780 IEEE SYSTEMS JOURNAL, VOL. 15, NO.

1, MARCH 2021

Finite-Time Robust Admissible Consensus Control of

Multirobot System Under Dynamic Events
Anuj Nandanwar , Narendra Kumar Dhar , Dmitry Malyshev , Larisa Rybak , and Laxmidhar Behera

Abstract—This article addresses the problem of event-based spaces, and d) cannot reach the goal if obstacles are near. The
consensus in a leader–follower multiagent system framework prone work in [4] is simple, robust to time-delays during communi-
to external bounded disturbance. The proposed approach has cation, and easy to implement in real-time applications. The
three parts. The first part defines a novel measurement error
based on sliding surface for super-twisting sliding-mode controller. common feature of [3]–[7] is that they consider continuous
The Lyapunov stability analysis is then used to derive a dynamic communication between the neighboring agents. The update of
event-triggering condition for control updates. The event-based control inputs occur at each sampling instant. During real-time
control updates guarantee stability along with the desired consen- application, continuous communication and computational re-
sus amongst agents (robots). The second part derives a bound on sources are important concerns for multiagent systems.
reaching time to the sliding surface, thereby guaranteeing finite-
time consensus control for each agent. The third part guarantees The event-based control scheme [8], [9] is an effective way to
the admissibility of event-based control updates for each agent. reduce the resource consumption in such system. Event-based
The robustness of the proposed approach is validated through controllers generate control input in aperiodic manner based
simulation and real-time experiments using three Pioneer P3-DX on certain rule(s) or condition(s). In [10], an event-triggered
mobile robots in a multiagent framework. The real-time experi- controller has been developed for linear stochastic system of
mental results prove the reduction in computational burden of the
entire system as control updates for two followers are found to be first-order. Based on distributed events, a self-triggered approach
approximately 28.33% and 23.33%, respectively, in the presence has been used to save energy for the system [11]. Dimarogonas
of disturbances. et al. [12] have presented a preliminary work on event-based
Index Terms—Admissibility, consensus control, event-trigger, controller for MAS. The event-triggering strategy is applied
Lyapunov stability, sliding-mode control. to handle the problem of consensus in MAS with agents hav-
ing single-integrator dynamics [13], general linear model [14],
I. INTRODUCTION and second-order system dynamics [15]. Zhu and Jiang [16]
N THE recent two decades, distributed cooperative control addressed the consensus problem among leader–follower with
I and consensus problem for multiagent systems (MASs) [1]
have garnered large attraction because of their wide-spread
general linear model and input delay, whereas Yu et al. [17]
investigated the consensus problem in leader–follower frame-
application in various areas such as area coverage and ex- work using event-triggering for discrete first- and second-order
ploration, search and rescue mission [2], synchronization [3], systems with packet losses and time-varying delay. The major
and distributed optimization. Numerous works have been pro- drawback of all these works is that they show consensus asymp-
posed related to consensus of MAS such as heterogeneous totically but do not guarantee it in finite-time.
linear agent [3], first-order consensus with communication time- The event-based consensus of leader–follower and leaderless
delay [4], second-order consensus [5], and nonholonomic sys- MAS in finite-time has been presented in [18]. The work in [19]
tem [6]. A very popular approach to achieve consensus has been studies time-triggered formation control of MAS in finite-time
artificial potential function [5]. It has some limitations [7]: a) with nonholonomic robots as agents. The works [18] and [19]
local minima leading to traps, b) no way out when obstacles do not consider the effect of disturbance. An event-based H∞
are nearby, c) oscillations occurring near obstacles and narrow control is designed for consensus of discrete time-varying MAS
with external disturbance [20]. This approach does not guarantee
consensus in finite-time.
Manuscript received October 6, 2019; revised January 31, 2020; accepted
February 29, 2020. Date of publication April 17, 2020; date of current version The sliding-mode control (SMC) is a very popular robust con-
March 9, 2021. This work was supported in part by the Department of Science trol technique and it rejects bounded disturbances and matched
and Technology of India under the research project DST-EE-2018309 and in uncertainties. The second-order SMC (SOSMC) very efficiently
part by the Russian Foundation for Basic Research under the research project
18-57-45014 IND_a. (Corresponding author: Anuj Nandanwar.) fulfills this purpose. It belongs to the class of higher order
Anuj Nandanwar, Narendra Kumar Dhar, and Laxmidhar Behera are with SMC (HOSMC). One such controller is super-twisting SMC
the Department of Electrical Engineering, Indian Institute of Technology (STSMC) [21]. This approach steers the sliding variable to
Kanpur, Kanpur 208016, India (e-mail: [email protected]; nardhar@iitk.
ac.in; [email protected]). zero without using its time derivative for systems with relative
Dmitry Malyshev and Larisa Rybak are with the Department of Mechani- degree of two. The approach proposed in [22] shows exponential
cal Engineering, Belgorod State Technological University, Belgorod 308012, convergence for linear systems, whereas there is no guarantee of
Russian Federation (e-mail: [email protected]; [email protected]).
Digital Object Identifier 10.1109/JSYST.2020.2979271 finite-time convergence for higher order systems as the SOSMC

