The Positioning of Systems Powered by McKibben Type Muscles Wiktor Parandyk, Michał Ludwicki, Bartłomiej Zagrodny, and Jan Awrejcewicz Lodz University of Technology, Lodz, Poland Department of Automation, Biomechanics and Mechatronics parandyk.wiktor@gmail.com Abstract. In this paper a continuous control of the mechanical system positioning, powered by a pneumatic actuators (McKibben type muscles) is presented. The control system consists of appropriate sensors which allows to monitor the values of the characteristic parameters, i.e. displacement and pressure. Moreover, throttle valve controlled by stepper motor is used as regulated elements. Measured signals (displacement of the actuator and the load, calculated indirectly) provide feedback loop to the control system which operate the throttle valves. Proposed system, build of one valve (actuated by stepper motor), McKibben muscle, air compressor and electronic compartments allows for continuous control of the air flow, variable speed of shortening or stretching of artificial muscle and its smooth stop at the desired (set) position. Data acquisition system, used for measuring the characteristic parameters and for valve operation support is realized by an universal measurement and control multimodule in addition to the LabVIEW software package. Keywords: muscle McKibben type, positioning, continuous control. 1 Introduction Presented experimental set-up is dedicated to test the possibility of positioning pneumatic McKibben type actuators using throttle valves and position feedback. In this case we propose indirect control of the muscle position which is a resultant value of the valve opening level. In the other words the control is conducted on the compressed air flow ratio at the actuator inlet. The FPA (flexible pneumatic actuators [1, 2]) group was developed to face the biocompatibility problem. First of all a flexible structure of each actuator located in the FPA group allows for a various form of application. In contrast to the classic linear, pneumatic cylinder in rigid housing there are no limitations due to the mounting space. Focusing on the McKibben type artificial muscle, it is convenient for it to work in horizontal/vertical orientation or to work bent even 90 degrees. The universal configuration character and flexibility cause that described actuators are finding increasing use as an alternative drive solution for robotics (see [3, 4]), automatic control systems and industry. Springer International Publishing Switzerland 2015 J. Awrejcewicz et al. (eds.), Mechatronics: Ideas for Industrial Applications, Advances in Intelligent Systems and Computing 317, DOI: 10.1007/978-3-319-10990-9_13 133
134 W. Parandyk et al. The operating principle of the pneumatic McKibben type drive, presented in the Fig. 1, is based on the fusion of elastic properties of the inner rubber-type core with the longitudinal stiffness of the polyester cross-braid outermost layer. The increasing air pressure causes stretching of the rubber core, as a result braid fibers move relatively to each other allowing the radial displacement and muscle shortening. Fig. 1. Pneumatic McKibben type actuator: 1 polyester cross-braid, 2 rubber-type core Because of a strong nonlinear character of artificial muscles (see for example [5]) connected with a rubber-type material properties and the friction occurrence between both the core and the braid, the displacement function of the actuator is dependent on two variables the pressure and the load. Due to mentioned difficulties and necessity of using various types of sensors, the most convenient way of the control is a proportional regulator (see the work [6] and [7]) with the displacement feedback usage (the control system scheme is presented on the Fig. 2). 2 Experimental Set-Up The experimental set-up (Fig. 2 and Fig. 3) consists of the McKibben type actuator connected to the linear bearing guided, load handle, mounted on the rigid frame. The control hardware is dedicated to cooperate with National Instruments DAQ module with a LabVIEW software. The indirect control of the actuator position is carried by the air flow ratio driven by a throttle valve coupled with a stepper motor (Fig. 4). Namely, the control is conducted on the compressed air flow ratio at the actuator inlet. The position feedback signal is measured by a linear displacement sensor connected to an analog input of the DAQ module. The regulator structure was developed as a part of the LabVIEW control program.
