Abstract
In this paper, the problem of guidance and motion control of mobile robots is addressed and solved within the novel framework of the mixed finite-time/\(\mathcal H_\infty \) control theory of nonlinear quadratic systems (NLQSs). Starting from a NLQS describing the dynamics of omnidirectional mobile platforms, the main tasks performed for controlling in closed loop the motion of omnidirectional robots can be conveniently formulated as a mixed finite-time/\(\mathcal H_\infty \) control problem. A robust motion controller, which can effectively rejects disturbances deviating the robot platform from a planned path, can be designed after choosing a linear state-feedback structure for the controller. The synthesis problem is solved through some sufficient conditions contemplating both norm-bounded disturbances and sets constraining initial and terminal conditions, together with a finite-time bound on the output transient. Therefore, for all the allowable uncertainties, in presence of nonzero initial conditions and exogenous disturbance inputs which are possible within an unstructured environment, the motion control tasks can be accomplished through optimal \(\mathcal H_\infty \) performance by simultaneously guaranteeing that the NLQS, which governs in closed loop the robot platform, is finite-time bounded. Finally, the applicability and control performance of the design approach have been evaluated through numerical simulations.
Similar content being viewed by others
Avoid common mistakes on your manuscript.
1 Introduction
This paper investigates the performance of a novel methodology for controlling nonlinear quadratic systems (NLQSs), based on a mixed approach combining finite-time control (FTC) and \(\mathcal H_\infty \) control theory.
As for FTC, finite-time stability (FTS) can be traced back to the papers [1,2,3]; roughly speaking, given a zero-input dynamical system, FTS requires that the state trajectories, starting from any initial condition belonging to a preassigned inner set surrounding the origin, remains confined into a given outer set in a finite-time interval. The study of FTS has seen a resurgence in interest over the past twenty-five years, and numerous references on the topic have been made available in the literature (see, for instance, the book by Amato et al. [4] and the bibliography therein).
While FTS looks at the trajectories behavior in the absence of inputs, the definition of finite-time boundedness (FTB), given in [5], requires that the state remains bounded whenever both nonzero initial conditions and the presence of external disturbances are simultaneously assumed; the initial state is still unknown and constrained to belong to a prespecified set (typically an ellipsoid), while the disturbance has to be the member of a well defined class of input signals. A sufficient condition for FTB has been given, for linear time-invariant (LTI) systems under disturbances generated by an exo-system, in [5] where the properties of quadratic Lyapunov functions have been exploited. More recently, through an approach based on operator theory, a necessary and sufficient condition for FTB has been provided in [6], by assuming that the disturbance is a square-integrable signal; moreover, such condition deals with the more general setting of linear time-varying systems.
Within FTC, the definition of FTB can be seen as the generalization of the concept of input–output (IO)-FTS (see [7, 8]) when the initial condition is nonzero; for this reason FTB is a strongest property than IO-FTS, since it guarantees state boundedness in the simultaneous presence of disturbance inputs and nonzero initial conditions.
To highlight the novelty of our contribution in the context of existing literature, we recall that, so far, the prevalent approach to FTC has been often referred to linear systems, e.g., see [9,10,11,12,13] and the book [14]. Some relevant applications to the context of NLQSs concern robotic control tasks [15], biomolecular systems [16, 17] and epidemiological processes [18]; to this regard, see also the recent paper by Nesci et al. [19] for further theoretical advancements.
In a different context, the standard \(\mathcal H_\infty \) control problem is solved by minimizing the \(\mathcal H_\infty \)-norm of the operator between the disturbance and the system output, see [20]. In this framework, the role of the transients is discarded since the minimization of the output energy, which can be achieved through the optimization of the \(\mathcal H_\infty \) norm, cannot guarantee further constraints on the output peaks over finite-time intervals. In [21], the position control of a robotic manipulator having a planar kinematic chain has been recently framed into the \(\mathcal H_\infty \) control of time-varying polytopic NLQSs.
The combination of the above-mentioned control techniques has led to the definition of IO-FTS with a \(\mathcal H_\infty \) bound, originally stated in [22] for (uncertain) linear time-invariant systems, in order to provide simultaneous constraints which guarantee that the system output is energy bounded and its instantaneous values are point-wise bounded over a finite interval of time. The control technique proposed in [22] has been tested in a real engineering application, namely the vibration control of vehicle active suspensions (see [23]), since the mixed design conditions allow to improve the ride comfort and safety of the vehicle, through an attenuation of the perturbations due to road asperities on the dynamics of the car body, in the presence of further constraints on the maximum allowable excursion of the suspension stroke and thresholds on the saturating force actuators.
