Fractional Order Control for the Grasping by Coiling

The paper studies the grasping control by coiling for a class of grippers with continuum arms. First, the fractional order model of continuum arm is proposed. The viscoelastic components are assimilated by the fractional Kevin-Voigt model and the fractional-order dynamics are inferred. A hybrid control technique with two control loops for position and force control, respectively, is proposed. The Lyapunov method for position control is applied. Sector-type constraints for input torque are implemented and frequential conditions, Popov Circle Criterion, that ensure asymptotic stability, are inferred. A conventional PD controller is proposed for the force control. Numerical simulations illustrate the performances of the control system.


Introduction
In this paper, a class of arms with continuum elements that performs the grasping function by coiling is analysed. This function is often met in the animal world as octopus tentacle, the elephant's trunk or constrictor snakes. In robotic applications, this function is used to manipulate objects with geometric shapes that contain borders with concave areas (Figure 1), to objects with certain fragility and resistance to manipulate, etc. In this area of application, the architectures of continuum arm type play an important factor in determining the control laws. The specialized literature contains a large number of papers addressing the study of these models known as systems with distributed parameters. A summary of these results can be found in [1]. Among the pioneering works we can mention [2][3][4] which introduce the basic concepts of these models, the state equations. The dynamic models and the control laws for constrained motions are inferred in [5,6]. In [7] are discussed kinematic algorithms to avoid the obstacles of hyper-redundant manipulators. An extension of Jacobian based methods for infinite dimensional systems is presented in [8]. In [9] are studied the grasping configurations of continuum arms. Similarities of these DOI: 10 systems with certain classes of biological models are analysed in [10]. Kinematic control for real time implementation is discussed in [11]. Sliding mode control techniques for hyper-redundant robots are studied in [12]. Architectures based on concentric tubes for the continuum arm are presented in [13]. Spatial weight error control for a class of hyper-redundant robots is treated in [14]. Shape control of the tentacle robots is presented in [15,16]. This paper studies the grasping control by coiling for a class of grippers with continuum arms. First, the fractional order model of continuum arm is proposed. The viscoelastic components are assimilated by the fractional Kevin-Voigt model and the fractional-order dynamics are inferred. A hybrid control technique with two control loops for position and force control, respectively, is proposed. The Lyapunov method for position control is applied. Sector-type constraints for input torque are implemented and frequential conditions that ensure asymptotic stability, similar to the Popov Circle Criterion for classical systems, are inferred.
A conventional PD controller is proposed for the force control. Numerical simulations illustrate the performances of the control system. The paper is structured as follows: Section 2 presents technological principles, section 3 treats the dynamic model, section 4 describes the control system, section 5 verifies the control techniques by numerical simulations, section 6 presents discussions and section 7 resumes the conclusions.

Technological Principles
The main component of the grasping system is the continuum arm that manipulates the object by exerting a sufficiently high pressure on the contact surface so that it can execute its displacement on a predetermined path. We denote this system as "Continuum Arm Grasping System (CAGS)". CAGS can achieve any position and orientation in 3-D space to surround the manipulated load on the surface of different geometric shapes. CAGS constitutes the terminal component of a conventional robot intended for current part handling functions.
is the Caputo fractional derivative of order β . The dynamic model is derived by using Lagrange equations [5,14] for this class of hyper-redundant arms. We consider a light weight arm with following parameters: I ρ is inertial rotational density, E is equivalent elastic modulus and I is the moment of inertia. It is assumed that during the grasping operation, the arm is subjected to a disturbance, the moment m , distributed along the contact surface. In terms of the orientation variables 1 2 , u u this model can be rewritten as The main component of the CAGS mechanical structure is a long flexible backbone. It is surrounded by an elastic envelope and a viscoelastic solution fill the whole system ensuring a large flexibility ( Figure 2). The actuation system consists of antagonistic Bowden cables with DC motor driving. The mechanical architecture is composed of several segments with independent Bowden cable driving system that allows a decoupled control strategy for each arm segment. The measuring system consists of two components: A measuring system of the geometric parameters of the orientation of the arm and one for measuring the compression forces during the grasping operation. The first system is achieved by measuring the lengths of the Bowden cables [16]. The second system is consisted by an array of Force Sensitive Resistors (FSR) ( Figure 2). They are constructed of several thin flexible layers that varies in resistance as pressure is applied and released [17].
The backbone of the arm is defined by a 3-dimensional curve C and a length parameter s , where l is the length of the arm. The arm orientation is given by two angles 1 u ,