1937-9234 © 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See https://www.ieee.org/publications/rights/index.html for more information.

Authorized licensed use limited to: Indian Institute Of Technology (IIT) Mandi. Downloaded on July 12,2023 at 11:22:02 UTC from IEEE Xplore. Restrictions apply.
NANDANWAR et al.: FINITE-TIME ROBUST ADMISSIBLE CONSENSUS CONTROL OF MULTIROBOT SYSTEM UNDER DYNAMIC EVENTS 781

algorithm lacks local lipschitzness at origin. The SMC ap-

proach [23] handles tracking problem in finite-time for systems
showing lipschitzness in their dynamics and prone to external
disturbances. An integral SMC approach achieves consensus
in finite-time for distributed second-order MAS with bounded
disturbances [24]. Liu et al. [25] proposed a second-order super-
twisting algorithm for the leader–follower MAS with system un-
certainties and external disturbances. Yang et al. [26] proposed
a methodology which is simple to design and reduces the com-
putational burden during practical applications. It guarantees Fig. 1. Schematic of robot model.
finite-time reachability and fault identification during operation
even in the presence of stochastic noise, disturbance in output,
1) design of dynamic event-based STSMC methodology for
and time delays. The simultaneous reconstructions of faults and
consensus in MAS;
states are achieved using reduced-order sliding-mode observer
2) the guarantee of finite-time consensus and stability of
for time-delayed Markovian jump systems in [27]. Apart from
MAS;
simultaneous reconstructions, the methodology in [28] annuls
3) the guarantee of lower bound on interevent execution time
the switching issue of sliding surfaces and avoids the necessity
for admissibility of control updates; and
of reachability assay. All these works follow time-triggered
4) the real-time desired consensus in leader-follower frame-
approaches.
work with directed graph topology.
An event-based SMC technique seems to be the best-suited
The rest of this article is organized as follows. The pre-
approach for disturbance rejection as well as reduction in com-
liminaries and problem formulation are given in Section II.
munication and computational cost in a complex MAS. It is
Section III has three subsections. The first one derives dynamic
very difficult to derive a triggering strategy for finite-time con-
event-triggering condition. The second and third subsections
sensus in the presence of nonlinear consensus protocol. For this,
present finite-time consensus and admissibility of event-based
Behera and Bandhopadhay [29] assumed a nonlinear function
control, respectively. Section IV presents simulations, real-time
in the locally Lipschitz continuous-time system dynamics with
experiments, and comparative analyses. The conclusions are
a unique equilibrium point. Nair et al. [30] have proposed an
drawn in Section V.
event-based integral SMC for multirobot consensus but the
events for control updates are obtained through static threshold.
II. PRELIMINARIES AND PROBLEM FORMULATION
As compared to HOSMC, integral SMC in [30] ignores the
reaching phase. It forces the entire system response to pass A. Multiagent System Framework
through sliding mode. Though it reduces the sliding function
A distributed MAS framework is considered for p robots as
to zero, it cannot do so with the derivatives of higher order.
agents. These agents have single-integrator dynamics. They can
It cannot guarantee all the states to be finite-time stable. The
interact and transmit data among each other. The communication
other limitation of [30] is the chattering phenomena due to dis-
topology between them is assumed to be a directed and weighted
continuities generated by the control law. Though the approach
graph devoid of self-loops, G = (V, E, A) [30].
in [31] considers heterogeneous agents for leader-tracking prob-
Notation: For any vector ε = [ε1 , ε2 , . . ., εn ]T , a) sgn(ε) =
lem using dynamic controller, the triggering rule is again a static
[sgn(ε1 ), sgn(ε2 ), . . ., sgn(εn )]T , b) |ε|η = [|ε1 |η , . . ., |εn |η ]T ,
one. The transmission of information is affected by time-varying
and c) |ε|η sgn(ε) = [|ε1 |η sgn(ε1 ), . . ., |εn |η sgn(εn )]T . The pa-
communication delays. The approach in [32] investigates dy-
rameter√η ∈ R. The norm used in the article is defined as
namic event-triggered leader–follower consensus problem with
||ε|| = εT ε.
matched and unmatched disturbances using traditional sliding-
mode controller in finite-time. The approach does not consider
B. System Dynamics and Control Input
HOSMC. Apart from this, an infinite number of switching occurs
in finite time, i.e., Zeno behavior [33], a major issue to be taken The nonholonomic mobile robots in the MAS framework
care of in an event-based technique. To solve this problem, the move on a X–Y plane. Each agent of the group has the same
control protocol should ensure that the interevent intervals have built and dynamics. The kinematic model of any agent i as per
a strictly positive lower bound value. Wang et al. [34] devised Fig. 1 is similar to [35]
an approach for consensus of asynchronously distributed MAS. x̄˙ i = vi cos(θi ), ȳ˙ i = vi sin(θi ), θ̇i = ωi (1)
It is a periodic event-based strategy which effectively removes
Zeno behavior during sampling of data. where x̄i = [x̄i ȳi ]T and θi represent position vector and turn
Motivated by the aforementioned discussion, the work pre- angle, respectively. vi and ωi are linear and angular velocities,
sented in this article aims to design and implement an event- respectively.
triggered STSMC for finite-time robust consensus in MAS An off-axis point Q is considered as the point of operation
prone to disturbances. The controller is applied to a real-time for robot i. The kinematic model (1) is first transformed to
leader–follower MAS framework. In summary, this article has single-integrator form. The off-axis point is located at (xi , yi ),
the following contributions: where xi = x̄i + l cos(θi ) and yi = ȳi + l sin(θi ) with distance