In this work, we aim at controlling a three-wheeled omnidirectional mobile robot, using a mixed FTB/\(\mathcal {H}_\infty \) approach. As shown in the following sections, the robot dynamics can be described by means of a nonlinear quadratic dynamical model. Since, the results developed in [22] are only applicable to the class of linear systems, a key contribution of this work is to extend the results provided in [22] to the design of a state-feedback controller for NLQSs. Furthermore, we also replace the IO-FTS requirement, used in [22], with a FTB one, which enables taking into account also a constraint on the initial conditions of the system. Therefore, in the following, the proposed methodology is referred as the mixed FTB/\(\mathcal H_\infty \) approach which—over finite-time interval—will guarantee, in the presence of an unknown initial condition and exogenous disturbances, a point-wise bound on the system state together with the satisfaction of a quadratic-integral performance measure.
Indeed, whenever a FTC point of view is exploited to approach to the \(\mathcal H_\infty \) control problem, a performance index, which can take into explicit account the robustness properties versus all the allowable initial conditions of the system state, can be straightforwardly considered. To this regard, the work [24] contemplates nonzero conditions in the definition of a \(\mathcal H_\infty \) performance measure of the formFootnote 1
where \(\mathcal L_2^n(\Omega )\) denotes the set of square integrable vector-valued signals with n components defined in the set \(\Omega \), and, given \(z(\cdot )\in \mathcal L_2^n(\Omega )\),
represents the classical 2-norm in \(\mathcal L_2^n(\Omega )\). In (1), \(\gamma \) is a positive scalar, x(T) is the state vector at the terminal time T, and \(x_0\) is the initial state vector taken from \(\mathcal {X}_0\subset \mathbb R^{n}\) representing a set of allowable initial conditions; the matrices R and S weight the initial and terminal states, respectively.
The left hand side in (1) can be regarded as the induced norm of the linear operator generated by the closed-loop system, which maps the pair (x(0), w) to (x(T), y), see [24]. The \(\mathcal H_\infty \) performance bound expressed by condition (1) implies—under uncertainty of the initial condition—a constraint on the sensitivity of the system response, in terms of both terminal state and output energy, to energy-bounded disturbance input \(w(\cdot )\) and uncertainty in the initial conditions.
To reach our goal, we shall follow a machinery that is reminiscent of the papers [25, 26], where the authors proposed some sufficient conditions for the finite-time stabilization via linear state feedback of NLQSs in order to minimize a quadratic cost function. To this regard, the recent work [27] has dealt with finite-time boundedness and stabilization of NLQSs in the presence of disturbance inputs; some state-dependent conditions, which are based on annihilators (see [28]) and Finsler’s lemma, have been provided therein.
A preliminary version of the theoretical results provided in the present paper has been presented in [29].
The reminder of this paper is as follows. Section 2 provides some preliminary definitions and the problem statement. In Sect. 3, the guidance and motion control of omnidirectional mobile robots is formulated within the framework of mixed FTB/\(\mathcal H_\infty \) control of NLQSs. A sufficient condition for the existence of a mixed FTB/\(\mathcal H_\infty \) controller, which represents the main contribution of this work, is reported in Sect. 4. The applicability and usefulness of the conceived design methodology are shown in Sect. 5, whereas the conclusions are reported in Sect. 6.
2 Mixed FTB/\(\mathcal {H}_{\infty }\) control of NLQSs
2.1 Definition of nonlinear quadratic systems with disturbance input
Let us define a NLQS as
where \(x(t) \in \mathbb {R}^n\) is the system state, \(u(t) \in \mathbb {R}^{p}\) is the control input, \(w(t) \in \mathbb {R}^{n_w}\) is an exogenous disturbance,
with \(B_i \in \mathbb {R}^{n \times n}\), \(i=1,\dots ,n\), and
with \(N_i \in \mathbb {R}^{n \times m}\), \(i=1,\dots ,n\) .
With a time interval taken as \(\Omega := [0,T]\), it assumed that the disturbance input \(w(\cdot )\) belongs to the class of square-integrable signals
2.2 Problem statement
The mixed FTB/\(\mathcal H_\infty \) control problem is stated here for NLQSs; a linear state feedback controller is adopted in the form
where \(K \in \mathbb {R}^{m\times n}\) is the control gain matrix. By interconnecting system (3) with the controller (6), the closed-loop system is obtained as
where
Preliminarily to the definition of the mixed FTB/\(\mathcal H_\infty \) control problem, it is needed to recall from [5, 6] the definition of FTB.