Dynamic Model
The dynamic behaviour of the CAGS is deter- The nonlinear term ( ) h u is determined by the gravitational component and satisfies the constraints [14,15].
Where, χ is a positive constant. The system output is evaluated as a weighted average The evolution error will be The dynamic model (2) The fractional order model will be [18][19][20][21] ( ) . (12) The following fractional order model can be in-ferred from (9)-(12) (for simplification of notations the index i is eliminated starting from here), [ ]

Grasping Control System
The control of the grasping operation is a hybrid control system, position-force control that com-prises two phases. In the first phase, the free move-ment of the arm, without contact with the object, Where , k k α β define the new sector parameters. Algorithm 1. The dynamic system (19), with the control variable (20), (21) is asymptotically stable if: Where the following notations were used: Where, from (21) yields The frequency inequality (23) can be interpreted in terms of the constraints (21), (26) as [22], is controlled to the target position t u (Figure 4), in the second phase; the grasping operation is per-formed, the arm exerting on the object the needed compression force. The control system is shown in Figure 5.
Two cases will be analyzed: The control system for a free movement to a desired position and the control system for the control of the grasping com-pression forces.

Grasping position control
Dynamic model is rewritten in terms of error Consider the control law The force constraints (4) determine sector type restrictions on the gain k ,

Grasping force control
Consider that the stress-strain ( ) , f  relation of the Kevin-Voigt model (1) can be approximated by > , > p p p p . that verify Proof: The proof is simple and is obtained by Lyapunov conventional techniques [22,27].

Numerical Simulations
Consider a grasping control problem for a 2-D motion in a horizontal plane (Figure 7). The following parameters characterize the mechanical system: the length of the arm segment 0.15 l = m , the rotational inertial density Consider that ( ) , C α β σ σ to be the circle in the complex plane ( Figure 6) defined by the points + 0 j α σ − and + 0 j β σ − . In this case, the condition (26) can be defined as: all points of plot of ( ) G jω must be outside the circle ( ) , C α β σ σ , or as the plot of ( ) G jω does not intersect ( ) , C α β σ σ (Figure 6).
Because ( ) G jω defines the frequency characteristics of the arm and ( ) , C α β σ σ represents the input constraints, the stability condition can be rewritten as:" the both domains must not intersect".
The relation (28) can be considered as an extension of the Popov's Circle Criterion for this class of fractional order models [22].
Convergence Algorithm Proof. The following Lyapunov function is used Where, = > 0 T P P . This Lyapunov function verifies asymptotic stability conditions [19,23,24]. From [21][22][23][24][25][26], the fractional derivative of (29) becomes From (19), (21), (25), this inequality is rewritten as The system is fractional order stable but is not Hurwitz stable [28][29][30]. The matrix   ing for a class of grippers with continuum arms. First, the fractional order model of continuum arm is proposed. The viscoelastic components are assimilated by the fractional Kevin-Voigt model and the fractional-order dynamics are inferred. A hybrid control technique with two control loops for position and force control, respectively, is proposed. The Lyapunov method for position control is applied. Sector-type constraints for input torque are implemented and frequential conditions that ensure asymptotic stability, similar to the Popov Circle Criterion for classical systems, are inferred. A conventional PD controller is proposed for the force control. Numerical simulations illustrate the performances of the control system.

Discussion
The dynamic model of the continuum arm is associated with the light weight model, for which the gravitational component has a small influence. If this component increases, the stability conditions are getting worse.
In Figure 10 are shown two plots of the linear and nonlinear continuum arm models, respectively. We remark that the nonlinear component moves the plot of ( ) G jω to the left, toward the circle C which leads to a decrease in stability. For high value of nonlinearity coefficient χ , when ( ) G jω intersects the circle C , the system becomes unstable.

Conclusions
The paper studies the grasping control by coil-