Complexity

Volume 2018, Article ID 5468090, 8 pages

https://doi.org/10.1155/2018/5468090

## Fuzzy Adaptation Algorithms’ Control for Robot Manipulators with Uncertainty Modelling Errors

^{1}The State Key Laboratory for Manufacturing Systems Engineering and Systems Engineering Institute, Xi’an Jiaotong University, Xi’an 710049, China^{2}The School of Automation, Xi’an University of Posts and Telecommunications, Xi’an, China

Correspondence should be addressed to Yongqing Fan; moc.361@f-qgnoy

Received 19 February 2018; Accepted 27 March 2018; Published 10 May 2018

Academic Editor: Jing Na

Copyright © 2018 Yongqing Fan et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

#### Abstract

A novel fuzzy control scheme with adaptation algorithms is developed for robot manipulators’ system. At the beginning, one adjustable parameter is introduced in the fuzzy logic system, the robot manipulators system with uncertain nonlinear terms as the master device and a reference model dynamic system as the slave robot system. To overcome the limitations such as online learning computation burden and logic structure in conventional fuzzy logic systems, a parameter should be used in fuzzy logic system, which composes fuzzy logic system with updated parameter laws, and can be formed for a new fashioned adaptation algorithms controller. The error closed-loop dynamical system can be stabilized based on Lyapunov analysis, the number of online learning computation burdens can be reduced greatly, and the different kinds of fuzzy logic systems with fuzzy rules or without any fuzzy rules are also suited. Finally, effectiveness of the proposed approach has been shown in simulation example.

#### 1. Introduction

Technologies of learning and adaptation in complex dynamical systems have been become an indispensable part of modern high-tech and attracted the attention in different range of engineering fields. More recently, much effort has been made to handle applications in medical researches [1], virtual reality [2], teleoperation and entertainment [3], and so on. In robotic systems, many technologies about dynamical complex systems can be described as learning human-like skills, such as making the states following a given reference model in [4]; the robot slave arm is employed to drive the adaptation algorithms according to human tutors behavior [5]. Usually, the motion of a master device is controlled by human operator controls, and the master device generates commanding trajectories that are passed to slave systems in order to execute the task by the communication channels [6]. Many kinds of learning and adaptation algorithms control strategies for teleoperated robot systems have been reported in order to improve the controllability and reliability in [7, 8]. Based on wave variables and corrections, a novel control scheme is established to deal with the stability of the closed-loop systems in [9]. From the viewpoint of mathematics, ordinary differential equations (ODE) can be represented between the master and slave of robot systems, so the synchronization between them is closely similar to observer problem [10, 11], which means that some mathematical methods in control theory may be employed to solve the problem of synchronization for master-slave robot systems. A novel shared control system with task motion and self-motion is to achieve accurate object manipulation for Baxter robot manipulator by users mind in [12]. In order to ensure robustness and realization capability of the controllers, disturbance observer is designed by using linear matrix inequality (LMI) in [13].

Generally speaking, it is difficult to research on controlling system with unknown parameters or uncertainties models [14, 15], and hence how to control complexity systems such as robot manipulator systems with uncertain terms becomes a topic deserving of study. Adaptive learning algorithm of parameter estimation is utilized to control robot arm systems with unknown parameters [16]. In [17], an extended state observer is designed to estimate the unknown friction and disturbances in systems such that tracking performance is achieved. For the problem of unknown model of vehicle engine torque, unknown input observer and adaptive parameters estimation are proposed in [18]. Many other results have shown that neural network (NN) [19–23] and fuzzy logic systems (FLSs) are universal approximators [24–26] with online learning capability to emulate complicated nonlinearity and uncertain functions. For an uncertain three-link flexible-joint electrically driven manipulator with complex nonlinear functions, an adaptation algorithms observer combining with NNs is used to estimate the uncertain information such as the link and actuator velocity in systems in [27]; a link position tracking controller with estimated states is proposed. A novel control is investigated in [28], incorporated with NNs and adaptation algorithms technology; the output feedback controller with an observer is built to stabilize the closed-loop systems, and the NN is to satisfy or improve some performance of robot dynamical systems. In order to extend the semiglobal stability to global stability by using NNs controller, a switching mechanism in [29] is formatted for the control task of bimanual robots systems. An augmented NN combining with prescribed performance function is utilized to compensate for unknown active vehicle suspension systems in [30], which can ensure that the transient and steady-state suspension response is guaranteed. Because the extreme learning machine (ELM) has many advantages compared to the traditional neural network, such as convenience for implementation, the fast training speed, and the least manual intervention, [31] brought a novel ELM control to tackle the problems of uncertain robot manipulators. In [32], an amplitude limited controller was employed to active compensation for unknown parameters effects in dynamical systems, and the controller designed can guarantee semiglobal asymptotic stability of the task-space errors. Because FLSs have high interpretability of rich expert experience comprised with NNs, many researchers incline to deal with uncertainties terms in systems by using FLSs combining with the adaptation algorithms technology. From a mathematical point of view, the fuzzy adaptation algorithms control ideas in [24, 25, 32] depend on the output of fuzzy logic systems, mainly represented as a linear basis function, and the combinatorial coefficients can be adjusted automatically online. On one hand, many fuzzy rules must be used in order to improve the accuracy of approximation, so great rules are used for this intent. But, on the other hand, too many fuzzy rules will make much more burdens of online learning computation, and time delay must be required; these phenomena can destroy or violate the stability and performance index of the robots dynamical systems.