Definition 1
(Finite-time boundedness) Given the time interval \(\Omega := [0,T]\), some positive definite matrices \(\Gamma _0\) and \(\Gamma \), such that \(\Gamma <\Gamma _0\), system (3), with \(u(t)=0\), is said to be finite-time bounded wrt \((\Omega , \Gamma _0, \Gamma )\) if, for any disturbance \(w(\cdot )\in \mathcal W\), the following holds
\(\Diamond \)
Within the mixed FTB/\(\mathcal H_\infty \) control theory, both the analysis and synthesis problems can exploit the following cost function
where \(\gamma >0\) is the \(\mathcal H_\infty \)-performance gain. The cost function (9) defines a quadratic-integral performance index weighting the energies of the output and disturbance input, respectively. Moreover, the matrices R and S, which are positive definite, weight the effects that both the uncertainty of the initial conditions and of the terminal state have on the transient performance.
Through a formalization in terms of \(\mathcal L_2\)-gain, it is possible to extend the \(\mathcal H_\infty \)-norm control to NLQSs (see [25]). Therefore, given \(\Omega \) and the scalar \(\gamma >0\), the existence of a state feedback control law in the form (6), such that
implies that
The definition of the performance index in (10) is well-posed under the hypothesis that the NLQS exhibits FTB; therefore, in the presence of any exogenous disturbance \(w(\cdot )\in \mathcal W\), over the interval \(\Omega \), the system state remains bounded for a prescribed set of initial conditions.
The ellipsoidal sets, which are induced by the matrices \(\Gamma _0\) and \(\Gamma \) in Definition 1, are defined as
To clarify the relationship between FTB and the \(\mathcal H_\infty \)-norm performance index (9), the following definition is needed.
Definition 2
Consider system (3) subject to a disturbance \(w(\cdot )\in \mathcal W\). Given the time interval \(\Omega \), the cost index (9), a positive scalar \(\gamma \), the ellipsoidal sets (11) and the positive definite matrices R and S, the controller (6) is said to be a mixed FTB/\(\mathcal H_\infty \) controller for the NLQS (3) if the following hold:
- (i):
-
The closed-loop system (7) is FTB wrt \((\Gamma _0,\Gamma ,\Omega )\);
- (ii):
-
The performance index (9) satisfies, for the closed-loop system (7),
$$\begin{aligned} J(\Omega ,R,S)<0,\forall w(\cdot )\in \mathcal W,x_0 \in \mathcal {E}_0. \end{aligned}$$\(\Diamond \)
Remark 1
The concept of mixed FTB/\(\mathcal H_\infty \), which contemplates both nonzero initial conditions and \(\mathcal L_2\) norm-bounded disturbances, is a novelty for NLQSs. Moreover, to extend prior results available only for linear systems, we propose a sufficient condition assuring simultaneously, for a NLQS under study, both a prescribed \(\mathcal H_\infty \) performance index over finite time and FTB. Instead, in the case of mixed IO-FTS/\(\mathcal H_\infty \) for linear systems, some distinct conditions, on \(\mathcal H_\infty \) and IO-FTS, respectively, are separately invoked (see Amato et al. [14, p. 71]). \(\triangle \)
2.3 Some preliminary lemmas
Some technical lemmas, which are useful to derive the main result, are presented in what follows.
Lemma 1
Consider the NLQS (3) subject to any disturbance input \(w(\cdot )\in \mathcal W\). Given the time interval \(\Omega :=[0,T]\), the performance measure (9), a positive scalar \(\gamma \), the ellipsoidal sets (11) and the positive definite matrices R and S, assume there exist a symmetric positive definite matrix Q, a positive scalar \(\alpha \) and a matrix K, such that
where
\(P=\Gamma _0^{\frac{1}{2}} Q^{-1} \Gamma _0^{\frac{1}{2}}\), and \(\lambda _{min}(\cdot )\) indicates the minimum eigenvalue of the argument. Then, the state feedback controller (6) is a mixed FTB/\(\mathcal H_\infty \) controller for the NLQS (3). \(\blacksquare \)
Proof
Let us consider a quadratic Lyapunov function \(v(x)=x^T P x\); its time derivative along the trajectories of the closed-loop system reads
The cost function (9), for any trajectory \(x(\cdot )\) starting from \(x_0 \in \mathcal {E}_0\), is such that
where the last inequality follows from the fact that the matrices \(S-P\) and \(P-\gamma ^2R\) are both negative definite in view of (12b).
From (15), a sufficient condition for \(J(\cdot ,\cdot ,\cdot )<0\) is
The satisfaction of condition ii) in Definition 2 can be inferred after recognizing that, from (14), condition (12a) implies (16) over \(\Omega \) if any trajectory \(x(\cdot )\) starting from \(x_0\in \mathcal {E}_0\) remains confined in the set \(\mathcal {E}_\Omega \).
