFC) that applied to a five-link biped robot model has been investigated and simulated. The NPC control law of the system is based on prediction model which is obtained via Taylor series expansion. From the receding horizon cost function minimization, the optimal control is computed directly. AFC gives an excellent trajectory performance and robust against various types of known/unknown uncertainties and able to cancel the disturbances’ effects on the biped system. The five-link bipedal walking model consists an upper body and two legs (thigh and shank in every leg) which constraint within the sagittal plane. The effectiveness of the proposed technique is studied in this case. It is found that this technique gives robust and stable to the control system although it is under disturbances’ influence. Key-Words: - human biped model, Nonlinear Predictive Control, Active Force Control 1 Introduction Robot manipulator is a nonlinear, high dynamic and uncertain system. Therefore, the controller of the robot system has to be able to deal with system uncertainties such as unmodeled dynamics, variable payloads, friction torques, torque disturbances, parameter variations, and measurement noises which are exiting in the practical environments [1]. In few decades ago, many techniques have been proposed to achieve robust biped robot system controller [2][3][4][5][6][7][8]. Research works on nonlinear predictive control applied to robot manipulator are continuous growing actively since a few years ago. Predictive Computed-Torque Control was implemented by V. Becerra et al. [9] using six Degrees of Freedom (DOF) PUMA 560 Manipulator Robot. R. Hedjar and P. Boucher proposed the nonlinear receding- horizon controller of rigid link robot manipulator approximation using output feedback via link position measurements to improve the trajectory tracking problem [10]. In [11], K. Bdirina et al. applied nonlinear predictive control to compute time optima solutions for a two link manipulator to control angle positions. A. Marabet and J. Gu constructed a robust tracking-control scheme based on state estimation for rigid -link robot manipulator [1]. A. Moreno and R. Mallgui applied a embedded robust nonlinear predictive control system [12] based on the algorithm developed in [1] that can control the angular position of the two DOF robot manipulator’s base and arm concurrently. Active Force Control (AFC) was introduced by Hewit and Burden [13] is a simple, stable, robust and effective compared to other conventional approaches in controlling dynamic system [14]. AFC gives an excellent trajectory performance and robust against various types of known/unknown external disturbances, uncertainties and varied operating conditions. AFC has been applied to many types of dynamic systems such as robot arm [15][16][17], mobile manipulator robot [18][19], brake system [20], human biped model [4][7], cutting tool [21] and worm-like micro robot [22]. In this study, a Nonlinear Predictive Control (NPC) incorporated with an Active Force Control (AFC) scheme is applied to a five-link biped robot model that has been studied and simulated. The algorithm based from predictive model [1] is carried out using Taylor series expansion to tracking control Manufacturing Engineering, Automatic Control and Robotics ISBN: 978-960-474-371-1 156