Staublı rx160 Manipulatörün Sensörsüz Kuvvet Kontrolü

Abstract

Son yıllarda gelişen teknoloji ile endüstriyel robotların saniyede kullanımı oldukça artmaktadır. Saniyede kullanılan robotlar sayesinde firmalar zamandan, iş gücünden, maliyetten tasarruf ederken, hem de ürettikleri ürünleri daha kaliteli ve hassas üretebilmektedirler. Sanayide kullanılan endüstriyel seri manipülatörlerin çoğu gerek düşük maliyeti gerekse de kolay uygulanabilirliği nedeni ile geleneksel PID(orantısal, integral, türev) kontrol stratejisi ile sabit kazanç katsayıları değiştirilerek kontrol edilmektedir. Fakat bu yöntem robotların doğrusal olmayışı ve eklemlerin bir birine bağlantılı olması nedeniyle hızın arttığı çalışma aralıklarında iyi sonuçlar vermemektedir. Diğer bir taraftan ise sadece konum kontrolü strateji ile endüstriyel robotları kontrol etmek nokta kaynağı, sprey boyama yada hafif yüklerin bir yerden bir yere taşınması sırasında daha çok robot uç noktasının uzayda serbest hareketinin önemli olduğu uygulamalarda başarı sağlamaktadır. Lakin çapak alma, taşlama ve montaj gibi işlemler sırasında robot uç noktasının çevresi ile ciddi etkileşimlere girdiği durumlarda pozisyon kontrolü yerine empedans, hibrid gibi hem konum hem de kuvvet kontrol algoritmalarının beraber uygulandığı kontrolörler önem kazanmaktadır. Bu tezde, İTÜ Mekatronik Eğitim ve Araştırma Merkezinde bulunan altı serbestlik dereceli Staubli RX-160 manipülatörleri kullanılarak robotun geometrik, kinematik ve dinamik modeli elde edilip, konum ve sensörsüz kuvvet kontrolleri gerçekleştirilmektedir. Yapılan tez başlıca 5 bölümden oluşmaktadır. İlk bölümde tez çalışması sırasında kullanılan Staubli RX-160 manipülatörünün hem donanım hem de yazılım kısmı anlatılmaktadır. İkinci bölümde Staubli firmasından alınan ölçü ve parametreler sayesinde Matlab programı kullanılarak, robotun önce ileri geometrik modeli sonra ise ters geometrik modeli elde edilmektedir. Elde edilen ileri geometrik model sayesinde robot eklemlerinin hızları ile uç noktanın hızı arasındaki ilişkiyi veren Jakobiyen matris elde edilmektedir. Daha sonra Staubli firmasından alınan robot uzuvlarına ait kütle ve eylemsizlik matrisleri kullanılarak robotun dinamik modelini açıklayan Euler-Lagrange denklemleri elde edilmektedir. Bu bölümde en son olarak analitik denklemlerden elde edilen robot dinamiğine sürtünme ve yay modelleri eklenerek robotun nihai dinamik modeli kurulmaktadır. Üçüncü bölümde tez sırasında robota uygulanan kontrol algoritmaları elde edilmektedir. Elde edilen ilk kontrol algoritması, görev uzayında uygulanan ters dinamik konum kontrolüdür. Ayrıca kontrol algoritmalarında referans giriş olarak kullanılan yörüngeler de bu bölümde tanımlanmaktadırlar. Robot ile çevresi arasında temas olduğunda, tepki kuvveti oluşmaktadır. Oluşan tepki kuvvet ve moment değerleri genellikle kuvvet-moment sensörleri ile elde edilmektedir. Fakat tez sırasında tepki kuvvet ve momentleri farklı olarak bozucu tork denklemi kullanılarak elde edilmektedir. Son olarak, kuvvet kontrol algoritmaları olan hibrid konum/kuvvet ve empedans kontrol algoritmaları açıklanıp, benzetimleri yapılmaktadır. Dördüncü bölüm, ikinci ve üçüncü bölümlerin sonunda elde edilen model ve kontrol algoritmalarının deneysel olarak doğrulandığı bölümdür. Konum ve sensörsüz kuvvet algoritmalarının deneysel sonuçları bu bölümde ayrı ayrı elde edilmektedir. Son bölüm sonuç bölümü olup, tez sonunda elde edilen gözlemler, yaşanan zorluklar anlatılıp, çeşitli sonuçlar çıkarılmaktadır.