Now, we are ready to show that (12a) and (12d) assure condition i) of Definition 2. Since, in the first block at the left hand side of condition (12a), we have that \(C^T C>0\), it is readily seen that
By adding \(-\alpha P\) to the first block at the LHS of (17), the above inequality implies
Condition (18) implies that
Given the candidate Lyapunov function, the induced ellipsoidal set is defined as
Condition (12c) guarantees
Following similar arguments to those ones of the proof of Theorem 1 in [26], by contradiction, we shall show that any trajectory starting from \(\mathcal E_0\) cannot exit \(\mathcal E\) for any \(t\in \Omega \).
Indeed, let us assume the existence a time instant \(t^\star \in \Omega \) such that \(x(t^\star )\) belongs to the boundary of \(\mathcal E\), it follows that
From the inclusions \(x(0) \in \mathcal E_0 \subset \mathcal E\) and continuity over time of the state trajectories, it is always possible to find a time instant \(0<\tilde{t}\le t^\star \) such that
-
(i)
\(x^T(\tilde{t})Px(\tilde{t}) = 1\) ;
-
(ii)
\(x(t)\in \mathcal E\) \(\forall t\in [0,\tilde{t}]\) .
From (12c), condition (19) is fulfilled for all \(t \in [0,\tilde{t}]\) if \(\mathcal E\subset \mathcal {E}_{\Omega }\). By multiplying both sides of (19) by \(e^{-\alpha t}\), we obtain
after integrating the above inequality from 0 to t, with \(t\in [0,\tilde{t}]\), it is readily obtained that
From conditions (23) and (12d), it follows, for all \(t\in [0,\tilde{t}]\),
The satisfaction of (24) at \(t=\tilde{t}\) implies
which contradicts (22). Therefore, a time point \(t^\star \in \Omega \) satisfying (22) does not exist, and x(t) cannot exit the set \(\mathcal E\) for \(t\in [0,T]\). The proof follows from the fact that \(\mathcal E\) is a subset of \(\mathcal E_{\Omega }\).
Although Lemma 1 provides a sufficient condition guaranteeing that a mixed \(FTB/\mathcal H_\infty \) controller for the NLQS (3) exists, it cannot be exploited in practice, since the LHS of the LMI (12a) depends on x. The effort of the next section will be devoted to translate conditions (12) into a set of algebraic LMIs.
To this end, we need to recall the following lemmas.
Lemma 2
([30]) Consider the positive definite matrix \(P\in \mathbb {R}^{n\times n}\), and a vector \(\nu \) such that \(\Vert \nu \Vert =1\). Then, any point on the boundary of the ellipsoid (20) can be parameterized as \(x = P^{-1/2}\Lambda \nu \), with \(\Lambda \) any matrix such that \(\Lambda ^T\Lambda =I\).\(\blacksquare \)
Lemma 3
([31]) Given any scalar \(\epsilon >0\), a triplet of matrices of compatible dimensions \(\Phi \), \(\Theta \) and \(\Pi \), with \(\Pi ^T \Pi \le I\), then
\(\blacksquare \)
3 Formulation of the control problem for omnidirectional mobile robots
3.1 NLQS modeling of three-wheeled robot platforms
In [32], the authors have derived the motion equations of omnidirectional mobile robot platforms consisting of three lateral orthogonal-wheel assemblies distributed at an equal distance from the center of gravity of the robot. Each assembly is configured to reproduce the behavior of an ideal spherical wheel. As an alternative, three mechanic wheels could be adopted without the necessity of complex wheel assemblies (see Fig. 1).
Through the on-board installation of an inertial measurement unit (IMU), it is possible to acquire all the feedback variables required for the implementation of a tracking control law (6) for the omnidirectional platform. Therefore, by adopting the mobile reference system in Fig. 1, the state vector can be taken as \(x=\begin{pmatrix} \dot{x}_m&\dot{y}_m&\dot{\phi }\end{pmatrix}^T\).
Starting from the motion equations expressed in the mobile reference system on the robot frame (see Watanabe et al. [32, p. 319]), the dynamics of a three-wheeled omnidirectional robot can be described through a NLQS (3), where the matrices in the state equations (3a) are given by
with
The parameter \(I_v\) is the moment of inertia of the robot, \(I_w\) is the moment of inertia of the wheel around the driving shaft, c the viscous friction factor of the wheel, m the mass of the robot, L the distance between any wheel assembly and the center of gravity of the robot, k the driving gain factor, r the radius of the wheel and \(u_i\), \(i=1,\dots ,3\), are the driving input torque on each wheel.