The Positioning of Systems Powered by McKibben Type Muscles 135 Fig. 2. The control system schematic structure Fig. 3. The experimental set-up structure: a) unpowered actuator, b) fully contracted, loaded actuator
136 W. Parandyk et al. Fig. 4. Air flow control components: 1 stepper motor, 2 throttle valve, 3 sliding clutch 3 Measurement and Control Software In this project, the measurement of actual pneumatic actuator length and control of the valves stepper motor are performed by National Instruments USB-6009 multifunction IO module, with all necessary AD and DA converters and digital port connected to a dedicated stepper motor controller (Fig. 5). The control algorithm as well as the communication with the IO module are developed in National Instruments LabVIEW 2012 Developement System. It is a good, easy to use and complete solution for this type of experiments. Fig. 5. The experimental set-up components: a) USB-6009 multifunction IO module, b) stepper motor controller
The Positioning of Systems Powered by McKibben Type Muscles 137 In the Fig. 6, complete LabVIEW control algorithm is showed. Length sensor of the actuator gives proportional, linear but noisy voltage signal (big blue analog input block on the left). This is a typical problem in AD signal acquisition. That is why standard Butterworth filter was used, to smoothen measured position value. It was necessary to make it valuable for the PID control algorithm. PID block calculates the error between real actuators length and the set one. Obtained correction value is rescaled to the stepper motor position, expressed in number of steps. The condition block on the right, sets proper direction of the motor, depending of the error sign and starts the motor if PID-calculated number of steps are far (with some tolerance) from the actual ones. As mentioned before, in this work, only proportional P control was used. Fig. 6. Main measurement and control algorithm in LabVIEW The stepper motor, which rotates the valve, is also controlled by the IO module and proper LabVIEW loop. Fig. 7 depicts stepper motor control loop, with ½ step division feature. Consecutive combinations of motor coils voltage supply are sent to the IO module throw the digital byte output. Fig. 7. Stepper motor control algorithm
138 W. Parandyk et al. In Fig. 8, graphical user interface (GUI) of developed software is showed. User can chose the value of constant speed of the motor, change each PID parameters and set desired actuators length set point. The plot presents actual actuators length, set point position and motor steps number, all vs. time. Among others, the control value of each PID parts is showed, the error, actual actuators length etc. The clock in bottomleft corner of the window shows actual motor angle of rotation, expressed in motor steps, proportional to the level of the valve opening. Fig. 8. LabVIEW measurement and control interface 4 Experimental Results Each test of the experiment consisted of closing the throttle valve, homing the actuator (by lowering the pressure inside), setting the set point to half length of the actuator and starting control program. After reaching final set point with an error of minimum 0.0001 values, measurement data were saved. This procedure was repeated for different PID proportional parameter (0.5, 1.0 and 1.5) and different load (no added load, 2.5 kg, and 5.0 kg) attached to the actuators length, 5 times each. Performed experiments showed good repeatability of the positioning sequence. Below in Figures 9 11, position control Y and its error vs. time is showed. Position of the muscle is normalized, for easier comparison. Fig. 9 shows, that if there is no load, reaching set position (marked with black circle) is fastest for the high P value (more than 1.0) while P equals 0.5 and 1.0 gives similar times but longer than for lower P. In Fig. 10, it can be noticed that increasing the load mass slows the position reaching. It also appears, that the best proportional coefficient is for P equals one, while other settings makes the positioning slower.
The Positioning of Systems Powered by McKibben Type Muscles 139 Fig. 9. Controlled position and control error vs. time for actuator with no added load Fig. 10. Controlled position and control error vs. time for actuator with 2.5 kg load Fig. 11. Controlled position and control error vs. time for actuator with 5.0 kg load
140 W. Parandyk et al. Adding next 2.5 kg to the load makes the setting time longer but only for P value lower than 1.5. Bigger P (equals 1.5) makes the positioning much faster than for lower loads. 5 Summary In this project, attempt of PID control algorithm of pneumatic McKibben type actuators contraction was developed. It is an extension of some earlier work [8], where the manual on-off control was used. As proved in previous sections, it is possible to automatically manipulate the air valve opening level, only measuring the actual length of the actuator as the feedback signal. This solution is probably not optimal one but valuable, especially for slow muscle movements, in one direction only, with no position overshooting. As for the future, it is important to determine proper, optimized PID parameters, probably in function of load value. There is also necessity to add second throttle valve, to lower the air pressure after overshooting desired position. That would make a possibility of moving the pneumatic actuator in both directions and increase the positioning quality. Acknowledgements. Project was supported by the participation of the students scientific mechatronics group of Departure of Mechanical Engineering, Lodz University of Technology. Great gratitude for the assistance during the experimental set-up construction. References 1. Dindorf, R.: The modeling of the pneumatic, artificial muscle systems, Department of Mechatronics, pp. 147 156. Kielce University of Technology, Kielce (2005) (in Polish, Modelowanie sztucznych układów mi niowych aktuatorami pneumatycznymi) 2. Daerden, F., Lefeber, D.: Pneumatic Artificial Muscles: actuators for robotics and automation, Vrije Universiteit Brussel, Department of Mechanical Engineering, Brussels 3. Kawashima, K., Sasaki, T., Ohkubo, A., Miyata, T., Kagawa, T.: Application of Robot Arm Using Fiber Knitted Type Pneumatic Artificial Rubber Muscles, pp. 4937 4942. Tokyo Institute of Technology, Yokohama (2004), doi:10.1109/robot.2004.1302500 4. Tondu, B., Ippolito, S., Guiochet, J.: A Seven-degrees-of-freedom Robot-arm Driven by Pneumatic Artificial Muscles for Humanoid Robots, pp. 257 274. Institut National de Sciences Appliquées, Touluse (2005), doi:10.1177/0278364905052437 5. Chou, C.-P., Hannaford, B.: Static and Dynamic Characteristics of McKibben Pneumatic Artificial Muscles, Department of Electrical Engineering, pp. 281 284. University of Washington, Seattle (1994), doi:10.1109/robot.1994.350977 6. Li, Y., Heong Ang, K., Chong, G.C.Y.: PID control system analysis and design. IEEE Control Systems 26(1), 32 41 (2006), doi:10.1109/mcs.2006.1580152 7. Ko, B.-S., Edgar, T.F.: PID control performance assessment: The single loop case. AIChE Journal 50(6), 1211 1218 (2004) 8. Parandyk, W., Zagrodny, B., Awrejcewicz, J.: Selected problems of biocompatibility of the pneumatically controlled arm. Pomiary Automatyka Robotyka 17(1), 71 75 (2013)