IFAC PapersOnLine 50-1 (2017) 7592–7597 ScienceDirect Available online at www.sciencedirect.com 2405-8963 © 2017, IFAC (International Federation of Automatic Control) Hosting by Elsevier Ltd. All rights reserved. Peer review under responsibility of International Federation of Automatic Control. 10.1016/j.ifacol.2017.08.1003 © 2017, IFAC (International Federation of Automatic Control) Hosting by Elsevier Ltd. All rights reserved. Keywords: Tumour mimic model, cascaded control, iterative learning control, pneumatics, friction compensation. 1. INTRODUCTION In the framework of medical research, suitable mechanisms with mimic models are often required. They are either used to validate experimental set ups in radiology or computed tomog- raphy (CT) imaging. In many cases, these specially designed mechanisms or small-scale robots are actuated pneumatically. In comparison to regular electro-mechanical or hydraulic so- lutions, pneumatic drives benefit from low investment costs, a clean working medium and steady-state actuation without thermal problems. From a control point of view, however, a characteristic drawback of pneumatically actuated systems is given by the air compressibility, which has to be taken into account properly at control design. In addition, the nonlinear kinematics of the mechanical mechanism or the small-scale robot and nonlinear friction have to be considered. Especially the friction in the pneumatic actuators ends up becoming a significant barrier to precise motion. In Fischer et al. (2008), a pneumatically actuated robotic as- sistant is presented, used for magnetic resonance imaging. The robot is capable of placing a needle to the patient with high accuracy. Another pneumatically driven robot is described in Kapoor et al. (2009). Here, the robot has to fulfil medi- cal tasks with a hands-on cooperative control. A backstepping control approach is presented for a pneumatic needle-guiding robot in Franco et al. (2015). A pneumatic needle driver is also discussed in Comber et al. (2016). The applied hybrid control concept meets given performance criteria for positions as well as speeds in both translational and rotary directions. Fig. 1. Test rig in front of a Philips PET/CT scanner at the University Medical Center of Rostock. A possible application is depicted in Fig. 1. Here, the tumour mimic robot is used inside a PET/CT scanner from Philips to simulate the breathing-induced motion of a human lung tumour. A clinically validated mathematical model for such a motion is presented in Seppenwoolde et al. (2002) and Prabel et al. (2016). This paper is structured as follows: First, a control-oriented model of the mechatronic system including identified valve * Chair of Mechatronics, University of Rostock, D-18059 Rostock, Germany, (e-mail:{Alexander.Wache, Harald.Aschemann, Robert.Prabel}@uni-rostock.de). ** Department of Nuclear Medicine, University Medical Center of Rostock, D-18057 Rostock, Germany, (e-mail: {Jens.Kurth, Bernd.Krause}@uni-rostock.de). *** Chair of Mechanical Engineering Design/CAD, University of Rostock, D-18059 Rostock, Germany, (e-mail: Stefan.Zorn@uni-rostock.de). Abstract: In this paper, a model-based tracking control is proposed for a mechanism dedicated to accurately reproduce the breathing-induced motion of a human lung tumour. A lung tumour mimic model should perform the same smooth motion as a real one in a human body during inhalation and exhalation. In former work, a 3-dimensional mechanism equipped with three pneumatically driven axes has been developed and built up for this purpose. The discrete-time control structure consists of cascaded control loops: In the fast inner loops, the chambers pressures of the pneumatic cylinders are controlled, whereas the outer loops are related to the position control of the cylinders. Furthermore, iterative learning controllers are employed to compensate model uncertainties and lumped disturbance forces. The proposed overall control structure has been implemented and successfully validated on the innovative test rig. Alexander Wache * Harald Aschemann * Robert Prabel * Jens Kurth ** Bernd J. Krause ** Stefan Zorn *** Iterative Learning Control of a Pneumatically Actuated Lung Tumour Mimic Model