3.2 Guidance and tracking control system
The application to the quadratic model of a three-wheeled omnidirectional of the proposed mixed FTB/\(\mathcal H_\infty \) control framework can help to evaluate its usefulness in a real case-study.
The main components and line signals of the closed-loop system implementing the guidance and tracking control of the robot are represented in the block diagram of Fig. 2. In an unstructured environment, the robot can proceed forward along a desired direction by keeping the cruise velocity; in the presence of an obstacle or in the case of an unwanted configuration of the robot, the guidance controller must update the set-point values of the rotational and translation velocities, which are sent to the tracking controller, to reconfigure the path of the robot platform. The presence of an obstacle along the forward direction of the path can be detected through an array of optical or ultrasonic sensors placed on the anterior portion of the robot frame.
3.3 Mixed FTB/\(\mathcal {H}_\infty \) tracking controller
The control problem for the motion tracking of the omnidirectional robot can be conveniently formulated as a mixed FTB/\(\mathcal {H}_\infty \) control problem.
Indeed, to effectively attenuate the impact of the exogeneous inputs and uncertainty on the initial conditions over finite time, and to robustly follow the planned path, the tracking controller has to compensate some bounded-energy disturbances, also when such perturbations are of short duration and large magnitude. In this case, to gain the desired robustness, the control performance has to be mandatorily optimized over the transient.
Both the robustness and accuracy of the tracking control performance can be optimized after defining a \(\mathcal {H}_\infty \) cost function (9), in order to minimize the sensitivity of the tracking error to the energy-bounded disturbances perturbing the trajectory of the omnidirectional robot. Moreover, by specifying in (9) a set of non-null initial conditions of the kinematic variables, on which the mixed FTB/\(\mathcal {H}_\infty \) control performance has to be guaranteed, it is possible to enlarge the operating range of the closed-loop control system.
By exploiting the results presented in Sect. 4, after recasting the quadratic model of a guidance and tracking control system of omnidirectional mobile robots in the form (7), it is possible to design a state-feedback tracking controller having the properties stated in Definition 2.
4 Mixed FTB/\(\mathcal H_\infty \) controller design
According to Definition 2, the existence of a mixed FTB/\(\mathcal H_\infty \) controller for NLQSs can be proved through the sufficient condition provided by the following Theorem.
Theorem 1
Given the NLQS (3) subject to any disturbance \(w(\cdot )\in \mathcal W\), the time interval \(\Omega \), the performance measure (9), a positive scalar \(\gamma \), a pair of positive definite matrices R and S, and the admissible sets \(\mathcal {E}_0\) and \(\mathcal {E}_\Omega \) defined in (11), let \(X=\Gamma _0^{-\frac{1}{2}} Q \Gamma _0^{-\frac{1}{2}}\), and assume there exist positive scalars \(\alpha \) and \(\epsilon \), a matrix L and a positive definite matrix Q such that
with \(\Xi :=\begin{pmatrix} (B_1X+N_1 L)^T&\dots&(B_nX+N_n L)^T \end{pmatrix}\) and \(\widehat{\Gamma }=I\otimes \Gamma \,;\) then, letting \(K=LX^{-1}\), (6) is a mixed FTB/\(\mathcal H_\infty \) control law.
Proof
Starting again from the candidate Lyapunov function \(v(x)=x^T P x\) and the set (20) induced by v(x), Lemma 1 provides the main arguments used in this proof.
First, by exploiting the results in Lemmas 2 and 3 and after recognizing an affine structure in the involved matrix functions, we shall show that (28f) implies (12a).
To this aim, as first intermediate step, it is needed to show that (28a) and (28b) imply (21). Exploiting Schur complements (see Boyd et al. [33, p. 7]), from (28a) we obtain that \( X^{-1}>\Gamma \); in the same way, we have that (28b) implies \(X^{-1}<\Gamma _0 \). From the inversion \(P= X^{-1}\), inclusions (21) hold.
Now we prove that (28f) implies (12a). Condition (12a) is assured if
with \(\Theta (x)\) given in (13). Through pre- and post-multiplying of the LHS in (29) by
then letting \(X=P^{-1}\) and \(K P^{-1}=L\) and applying Schur complements, condition (29) can be equivalently rewritten
where \( \Psi :=\begin{pmatrix} (B_1X+N_1 L)^T&\dots&(B_nX+N_n L)^T \end{pmatrix} \,.\)
Note that since \(\mathcal {E}_\Omega \) is a convex set and the left hand side in (30) depends linearly on x, condition (30) holds for all x in the set \(\mathcal {E}_\Omega \) if and only if it is satisfied on the boundary of \(\mathcal E_\Omega \). Now, by virtue of Lemma 2, any point x on such boundary can be parameterized as \(x=\Gamma ^{-1/2}U v , U^T U =I,\Vert v\Vert =1\,\), and therefore, for any positive \(\epsilon \), we have
From (30) and (31), by exploiting Schur complement arguments, it is readily seen that condition (28f) implies (12a). Moreover, through further application of the Schur complements, the satisfaction of (12b) follows from (28c) and (28d). With \(\mathcal {E} \subset \mathcal {E}_\Omega \) in force and from (28e), the proof follows.
