International Journal of Engineering Research in Africa ISSN: 1663-4144, Vol. 18, pp 123-129 doi:10.4028/www.scientific.net/JERA.18.123 © 2015 Trans Tech Publications, Switzerland Submitted: 2015-04-09 Revised: 2015-04-10 Accepted: 2015-08-18 Online: 2015-10-07 Experimental verification of force feedback for rehabilitation robot MORE Marcel1,a*, LÍŠKA Ondrej1,b, KOVÁČ Juraj2,c 1 Technical University of Kosice, Faculty of Mechanical Engineering, Department of Automation, Control and Human Machine Interaction, Letná 9, 04200 Košice, Slovak Republic 2 Technical University of Kosice, Faculty of Mechanical Engineering, Department of Robotics, B. Němcovej 32, 04200 Košice, Slovak Republic a b c email@example.com, firstname.lastname@example.org, email@example.com Keywords: force feedback, force measurement, experiment, rehabilitation robot Abstract. Unlike conventional robots, the equipment provided with pneumatic artificial muscles cannot integrate standard systems for force measurement. Applied measurement system involves specific attributes and requirements for pneumatic muscles. Force feedback of rehabilitation device equipped with pneumatic muscles was experimentally verified under the laboratory condition. Introduction Rehabilitation robot driven by pneumatic artificial muscles (pneumatic artificial muscles PAM), developed at the Faculty of Engineering Technical University of Kosice is intended for rehabilitation of the upper limb joints (Fig. 1). It is pluri-articular i.e. it will practice several joints simultaneously. Each of the arms is driven by two pneumatic muscles in antagonistic involvement, controlled by solenoid (electro-magnetic) valves [1-4]. Device shall include a multilevel controlling system. Force feedback is an essential part of an automated device that is intended for active rehabilitation. In the device which is driven by the pneumatic artificial muscles, the force/strength measuring system cannot be used as it is in the conventional robots [5-7]. Therefore it is necessary to design a measurement system considering the specific characteristics and requirements of such equipment [8,9]. Analysis shows that the loading strength can be measured: at the point of contact of the robot with the man, on the robot arm or in drive itself . Three measurement methods were selected for experimental verification. The first one is the measurement in the device handle via touch sensors FSR. Other methods are based on the knowledge that the force applied to the robot arm creates a corresponding reaction in the drive. Load for actuator with pneumatic muscle can be determined from the change in filling pressure in each muscle. Load is also reflected in the mechanical parts of the drive and transmission, where it can be measured using the strain-gauge load cell (sensors). Measuring the force applied on the handle of rehabilitation device Card Humusoft MF624 is used when measuring the force applied on the handle of rehabilitation device and its transfer to the signal. Card is used to obtain data for Matlab-Simulink in which the controller device is designated. FSR Sensors FSR (Force Sensitive Resistor) consists of thin layers of polymer. When the effective surface is loaded by the force, electrical resistance of polymers decreases [10-12]. Their sensitivity is optimized for use of human touch [13-15]. Therefore the sensors are placed in the handle unit. These are located on a flexible backing and covered with a layer of rubber. Electrical connection is based on sensor function. They work as variable resistors. Sensors are connected in series with resistors having a constant resistance, and form a voltage divider. Resistor values are calculated All rights reserved. No part of contents of this paper may be reproduced or transmitted in any form or by any means without the written permission of Trans Tech Publications, www.scientific.net. (#102732250, Chalmers University of Technology, Göteborg, Sweden-27/10/17,14:19:38) 124 International Journal of Engineering Research in Africa Vol. 18 with respect to the measuring range of the sensors. Output divider with the largest margin ensures measurement with the highest resolution. Measuring chain for the FSR sensor represents Figure 2. Fig. 1 Prototype of rehabilitation device driven by PAM Fig. 2 Measurement chain for FSR sensors Pressure sensors Muscle loaded up is changing its length. By changing the length