As an attempt to solve the above discussed problem, the bounded norm about the optimal fuzzy parameter vector is assumed in [33, 34], with which the amount of online learning updated laws is decreased to only one. However, these design methods induced another trouble that the number of the input variables increases greatly with the increase of dimension in systems. Besides, the parameter of bounded norm is assumed to be known previously in these references. It should be also noted that the above fuzzy adaptation algorithms methods can not be used in different forms of FLSs, such as FLSs without if-then rules, or the outputs of FLSs can not be written as the linear functions [35, 36]. So it is noteworthy to exploit another design adaptation algorithm that can be utilized in the FLSs with linear combinatorial parameters or a great variety of the outputs forms of FLSs. In order to achieve this purpose, we consider the approximation accuracies of FLSs, utilized to be parameter estimated online instead of online estimation of the linear combination coefficients or the ideal norm of weighted vectors. The advantage of this design is that the adaptive online learning laws only focus on the approximation accuracies. However, only the index of approximation accuracy is not enough to synthesize the adaptation algorithms fuzzy controller. Therefore, other parameters independent of the structure of FLSs are necessary to be considered. For this objective, a time-varying parameter is introduced in FLSs, and a design method is used in the master robot system which guarantees good tracking effect compared with the conventional fuzzy adaptation algorithms controller.

This paper puts forward a fuzzy logic system with time-varying parameter, which is proposed to design adaptation algorithms controller for the problem of synchronization between the robot manipulators dynamical system with unknown nonlinear terms and reference model system; we design the fuzzy adaptation algorithms controller based on sliding surface and Lyapunov stabilized theory analysis. In Section 2, some assumptions and descriptions of dynamical system are given. FLS based on linguistics is introduced in Section 3. An adaptation algorithm control is designed in Section 4. Simulation example is performed to test the designed controller, and conclusion is given in the end.

#### 2. System Description and Assumptions

The dynamic equation of the robot manipulators can be represented as follows:where is the vector of joint angular positions. , , are the inertial matrix, Coriolis and centrifugal matrix, and gravity vector, respectively. is viscous friction, denote a positive definite matrix with bounded constant satisfying , and is a known constant. is continuous external disturbance, which satisfied . denotes vector of applied joint torques and will be designed.

*Assumption 1. *Matrix vector is uniformly bounded and uniformly continuous if () is uniformly bounded and continuous and satisfies , and is a known positive scalar.

From Assumption 1, we denote and obtain .

*Assumption 2. *Function vector is unknown on the compact set , and satisfies Lipschitz condition. That is to say, there exists Lipschitz coefficient satisfying .

*Property 3. *Matrix is a known positive symmetric matrix, the inverse matrix exists, satisfying , and and are two known positive constants.

If we define state vector , dynamic system (1) can be rewritten as where is a matrix with all elements 0, and denotes identity matrix. The following desired reference model for the plant can be represented as in [37]:where , , and are all constant matrices, , , and () are chosen such that the reference model (3) can be stable. In this paper, vectors , , and are assumed to be measurable.

Defining , , then we get the error dynamical system as where and .

*Assumption 4. *The matrix pair is completely controllable; it is means that, for any given matrix , there must exist a matrix such that the Lyapunov equation as follows has only one positive definition symmetry matrix solution :The control gain matrix can be obtained by solving linear matrix inequality (LMI) for any positive matrix and matrices , , and .