Authorized licensed use limited to: Indian Institute Of Technology (IIT) Mandi. Downloaded on July 12,2023 at 11:22:02 UTC from IEEE Xplore. Restrictions apply.
782 IEEE SYSTEMS JOURNAL, VOL. 15, NO. 1, MARCH 2021

l between center of gravity and Q. The agent dynamics based

on off-axis point is
ẋi = vi cos θi − lωi sin θi
ẏi = vi sin θi + lωi cos θi . (2)
The linear and angular velocities are defined as
vi = uix cos θi + uiy sin θi
ωi = (−1/l)(uix sin θi − uiy cos θi ) (3)
where ui = [uix uiy ]T is the control input vector generating Fig. 2. Block schematic of the proposed approach.
these velocities. On using (3), (2) is simplified as
ẋi = uix ẏi = uiy . (4) Si (t) = [s1 (t), s2 (t), . . . sn (t)]T and qi is the consensus com-
ponent in (9). α ∈ (0, 1) determines the rate of consensus,
The robot is prone to external disturbances (dix , diy ). Hence, γi (γi > 0) is a constant parameter, aij and bi are the connection
the dynamics (4) on incorporating the disturbance becomes weights between the agents, and pi (1 ≤ pi < p) is the number
ẋi = uix + dix ẏi = uiy + diy . (5) of agents neighboring to ith agent. With the onset of sliding
phase, Si (t) = 0. To achieve the desired consensus, we have
The dynamics (5) can be written compactly in the single- considered a super-twisting control law
integrator model form as 1
ũi (t) = qiα (t) − K1 |Si (t)| 2 sgn(Si (t)) + i (t)
ẋi = ui + di (6)
where ˙ i (t) = −K2 sgn(Si (t)). (11)
where xi = [xi , yi ]T and di = [dix , diy ]T represent position
vector and lumped uncertainty for external disturbances and The gain matrices K1 = diag{k11 , k12 , . . . , k1n } and K2 =
unmodeled dynamics, respectively. The control input ui for diag{k21 , k22 , . . . , k2n }, with each of their elements being posi-
agent i is computed using event-based STSMC proposed in tive gains. The control law ũi (t) in (11) is updated only at events
Section III. (described in Section-III) and it remains the same until the next
The leader and follower robot dynamics in MAS framework event triggers. The control law ∀ t (tik ≤ t < tik+1 ) is
1
based on (6) are therefore defined as ũi (t) = qiα (tik ) − K1 |Si (tik )| 2 sgn(Si (tik )) + i (tik ) (12)
ẋi (t) = ui (t) + di (t), i = 1 . . . p tik
where is the latest event instant.
ẋ0 (t) = u0 (t) (7) Remark 1: The superscript (i = 1, . . . , p) and subscript (k =
1, 2, . . .) in tik are agent and event number, respectively. 
n n n
where xi (t) ∈ R , ui (t) ∈ R , and di (t) ∈ R represent po- A novel measurement error (event-triggered error) ei for each