also the volume of muscle varies and in closed system it causes pressure changing. Pressure in muscles determines the force which Journal of Engineering Research in Africa Vol. 18 125 the arm of rehabilitation robot is loaded up. Values of pressure in the muscles determine the force which arm rehabilitation device is loaded up. It allows applying sensors for measuring the pressure in the pneumatic muscles in the range of 0-6 bar which have compatible output. Sensors SSCDANT005PGAA5 were used in the experiment. While pressure regulating in muscle, fluctuation of pressure occurs therefore pressure sensors were placed as close as possible to the muscle [16-19]. Pressure fluctuation in this case is the smallest. Measuring chain for pneumatic pressure sensor is in Figure 3. Fig. 3 Measurement chain for pneumatic pressure sensor Strain-gauge load sensors. Drive load is determined by measuring the voltage on transmission. Experimental device uses for power transmission the wires integrated in bowdens. Sensors were placed between the wire and the pneumatic muscle. In this case there is a load on the axis of the wire. To measure the tensile and compressive stresses were applied strain-gauge sensors EMS30. Sensors do not directly produce signal therefore the measuring system contains the electronic units EMS168 which provides the required power. Signal is amplified and filtered to yield normalized voltage and current output. Measuring chain for load cells is described in Figure 4. Fig. 4 Measurement chain for strain-gauge sensors Experiment Measuring chains were applied on evolved rehabilitation device and their response was tested against the load on arm by strength. Measurements were repeated under the same conditions in all experiments. 126 International Journal of Engineering Research in Africa Vol. 18 Aim of the study was to investigate the dispersion of values obtained by different measurement methods for repeated loading constant force [20-22]. While comparing the results factors that affect the various different forms of measurement were eliminated. The effect of gravity is eliminated by placing the arm device in the way that the centre of the operating range is perpendicular to the ground. Because the device has several degrees of freedom, these while loading would transfer the strength out of the measuring chain [23-25]. Therefore all the joints in which the measurement does not occur were fixed. Subsequently, the pneumatic muscles propelling joint inflate the working pressure which lead to close of pneumatic system . By a simple device that allows producing constant load, the arm was burden alternately on both sides. The reason for the symmetrical loading was that use of pneumatic muscle leads to creeping. This causes a change of the average value according to the number of repeated measurements. For data measuring the program created in Matlab-Simulink was used. Signals from sensors were processed by measuring card MF624. Signal measured by this card was used directly in Simulink. All input signals are filtered by a low pass filter to remove the fluctuations of values. Whereas the measurement was conducted as static, the filter did not affect the process of the measured values. Prepared program stores the value in a file format that handles Microsoft Excel. Record of measured values was processed in Matlab-Simulink. Results of experiments The experiments were repeated in four batches or series respectively. Each contains 30 measurements. In each measurement was recorded load value. This value is in all measurement modes determined as the difference between the values of the two sensors (air pressure difference in muscle, the voltage difference in the cable, etc.). These values cannot be directly compared, since each type of sensor measures with different strength loading or zero offset. Therefore all values are first converted to the same range (0-100) by the formula (1). Fig. 5 Graph of adjusted values measured within one of series Journal of Engineering Research in Africa Vol. 18 127 X max was used as the highest load value in the series of measurements. Xadapted = ( Xmeasured ) × 100 X max (1) Graph of the adjusted values of one set of measurements is shown in Fig. 5. Dispersion of the measured FSR sensors is much greater than for the other methods of measurement. As the number of measurements or time respectively increases, the average value of FSR sensors is primarily deviating. Correlation of the strain-gauge values and values of the pressure sensors shows that the origin of the dispersion is not in the sensor itself, which measure the load on the drive, but may be formed e.g. in the transmission or at the load device. From each series of measurements was calculated experimental variance of the formula (2) where n is the number of measurements in the series, X is the measured value and is the average of the values in the series. Calculated results are in Table 1. 