#### 3. Description of Fuzzy Logic Systems

In this section, the Mamdani type FLS with fuzzy rules and the th rule is to be considered as the following format: If is and is and and is ,where and represent fuzzy sets of variable linguistics; and denote the membership function, respectively.

According to the reference in [26], then the output of fuzzy logic system (6) can be written by the following form:where . Now, a nonzero time-varying parameter is introduced in (7); then we can get the following result:

*Remark 5. *The work mechanism of the time-varying is to make the input signal enter ; the final output of the FLS is obtained as the form of (8); then the approximation accuracy can be online updated by parameter .

Lemma 6. *If the uncertain continuous nonlinear functions satisfy Lipschitz condition with coefficients and FLS exists, making , where , then the following inequality holds:*

*Proof. *Since satisfies Lipschitz condition with coefficient , then . If is true, we have the following result: This is the process of the proof for Lemma 6.

*Remark 7. *From the analysis of Lemma 6, we know that the approximate accuracy of (8) can be deduced by employing (7) and the time-varying parameter , which reveals that the accuracy of approximation can be guaranteed by the time-varying and the Lipschitz coefficient of unknown functions.

*Assumption 8. *There exist positive unknown constants and fuzzy logic systems such as the form of (7) such that .

If Assumption 8 is true, then we introduce the symbols , and we can easily get inequality . From Assumption 2, we define . For brevity, we denote and to represent the estimation values of and , respectively, and let and represent the estimate errors.

*Control Objective*. In this paper, a fuzzy adaptive controller is designed to make the output of robot manipulator system (1) track reference model (3) in finite time. Meanwhile, all the signals can be ensured to be bounded in the desired closed-loop system.

#### 4. Fuzzy Adaptation Algorithms Control Design

In the light of the control target, the following designed scheme fuzzy adaptive switching controller is given:where , , and where , , , . The parameters , , , , , , and are the given positive constants.

Theorem 9. *Consider the robot manipulators (1) and reference model (3) with Assumptions 1–8. The controller (11) with the adaptation algorithms (12) can be synchronized between the velocity and acceleration system (1) with reference model (3). At the same time, the parameter and estimate values and all can be ensured to be uniformly ultimately bounded.*

*Proof. *The procedure of the proof of Theorem 9 includes the following two cases.*Case **1* (). With this condition, we make . It is clear that ; we define the positive function , by using controller (12) with the case of Lipschitz coefficient under Assumption 2, and then the derivative of which relates to along the error dynamical system (4) can be obtained as According to [38], (13) implies that the extended state vector can reach the sliding surface in a limited period of times.*Case **2* (). We choose the candidate Lyapunov function . If Assumptions 1–8 are all satisfied, then the derivative of following the error dynamical system (4) is given by By using controller (11), we can obtain that Since and the inequality is known, then we have the following inequality:From Lemma 6 and controller (11), we can obtain In terms of adaptive laws (12), the inequality holds as follows: This is because the following inequality is true:Similarly, we have Substituting (19) and (20) into (18), it follows that Let and ; then (21) becomesand multiplying on both sides in the above inequality we have integrating both sides of (23) over , the following inequality holds: The above inequality means that all signals in the closed-loop systems are uniformly ultimately bounded, so the state of error dynamical system (4) is bounded. This is the whole proof of Theorem 9.

#### 5. Simulation Example

To illustrate the control design method in this section, the dynamic equations given as follows with a two-link manipulator are studied: where , , , , , , , , and We know , , and . The parameters in (25) are given as m/s, m, m, kg·, kg·, = 0.5 kg, and = 1.5 kg. The reference model is given byin which , , , , and the initial conditions of system (25) and (26) are selected as and , respectively. Initial values of update laws are chosen as , , and , and the parameters in the simulation are chosen as , , , , , , , , , and . We assume that is unknown. We choose , and the membership functions are chosen as , , , , , and . The FLS (fuzzy logic system) is comprised of selecting 6 fuzzy rules () to represent . If is , is , is , and is , then is . If is , is , is , and is , then is . If is , is , is , and is , then is . If is , is , is , and is , then is . If is , is , is , and is , then is . If is , is , is , is , then is .

Similarly, the FLS is constituted by following 6 fuzzy rules to represent as follows: If is , is , is , and is , then is . If is , is , is , and is , then is . If is , is , is , and is , then is . If is , is , is , and is , then is . If is , is , is , and is , then is . If is , is , is , and is , then is .

Figures 1, 2, and 3 show the simulation results of the two-link manipulators with fuzzy adaptive controller scheme, respectively.