In recent years, using of industrial robots in industry has increased thanks to developing technology. Due to robots used in industry, companies can save time, manpower and cost also they can manufacture their products more quality and precise. However most of serial manipulators used in industry are controlled via traditional PID(proportional ,integral ,derivative) control strategies with changing constant gains because of its low cost and easiness. But this approach doesn’t give good results when the work of robot is fast because of its nonlinear and coupling effects. On the other hand controlling of industrial robots using only position control strategies can have success in application of point welding, spray painting or pick and place, where free motion of end effector of robot is important. But during such an application like deburring, grinding and assembly, where the end effector of robot and its environment are in interaction should be controlled with both position and force controls algorithms like impedance and hybrid control instead of using only position control. In this thesis, using Staubli RX-160 manipulators in ITU Mechatronics Education and Research Center, geometric, kinematics and dynamics model of robot have been obtained and then both position and sensorless force control have been applied. This thesis mainly consists of five chapters. In the first chapter, both hardware and software of used Staubli RX-160 manipulator are explained. In the second chapter, with taken dimensions and parameters from Staubli RX-160, firstly the forward and inverse geometric models of the robot are found in Matlab program and then helping forward geometric model, Jacobian matrix which explains relation between velocities of robot joints and end effector is obtained. Secondly, thanks to mass and inertia matrices taken from Staubli Company, the robot is modelled dynamically in Matlab using Euler-Lagrange equations and Newton Euler equations. Finally adding friction model and spring model to dynamics model, final model of robot has been obtained. Then, to test success of the kinematics and dynamics model of robot, Matlab Simulink has been used. At the end of second chapter. Two dynamics models are compared each other. In the third chapter, control algorithms applied to the robot are obtained. These algorithms are separated into two parts. First of them is motion control in task space. The second one is compliant motion control. To realize motion control in the task space, the trajectory in the task space is generated. Using geometric, kinematics, dynamics models and trajectory in the task space, the computed torque control is applied to the robot. The method of the sensorless force control is explained in the beginning of the compliant motion segment. Using this method, reaction forces and moments on the end effector of robot is found. So we don’t have to use any force-moment sensor in compliant motion of the robot. In the thesis, two different control is used to execute compliant motion. The first control is hybrid motion-force control. Using this control, we manage to control both position of robot and force between the end effector and the environment. The second control method is impedance control. In this control, the end effector of robot is assumed to be a spring with mass and damper. The last segment of the third chapter. Both position control and force control is simulated in Matlab Simulink. Showing different Simulink plots, results of control algorithms applied to robot can be seen. In the fourth chapter, lots of experimental tests are applied to Staubli Rx-160 manipulators. During experiments, the low level program, named LLI, is used to realize position and force control algorithms. LLI is a kind of C libraries. LLI program provides users four feedback signals and for commands at every four miliseconds. Respectively, These feedback signals are joint positions, joint position errors, joint velocities and joint torques. The command signals are joint positions, joint velocities, joint forward torques and joint torques. Also the the feedback signal of force-moment on the end effector can be used at every four milliseconds in the compliant motion control. The first applied control method to the robot is computed torque control in joint space. In the computed torque control, the acceleration of the robot is used as a control signal. To determine this control signal, PID method is used. Changing proportional, derivative and integral gains, the performance of robot is achieved. Because of nonlinearities and coupling effects of joints, the forward dynamic model is used to compensate these effects. These model parts are gravity, Coriolis and centrifugal, spring and friction torques. After achieving position control in joint space, the computed torque control in task space is achieved. In this control, the position and velocity of the end effector is used directly. The acceleration of the end effector is used as a control signal. After, this acceleration is converted into the acceleration of the joints using Jacobian matrices. During the experiments, the robot is controlled in the direction of x, y and z. So the robot is followed to desired position of the end effector coordinates in space. Before the experiments of force control, experimental setup is constructed. A device which includes a spring and ball is attached to end effector of the robot so the robot can’t be damaged by its environment. The firs control algorithm in the force control is hybrid control. In this control, robot can be controlled by using both position and force control. To select which robot axis will be controlled using position control algorithm or force control algorithm, a selection matrix is used. If the corresponding diagonal element of the matrix is 1, the axis is controlled in position control algorithm, otherwise it is controlled in force control algorithm. In hybrid control tests, the motion of robot is controlled by using computed motion controller until it reaches the surface, and then both force between end effector and its environment and motion of robot is controlled using hybrid controller. Also by using hybrid controller, the success of the sensorless force control algorithm is measured. First, the force between the end effector and environment is controlled using ATI force-moment sensor. Second, instead of using sensor, the force is controlled by using sensorless force control algorithm. To compare two different control algorithm, the data provided by LLI is used and plotted in Matlab. The last compliance control algorithm is the impedance control. In this control, the end effector of robot is assumed to be a spring-mass-damper device. When the robot contacts its environment its impedance is changed. The last and fifth chapter. The results of thesis and suggestion about the future works are mentioned in this chapter. Because the works with using Staubli RX-10 robots is related each other. The useful and critical advices is given the next working about these robots. Especially friction model of robot can be improved by identification methods. The use of this thesis, several scenarios like deburring, assembling can be applied the robot. Also at the end of the thesis, lots of Simulink block diagrams used in the thesis and Matlab codes are shared.