sition, control input, and bounded disturbance, respectively of agent i has been defined to derive the event-triggering condition.
the ith follower robot. x0 (t) ∈ Rn and u0 (t) ∈ Rn are position The event-triggered error is defined as
and control input, respectively, of leader robot. The leader 1 1
dynamics is assumed to be not affected by disturbance. Let ei (t) = |Si (t)| 2 sgn(Si (t)) − |Si (tik )| 2 sgn(Si (tik ))
x̃i (t) = xi (t) − x0 (t) + Δi and ũi (t) = ui (t) − u0 (t) be the ⎡ 1 1
⎤
deviations in position and control input of follower i from the |si1 (t)| 2 sgn(si1 (t)) − |si1 (tik )| 2 sgn(si1 (tik ))
⎢ .. ⎥
leader, respectively. Δi is the desired position separation of =⎢⎣ .
⎥ (13)
⎦
follower i from leader. The relative dynamics of agent i based 1 1
|sin (t)| 2 sgn(sin (t)) − |sin (tik )| 2 sgn(sin (tik ))
on the above deviations and (7) can be presented as
where Si (t) and Si (tik ) are sliding surfaces at current instant (t)
x̃˙ i (t) = ũi (t) + di . (8)
and latest event-instant (tik ), respectively.
The STSMC sliding surface defined for an agent i in the MAS Using (8), (9), and (12), the surface dynamics ∀ t (tik ≤ t <
i
framework is tk+1 ) is
 t 1
Si (t) = x̃i (t) − qiα (t)dt, i = 1 . . . n (9) Ṡi (t) = qiα (tik )−K1 |Si (tik )| 2 sgn(Si (tik )) + i (tik )+di − qiα (t).
0 (14)
where A schematic of the proposed approach is shown in Fig. 2. The
γi  event-based STSMC block first verifies whether event-triggering
qi (t) = − aij {(xi (t) − x0 (t) + Δi ) occurs or not. If yes, it then transmits control input ũi to agent
pi + 1 j∈p
i
i. The ZOH (zero-order hold) block holds the control input
− (xj (t) − x0 (t) + Δj )} + bi (xi (t) − x0 (t) + Δi )) until the next transmission occurs. The sampling time of ZOH
γi  is the same as that for the system operation. The following
=− aij (x̃i (t) − x̃j (t)) + bi x̃i (t). (10) three assumptions have been considered while deriving the
pi + 1 j∈p
i event-based control input.
Authorized licensed use limited to: Indian Institute Of Technology (IIT) Mandi. Downloaded on July 12,2023 at 11:22:02 UTC from IEEE Xplore. Restrictions apply.
NANDANWAR et al.: FINITE-TIME ROBUST ADMISSIBLE CONSENSUS CONTROL OF MULTIROBOT SYSTEM UNDER DYNAMIC EVENTS 783

Assumption 1: For the agent dynamics (7), the unmodeled The time derivative of Lyapunov function in (16) is
dynamics/disturbance lumped together as di is assumed to be
1
bounded, i.e., di ≤ δ|Si | 2 sgn(Si ), where δ ∈ R. V̇i = ζ̇iT P ζi + ζiT P ζ̇i (17)
Assumption 2: The difference Δqi = (qiα (tik ) − qiα (t)) 1
where ζ̇i = [ 12 |Si |− 2 (ϕi + di ) − K2 sgn(Si )]T and ϕi =
∀ t (tik ≤ t < tik+1 ) is bounded as Δqi ≤ Φi where Φi ∈ R. 1