Remark 2
Note that condition (28e) can be satisfied by imposing
For a fixed value of \(\alpha \), the conditions of Lemma 1 can be solved as a LMI optimization problem; a one-parameter search is used to find an optimal value for \(\alpha \).\(\triangle \)
5 Control law design and simulation results
Thanks to the omnidirectionality of the robot platform, the commands which are updated by the guidance controller can be independently actuated through the following acts of motion:
-
Orientation: the robot must rotate around its vertical axis in order to give to the mobile reference system the desired orientation.
-
Linear motion: the robot is commanded to follow a straight line keeping the desired direction.
The task that we shall accomplish for the testing of the control performance—in closed loop—is the tracking of a straight line along a desired direction, which consists of a basic movement for the navigation in an unstructured environment.
Therefore, to robustly follow the desired direction by following a rectilinear path at cruise velocity, zero set-point values can be assumed for the linear transversal and rotational velocities, taking into account that the output components of the model consist of two translation velocities \(\dot{x}_m\) and \(\dot{y}_m\) with respect to the transversal \(x_m-\) and longitudinal \(y_m-\) axes, respectively, and the rotation velocity \(\dot{\phi }\) about the z-axis of the robot platform.
We want the robot to move freely along the vertical axis, rejecting disturbances that would cause it to depart from its planned path. In order to maintain the planned direction, it is important to ensure that the controller is sufficiently robust to guarantee that the velocity \(\dot{x}_m\) has a value sufficiently close to zero over finite-time.
The values of the physical parameters in (26)–(27) are taken as
and zero initial conditions are assumed.
The proposed framework enables to optimize the control action by prescribing a maximum amplitude of each component of u(t); thus, the loop shaping method has been implemented. Define then the transfer function that weights the sensitivity function as
where \(M_{\max }=2\) dB is the high-frequency gain, \(M_{\min }=1/100\) dB is the low-frequency gain and \(w_B=1\,\mathrm{rad/s} \) is the gain crossover frequency. The corresponding differential equations are
From the series interconnection between (3) and (33) the augmented system reads
To accelerate the system response, the variable to be controlled was chosen to be an order of magnitude more than the gain crossover frequency
A time bound \(T=100\) is chosen, with the matrices in (11) given by
A bounded-energy disturbance belonging to the class of signal in (5), with \(d=1.5\cdot 10^3\), is assumed to be in input to the control system. The input signal is a step having an unitary amplitude. Such perturbation can represent an external force generating a lateral drift on the robot wheels, e.g., due to an alteration in the wheels grip and friction with the ground. This implies that the robot trajectory can be modified relatively to the rectilinear path.
The controller is designed on the basis of the result in Theorem 1. Therefore, a maximum allowable \(\mathcal H_\infty \)-performance gain is chosen as \(\gamma =3.1\), in order to guarantee robustness against disturbance input and possible uncertainties starting from non-null initial conditions. The matrices, which weight initial and terminal conditions, are chosen as
With \(\Omega = [0, 100]\), through the MOSEK optimization toolbox within the Matlab LMI environment, a solution for the mixed FTB/\(\mathcal H_\infty \) controller is found, with
As it is shown in Figs. 3 and 4, by rejecting the disturbance, the output is able to give a reference-tracking result that is satisfactory, despite the velocity along x is slightly greater than zero. Therefore, the controller achieves the attenuation of the effect of the disturbance on the control system output.
By simulating the system response, in a set of tests generated from the combination of different initial conditions and disturbance inputs, the expected performance of closed-loop system has been validated.
In addition, a robustness analysis based on parametric variation was conducted. The velocities were calculated assuming a 5% variation in \(a_1\) and \(a_2\). Figure 5 demonstrates that the speed along x is unchanged, whereas the speed along y exhibits a deviation of 0.5. Figure 6 depicts essentially the same scenario, assuming a 10% variation for parameters \(a_1\) and \(a_3\) and a 5% variation for parameter \(a_2\).
The simulation results demonstrate the benefit of combining \(\mathcal H_\infty \) and FTC constraints. Through \(\gamma \) it is possible to optimize the control performance, over finite-time horizon, both in terms of the energy of the system output and its robustness versus \(\mathcal L_2\) norm-bounded disturbances. Moreover, the loop shaping technique allows the simultaneous optimization of the control effort.