1 n s = ∑ ( X −) n − 1 i =1 2 2 (2) Results of all series measurements show that the most promising method is the measurement of strain-gauges. FSR sensors exhibit high variance. It is not advisable to use them as a single measurement method [27-29]. They can be complementary to another method of measurement (e.g. check the correct grip). Load measuring with pressure sensors (Tab. 1) has a high instability. With dynamic load these deviations measurements even multiply, so the method is not suitable for measuring the load. Measurement of strain-gauge sensors provides the smallest scattering measurements, but even the smallest differences between series of measurements and the positive and negative direction of loading [30,31]. This method is best placed to form force feedback rehabilitation equipment driven by pneumatic artificial muscles. Table 1 Variance of load measurement by different methods Positive deflection Pressure Strain-gauge sensor FSR sensors Negative deflection Pressure Strain-gauge sensor FSR sensors Series 1 19.605 131.307 19.403 27.118 185.416 29.164 Series 2 127.495 101.365 33.551 37.424 184.186 15.311 Series 3 68.785 119.652 30.737 48.126 207.223 27.092 Series 4 100.876 104.683 39.976 33.700 262.262 24.966 Average 79.190 114.252 30.917 36.592 209.772 24.133 Conclusion Force feedback for rehabilitation device driven by pneumatic artificial muscles due to its specific characteristics and requirements must be different from the systems used in other robots. Analysis verified three assessment methods, which were designed for measuring chains. In specific terms, which should eliminate adverse impacts, there were performed and evaluated thirty-four series of repeated load measurements with constant force. 128 International Journal of Engineering Research in Africa Vol. 18 Acknowledgement Contribution was prepared with the support of EU Structural Funds, Operational Programme Research and Development, Measure 2.2 Transfer of knowledge and technology from research and development into practice, the project "Research and development of intelligent unconventional actuators based on artificial muscles" ITMS 26220220103 and also with the support of project VEGA 1/0911/14 "Application of wireless technologies in new products and services for the protection of human health." References  M. More, O. Líška, Design of active feedback for rehabilitation robot, Applied Mechanics and Materials 611 (2014) 529-535.  A. Hošovský, K. Židek, Experimental validation of nominal model characteristics for pneumatic muscle actuator, Applied Mechanics and Materials 460 (2014) 1-12.  B. Jobbágy, J. Karchňák, D. Šimšík, Rehabilitation robotics and wearable sensors as trends of home rehabilitation, in: Proc. of 15th Int. Carpathian Control Conf. ICCC 2014, Velke Karlovice, IEEE, 2014, pp. 219-222.  A. Hošovský: The control if the PAM-based position servosystem with acceleration loop, Journal of applied science in the thermodynamics and fluid mechanics 2/1 (2008).  J. Piteľ, M. Balara, J. Boržíková, Control of the Actuator with Pneumatic Artificial Muscles in Antagonistic Connection, Transactions of VŠB – Technická univerzita Ostrava 53/2 (2007) 101106.  D. Šimšík et al., Design of inertial module for rehabilitation device, in: Proc. of 11th Int. Symp. on Applied Machine Intelligence and Informatics - SAMI 2013, Herľany, Slovakia, IEEE, 2013, pp. 33-36.  J. Piteľ, Modelling of the PAM Based Antagonistic www.cybletter.com/index.php?s=file_download&id=56. Actuator, information on  A. Hošovský et al., Model-based Evolution of a Fast Hybrid Fuzzy Adaptive Controller for a Pneumatic Muscle Actuator, International Journal of Advanced Robotic Systems 9 (2012) 1-11.  O. Líška, M. More, D. Janáčová, H. Charvátová, Design of Rehabilitation Robot Based on Pneumatic Artificial Muscles, in: Proc. of 1st Int. Conf. on Optimization Techniques in Engineering (OTENG '13), WSEAS Press, Istanbul, 2013, pp. 151-154.  