Bulanık adaptif kayan kipli robot kontrolu

Abstract

Bulanık Adaptif Kayan Kipli Robot Kontrolü konusundaki bu tezde, Tübitak MAM Robotik Bölümünde tasarlanan ve imal edilen endüstriyel tipteki MAMROB ER- 15 Robotunun dinamik modeli gözönüne alınarak, dayanıklı, integral etkili dayanıklı, dayanıklı uyarlamalı, kestirilmiş eşdeğer kontrollü çatırtısız kayan kipli ve bulanık adaptif kayan kipli kontrol yöntemleri geliştirilerek robot kontrolü yapılmıştır. Birinci aşamada, MAMROB robotunun L-E ve parametre doğrusallaştırılmış dinamik modelleri elde edilmiştir. İkinci aşamada, dayanıklı, integral etkili dayanıklı ve dayanıklı uyarlamalı kontrol yöntemleri geliştirilerek MAMROB robotuna uygulanmıştır, integral etkili dayanıklı kontrol yöntemi olarak geliştirilen bu yeni yöntem, kontrol kuralındaki dayanıklı doyma ve integral etkilerinin olumlu tarafları birleştirilerek elde edilmiştir. Bu yöntemlere ait simülasyon sonuçlan verilmiştir. Üçüncü aşamada, SABANOVIC 'in kestirilmiş eşdeğer kontrollü çatırtısız kayan kipli kontrol yöntemleri geliştirilerek MAMROB robotuna uygulanmış ve deneysel çalışmalara ait sonuçlar verilmiştir. Son aşamada, Sabanovic'in Kayan Kipli Kontrol yönteminin Adaptif Kayan Kipli Kontrol Yöntemi olabilmesi için, bulanık uyarlama mekanizması geliştirilerek, bulanık adaptif kayan kipli kontrol yöntemi, tezde bir yenilik olarak sunulmuştur. Literatürde teklif edilen bulanık kayan kipli kontrol yöntemlerinden farklı olarak ele alman uyarlama mekanizması 7.Bölümde ayrıntıları ile birlikte sunulmuştur.

In this thesis, some certain control techniques are developed and applied to a new six degree of freedom ( DOF ) robot manipulator (MAMROB) which is made by Tubitak MAM, CAD/CAM Robotics Department. First of all, we derived the dynamic model of the MAMROB as a new industrial robot. In order to obtain the dynamic model of robot, we prefer the Lagrange-Euler method. We used the true physical parameter values in our model. In addition to classical control techniques, we proposed a novel robust control approach based on parameter linearization method. In this control, we proposed a globally asymptotically stable robust control scheme by combining integral control with a robust saturation control law. This method takes advantage of both saturation control and integral control techniques. During the derivation of these control scheme, we realized the uncertainty bounds that are required in the synthesis of robust control laws which are difficult to estimate precisely because they depend not only on various physical parameters of robot, but also on the commanded trajectory and the manipulator motion states. In this thesis, we also used a new chattering free sliding mode control technique. In classical sliding mode control the most important problem is the chattering which is high frequency oscillations of the control signal. A chattering signal may cause fatigue in harmonic drives in robot arms. In our method, chattering is avoided by selecting a different Lyapunov function. This novel approach can be used directly, if the controller parameters set property. ' Due to nonproper selection of control parameters and load variations, chattering may arise and cause some problems even when this algorithm is used. In order to overcome this problem we proposed a novel adaptation technique based on fuzzy logic. Fuzzy adaptive sliding mode control has both robust control properties and adaptive capability. IX In the firs chapter, the general procedures are introduced and some technical information is given about MAMROB. The general kinematics and dynamics derivation methods for robot manipulators are introduced in the second and third chapters. In the fourth chapter, the dynamics and kinematics equations of MAMROB are derived for its real time controls and their simulation studies. Five different control approaches are simulated in the fifth chapter to test their control performances. These control approaches are PID, MRAC, Robust Control, Modified Robust Control and Robust Adaptive Control. L-E dynamic model of robot is as follows.... M(q)q+ C(q,q)q+ g(q) = x (1) Parameter linearized dynamic model of robot is as follows M(q)q+C(q,q)q+g(q) = Y(q,q,q)p = r (2) where, Y (nxm) is a known matrix called regression matrix of robot, p (mxl) is an unknown parameter vector. Let p0 e9?m and /?e$R+ respectively be nominal parameter vector and a positive number that defines a bound of parametric uncertainty. The parameter error and its norm can be defined as given below; \\p-Po\\*P 0) Let to be nominal torque vector corresponding to nominal parameter vector p0 and it can be defined as r0 = M0 (q)a + C0 (q, q)v + g0 (q) - Kr (3) *0=Y(q,q,v,a)p0-Kr (4) In Eq. (4), K is a positive definite diagonal matrix and vector valued functions v, a, r are defined as a = v = <ld-k6 -YTr if,\YTr\</ld-k