6 Conclusions
In this work, an extension of the mixed FTC/\(\mathcal H_\infty \) theory to the class of NLQSs has been provided; its concrete applicability has been evaluated in the context of the guidance/tracking control system design for omnidirectional mobile robots.
The main methodological contribution concerns a sufficient condition ensuring the mixed FTB/\(\mathcal H_\infty \) closed-loop performance for a given NLQS under linear state feedback control. A mixed FTB/\(\mathcal H_\infty \) law is designed through the solution of a LMIs optimization problem. The designed controller can guarantee, other than FTB, a prescribed \(\mathcal H_\infty \) performance, which can be optimized—over finite-time horizon—on the basis of a versatile set of constraints, involving both bounds on the system output together with initial and terminal conditions, as shown by a design application concerning the closed-loop control of a omnidirectional mobile robot characterized by a three-wheeled frame. Indeed, the dynamic equations in the mobile reference system of such robot have been described through a NLQS.
The optimality and robustness of control performance, achievable in a closed-loop control scheme by the proposed design methodology, have been validated through numerical simulations. Through the robotic application, it has been possible to show the advantages provided by the proposed FTB/\(\mathcal H_\infty \) control approach. Moreover, by complementing the design methodology with a loop shaping technique, an improved disturbance rejection has been achieved under the minimization of the control effort.
Data availability
The source code and data presented in this work are available upon request to the corresponding author.
Notes
Notation: The symbol \(\mathcal L_2^{n_w}(\Omega )\) denotes the subspace of vector-valued functions in \(\mathbb R^{n_w}\) which are square-integrable over the time interval \(\Omega :=[0,T]\) with Euclidean vector norm \(\Vert f\Vert _{\Omega }=[\int _{\Omega } \Vert f(\sigma )\Vert ^2 \textrm{d}\sigma ]^{1/2}\)
References
Lebedev A (1954) The problem of stability in a finite interval of time. J Appl Math Mech 18:75–94
Dorato P (1961) Short time stability in linear time-varying systems. Proc IRE Int Conv Rec Part 4:83–87
Weiss L, Infante E (1967) Finite time stability under perturbing forces and on product spaces. IEEE Trans Auto Control 12:54–59
Amato F, Ambrosino R, Ariola M et al (2014a) Finite-time stability and control. Lecture notes in control and information sciences. Springer, London
Amato F, Ariola M, Dorato P (2001) Finite-time control of linear systems subject to parametric uncertainties and disturbances. Automatica 37(9):1459–1463
Amato F, Ambrosino R, Ariola M et al (2019) On the finite-time boundedness of linear systems. Automatica 107:454–466. https://doi.org/10.1016/j.automatica.2019.06.002
Amato F, Ambrosino G, Cosentino C et al (2010) Input-output finite-time stabilization of linear systems. Automatica 46(9):1558–1562
Amato F, Carannante G, De Tommasi G et al (2012) Input-output finite-time stability of linear systems: necessary and sufficient conditions. IEEE Trans Automat Control 57(12):3051–3063. https://doi.org/10.1109/TAC.2012.2199151
Ichihara H, Katayama H (2009) Necessary and sufficient conditions for finite-time boundedness of linear continuous-time systems. In: Proceedings of the 48h IEEE conference on decision and control (CDC) held jointly with 2009 28th Chinese control conference, pp 3214–3219. https://doi.org/10.1109/CDC.2009.5400748
Wang Y, Liu Y, Zuo Z (2014) Finite-time boundedness of switched delay systems: the reciprocally convex approach. IET Control Theory Appl 8(15):1575–1580. https://doi.org/10.1049/iet-cta.2013.1136
Huang S, Xiang Z, Karimi H (2014) Input-output finite-time stability of discrete-time impulsive switched linear systems with state delays. Circuits Syst Signal Process 33(1):141–158
Kussaba HT, Borges RA, Ishihara JY (2015) A new condition for finite time boundedness analysis. J Franklin Inst 352(12):5514–5528. https://doi.org/10.1016/j.jfranklin.2015.09.005
Zhang S, Guo Y, Wang S et al (2019) Input–output finite-time stability of discrete-time systems under finite-time boundedness. Int J Syst Sci 50(2):419–431. https://doi.org/10.1080/00207721.2018.1554170
Amato F, Tommasi GD, Pironti A (2018) Finite-time stability: an input-output approach. Wiley, New York