Assumption 3: The event-triggered error magnitude ∀ t (tik ≤ qiα (tik ) − K1 |Si (tik )| 2 sgn(Si (tik )) + i (tik ) − qiα (t). The com-
t < tik+1 ) of any agent i is bounded as ei  ≤ etr . ponent ζ̇iT P ζi in (17) is
Remark 2: The consensus component qi is defined using the
difference of position of agents which change finitely based on ζ̇iT P ζi = 1
2 |Si |
− 12
(ϕi + di ) −K2 sgn(Si )
velocity inputs. Hence, Δqi in Assumption 2 is considered to   1

be bounded in interevent duration. The event-triggered error 1 4K2 + K12 −K1 |Si | 2 sgn(Si )
×
depends on difference of sliding surfaces as per (13). The 2 −K1 2I i
magnitude of this difference is finite for interevent duration and
1 1 − 12
hence validates Assumption 3.  = 2 |Si | (ϕi + di ) −K2 sgn(Si )
2
 1 1

C. Problems 4K2 |Si | 2 sgn(Si ) + K12 |Si | 2 sgn(Si ) − K1 i
1
The solution to the following three problems are integral to the −K1 |Si | 2 sgn(Si ) + 2i
design of the proposed finite-time robust admissible consensus 1 1 1
control in the presence of external bounded disturbance. = |Si |− 2 (ϕi + di )(4K2 |Si | 2 sgn(Si )
4
Problem 1: Given the agent dynamics (7), derive the dy- 1
namic event-triggering condition for each agent i such that the + K12 |Si | 2 sgn(Si ) − K1 i ) + 2K1 K2 |Si |
event-based control input from STSMC achieves consensus and 1
renders the system stable. − 4K2 i |Si | 2 sgn(Si ) . (18)
Problem 2: Guarantee the finite-time consensus by having
finite reaching time to the sliding surface for each agent? Similarly, the component ζiT P ζ̇i of V̇i is
Problem 3: Guarantee the admissibility of event-based control
ζiT P ζ̇i = |Si | 2 sgn(Si ) i
1
law for each agent?
The solutions have been presented in the following section.   
1 − 12
1 4K2 + K12 −K1 2 |S i | (ϕi + d i )
×
III. EVENT-TRIGGERED CONSENSUS CONTROL 2 −K1 2I −K2 sgn(Si )
The proposed methodology has three parts. The first part 1 1 1

derives the dynamic event-triggering condition necessary for = |Si |− 2 (ϕi + di )(4K2 |Si | 2 sgn(Si )
4
updating control inputs of respective agents in the MAS. The 1

second part guarantees finite reaching time to the sliding surface + K12 |Si | 2 sgn(Si ) − K1 i ) + 2K1 K2 |Si |
while the third part guarantees admissibility of event-based 1
− 4K2 i |Si | 2 sgn(Si ) . (19)
control.
Using (18) and (19), (17) can be written as
A. Event-Triggering Condition
1 1 1 1
Theorem 1 presents event-triggering condition required for V̇i = |Si |− 2 ϕi (4K2 |Si | 2 sgn(Si ) + K12 |Si | 2 sgn(Si )
2
control input updates of agents. The approach will ensure con- 1 1
sensus among agents and their stability. − K1 i ) + di (4K2 |Si | 2 sgn(Si ) + K12 |Si | 2 sgn(Si )
Theorem 1: Consider the dynamics of agent i (7) with its 1

event-based control law (12). If the condition − K1 i ) + 2K1 K2 |Si | − 4K2 i |Si | 2 sgn(Si ) . (20)
1
ei  > z2 − |Si (tik )| 2 sgn(Si (tik )) On applying disturbance bound as per Assumption 1 in (20)
1 1 1 1
−κ + κ2 − 4βρ V̇i ≤ |Si |− 2 4K2 ϕi |Si | 2 sgn(Si ) + K12 ϕi |Si | 2 sgn(Si )
where z2 = (15) 2
2β 1
− K1 ϕi i + 4K2 δ|Si | + K12 δ|Si | − K1 i δ|Si | 2 sgn(Si )
is satisfied, then an event triggers and control input for the
1
agent i is updated at that instant. The gradual events result + 2K1 K2 |Si | − 4K2 i |Si | 2 sgn(Si ) . (21)
in its consensus with other agents of MAS in the presence of
disturbance and further ensure stability. 1
Let zi = |Si | 2 sgn(Si ). On expressing (21) in terms of zi
Proof: The Lyapunov function considered for the agent i is
1 1 
Vi = ζiT Pi ζi (16) V̇i ≤ |Si |− 2 4K2 ϕi zi + K12 ϕi zi − K1 ϕi i + 4K2 δzi2
2
1 
+ K12 δzi2 − K1 i δzi + 2K1 K2 zi2 − 4K2 i zi
2
where ζi = [ |Si | sgn(Si )
and Pi = 12 [ 4K−K
2 +K1 −K1
2
i ] 1 2I ].