A. Hošovský, M. Havran, Dynamic Modeling of One Degree of Freedom Pneumatic MuscleBased Actuator for Industrial Applications, Tehnički Vjesnik 3/19 (2012) 673-681.  K. Židek, O. Líška, V. Maxim, Rehabilitation Device Based on Unconventional Actuator. Mechatronics: Recent Technological and Scientific Advances, Springer, Berlin, 2011, pp. 697-702.  J. Mižáková, J. Piteľ and M. Tóthová, Pneumatic Artificial Muscle as Actuator in Mechatronic System, Applied Mechanics and Materials 460 (2014) 81-90.  J.M. Hopen, A. Hošovský, The Servo Robustification of the Industrial Robot, Annals of DAAAM 2005, DAAAM International, Vienna, 2005, pp. 161-162.  M. Tóthová, Modeling and Simulation of Dynamics of Nonconventional Pneumatic Actuator of Manipulator - Dissertation Work, TUKE, Prešov, 2014.  M. Tóthová, Simulation Model of Pneumatic Actuator Based on Artificial Muscles, in: Proc. of Conference IMAP 2013, TUKE, Prešov, 2013, pp. 65-69. Journal of Engineering Research in Africa Vol. 18 129  M. Tóthová, Simulation of Pneumatic Muscle Actuator Dynamics, in: Proc of 1st Biannual Comparative European Research Conference (CER 2014), London, 2014, pp. 109-112.  M. Tóthová, A. Hošovský, Dynamic Simulation Model of Pneumatic Actuator with Artificial Muscle, in: Proc. of 11th IEEE International Symposium on Applied Machine Intelligence and Informatics (SAMI 2013), IEEE, Budapest, 2013, pp. 47-51.  M. Tóthová, Simulation of the Pneumatic Actuator with Artificial Muscles Using Modified Hill´s Muscle Model, in: Proc. of Automation in Theory and Practice (ARTEP 2013), TUKE, Košice, 2013, 41/1-41/8.  M. Tóthová, Simulation of the Pneumatic Actuator Dynamics Using Advanced Geometric Model of Artificial Muscle, in: Proc. of Automation in Theory and Practice (ARTEP 2014), TUKE, Košice, 2014, 22/1-22/8.  I. Mack et al., Interactive Force Sensing Feedback System for Remote Robotic Laparoscopic Surgery, in: Haptic Audio visual Environments and Games, 2009, information on: http://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=5356129.  T. Kelemenová et al., Model based design and HIL simulations, American Journal of Mechanical Engineering 1 (2013) 276-280.  A. Erwin et al., Interaction Control For Rehabilitation Robotics via a Low-Cost Force Sensing Handle, in: Proc. of the Dynamic Systems and Control Conference DSCC 2013, information on: http://mahilab.rice.edu/sites/mahilab.rice.edu/files/publications/DSCC2013-4073.pdf.  A. Hošovský, K. Židek, Experimental Validation of Nominal Model Characteristics for Pneumatic Muscle Actuator, Applied Mechanics and Materials 460 (2014) 1-12.  M. Tóthová, S. Hrehová, Dynamic Characteristics of the Antagonistic Pneumatic Muscle Actuator for Biomechanical Applications, in: Proc. of 2nd Int. Conf. on Sensors and Materials Manufacturing Science (ICSMMS 2015), BOSI EDU, Paris, 2015.  A. Hošovský, P. Michal, M. Tóthová, O. Biroš, Fuzzy Adaptive Control for Pneumatic Muscle Actuator with Simulated Annealing Tuning, in: Proc. of 12th IEEE Int. Symp. on Applied Machine Intelligence and Informatics (SAMI 2014), IEEE, Budapest, 2014, pp. 205-209.  J. Boržíková, M. Balara, Mathematical Model of Contraction Characteristics of the Artificial Muscle, Manufacturing Engineering 6/2 (2007) 26-29.  M. Balara, M. Tóthová, Static and Dynamic Properties of the Pneumatic Actuator with Artificial Muscles, in: Proc. of IEEE 10th Jubilee Int. Symp. on Intelligent Systems and Informatics (SISY 2012), IEEE, Subotica, 2012, pp. 577-581.  K. Židek, A. Hošovský: Wireless Device Based on MEMS Sensors and Bluetooth Low Energy (LESmart) Technology for Diagnostics of Mechatronic Systems, Applied Mechanics and Materials 460 (2014) 13-21.  P. Berkelman et al., A Miniature Microsurgical Instrument Tip Force Sensor for Enhanced Force Feedback During Robot-Assisted Manipulation, in: IEEE Transactions On Robotics And Automation, 2003, information on: http://www.cs.jhu.edu/~rht/RHT%20Papers/2003/Miniature.pdf  A. Erwin et al., Interaction Control For Rehabilitation Robotics via a Low-Cost Force Sensing Handle, in: Proc. of the Dynamic Systems and Control Conference DSCC 2013, information on: http://mahilab.rice.edu/sites/mahilab.rice.edu/files/publications/DSCC2013-4073.pdf.  F. Trebuňa, P. Sivák, Experimental methods of thermomechanics: Tensometry (Experimentálne metódy mechaniky: Tenzometria), TUKE, Košice, 2012 (in Slovak).