Amato F, Colacino D, Cosentino C et al (2014b) Guaranteed cost control for uncertain nonlinear quadratic systems. In: Proceedings of the 2014 European control conference (ECC), Strasbourg, France, pp 1229–1235
Salerno L, Cosentino C, Merola A et al (2013) Validation of a model of the GAL regulatory system via robustness analysis of its bistability characteristics. BMC Syst Biol 7(1):1–14
Cosentino C, Salerno L, Passanti A et al (2012) Structural bistability of the gal regulatory network and characterization of its domains of attraction. J Comput Biol 19(2):148–162
Amato F, Tommasi GD, Merola A (2012) State constrained control of impulsive quadratic systems in integrated pest management. Comput Electron Agric 82:117–121
Nesci F, Merola A, Dragone D et al (2022) Input–output finite-time stability and state-feedback stabilization of nonlinear quadratic systems. In: 2022 IEEE 17th international conference on control and automation (ICCA), pp 850–855
Doyle J, Glover K, Khargonekar P et al (1989) State space solution to standard \(\cal{H} _2\) and \(\cal{H} _\infty \) control problem. IEEE Trans Auto Control 34(8):831–847
Lu L, Zhu P, Ke J et al (2018) \(\cal{H}_\infty \) control of time-varying polytopic quadratic systems with applications to a robotic manipulator. In: 2018 Chinese automation congress (CAC), pp 74–79
Amato F, Cosentino C, Tommasi GD et al (2016) The mixed robust \(\cal{H} _\infty \)/FTS control problem analysis and state feedback control. Asian J Control 18(3):828–841
Xue W, Li K, Chen Q et al (2018) Mixed FTS/\(\cal{H} _\infty \) control of vehicle active suspensions with shock road disturbance. Veh Syst Dyn 57(6):841–854
Khargonekar PP, Nagpal KM, Poolla KR (1991) \(\cal{H} _\infty \) control with transients. SIAM J Control Optim 29(6):1373–1393
Merola A, Cosentino C, Colacino D et al (2017) Optimal control of uncertain nonlinear quadratic systems. Automatica 83:345–350
Amato F, Cosentino C, Merola A (2010) Sufficient conditions for finite-time stability and stabilization of nonlinear quadratic systems. IEEE Trans Automat Control 55(2):430–434
Bhiri B, Delattre C, Zasadzinski M et al (2017) Results on finite-time boundedness and finite-time control of non-linear quadratic systems subject to norm-bounded disturbances. IET Control Theory Appl 11(10):1648–1657
Trofino A (2000) Robust stability and domain of attraction of uncertain nonlinear systems. In: Proceedings of the 2000 American control conference, vol 5. ACC, pp 3707–3711. https://doi.org/10.1109/ACC.2000.879262
Merola A, Nesci F, Dragone D et al (2023) Mixed fts/\({\cal{h}}_\infty \) control for nonlinear quadratic systems subject to norm-bounded disturbances. In: 2023 IEEE 31th mediterranean conference on control and automation (MED), pp 630–635
Valmorbida G, Tarbouriech S, Garcia G (2010) State feedback design for input-saturating nonlinear quadratic systems. Automatica 46(7):1196–1202
Wang Y, Xie L, de Souza CE (1992) Robust control of a class of uncertain nonlinear systems. Syst Control Lett 19(2):139–149
Watanabe K, Shiraishi Y, Tzafestas SG et al (1998) Feedback control of an omnidirectional autonomous platform for mobile service robots. J Intell Rob Syst 22(3/4):315–330. https://doi.org/10.1023/a:1008048307352
Boyd S, Ghaoui LE, Feron E et al (1994) Linear matrix inequalities in system and control theory. SIAM Press
Acknowledgements
None.
Funding
No financial support to acknowledge. Open access funding provided by Università degli studi "Magna Graecia" di Catanzaro within the CRUI-CARE Agreement.
Author information
Authors and Affiliations
Contributions
CC and AM were involved in the conceptualization; AM and FA contributed to the methodology; FN and DD were involved in the formal analysis and investigation; FN and DD performed the validation; FN curated the data; AM, CC and FN were involved in writing—original draft preparation. FA and CC assisted in writing—review and editing; AM and FA were involved in the supervision. All authors have read and approved the final version of the manuscript.
Corresponding author
Ethics declarations
Conflict of interest
All authors declare that they have no conflict of interest to disclose.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Nesci, F., Amato, F., Dragone, D. et al. Control of a three-wheeled omnidirectional mobile robot via a mixed FTB/\(\mathcal H_\infty \) approach. Int. J. Dynam. Control (2024). https://doi.org/10.1007/s40435-024-01430-7
Received:
Revised:
Accepted:
Published:
DOI: https://doi.org/10.1007/s40435-024-01430-7