Authorized licensed use limited to: Indian Institute Of Technology (IIT) Mandi. Downloaded on July 12,2023 at 11:22:02 UTC from IEEE Xplore. Restrictions apply.
784 IEEE SYSTEMS JOURNAL, VOL. 15, NO. 1, MARCH 2021

1 1  Theorem 2: For the triggering condition (15) and the event-

= |Si |− 2 zi2 (4K2 δ + K12 δ + 2K1 K2 ) + zi (4K2 ϕi
2 based control law (12), the reaching time Tr to the sliding surface
 is finite and upper bounded by
+ K12 ϕi − K1 i δ − 4K2 i ) + (−K1 ϕi i )

1 1   2  1
= |Si |− 2 βzi2 + κzi + ρ (22) Tr =  − |Si (tS0 )| 2 sgn(Si (tS0 ))
2 δ
ςi   
where β = 4K2 δ + K12 δ + 2K1 K2 , κ = 4K2 ϕi + K12 ϕi − 1 
− ln −δ|Si (tS0 )| sgn(Si (tS0 ))  .
2 (30)
K1 i δ − 4K2 i , and ρ = −K1 i ϕi . For stability, V̇i ≤ 0. δ
Hence
Proof: From (9), the sliding surface dynamics can be given
1 1  
|Si |− 2 βzi2 + κzi + ρ ≤ 0 by
2
Ṡi = x̃˙ i − qiα . (31)
⇒ βzi2 + κzi + ρ ≤ 0 ⇒ (zi − z1 )(zi − z2 ) ≤ 0
Equation (31) can be further written using (8) as
−κ ± κ2 − 4βρ
where (z1 , z2 ) = . (23)
2β Ṡi = ũi + di − qiα . (32)
Remark 3: For z1 and z2 to have real values, κ2 − 4βρ ≥ 0. On using the control input ũi from (12) at t = tS0 in (32)
This is ensured by the parameters considered for the agents. Let
1
for agent i, z1 ≤ z2 .  Ṡi = qiα (tS0 )−K1 |Si (tS0 )| 2 sgn(Si (tS0 ))+i (tS0 )+di −qiα
Since z1 ≤ z2 , the sufficient condition for system stability is (33)
(zi − z2 ) ≤ 0 ⇒ zi ≤ z2 . (24) where Si (tS0 ) is the initial value (at t = tS0 ) when agent i starts
toward the sliding surface. As per Assumptions 1 and 2, (33)
On substituting the expression for zi in (24)
can be expressed as
1
|Si | 2 sgn(Si ) ≤ z2 . (25) 1
Ṡi ≤ Φi − K1 |Si (tS0 )| 2 sgn(Si (tS0 )) + i (tS0 )
From (13) 1

1 1
+ δ|Si | 2 sgn(Si ). (34)
|Si | sgn(Si ) = ei +
2 |Si (tik )| 2 sgn(Si (tik )). (26) 1
Let ςi = Φi − K1 |Si (tS0 )| 2 sgn(Si (tS0 )) + i (tS0 ) for sim-
On rewriting (25) using (26) plicity of expression. Inequality (34) can be written as
1
ei + |Si (tik )| 2 sgn(Si (tik )) ≤ z2 dSi 1
Ṡi = ≤ δ|Si | 2 sgn(Si ) + ςi . (35)
1 dt
⇒ ei ≤ z2 − |Si (tik )| sgn(Si (tik )).
2 (27)
The maximum rate of change of sliding surface is
The system is stable for the condition obtained in (27). The above
condition can be further written as dSi 1
= δ|Si | 2 sgn(Si ) + ςi . (36)
1 dt
ei  ≤ z2 − |Si (tik )| 2 sgn(Si (tik )) . (28)
Remark 6: Equation (36) has vector components on the left-
Remark 4: The components on both sides of (27) are vector and right-hand sides. To proceed further, L2 norm has been
quantities. Hence, a simplified form is written in (28).  used. 
On the contrary, an event is triggered when The maximum reaching time can therefore be obtained from
(36) as
1
ei  > z2 − |Si (tik )| 2 sgn(Si (tik )) . (29)  Si (tS )  tS
f dSi f

1 = dt (37)
Equation (29) is therefore the event-triggering condition for Si (tS0 ) δ|Si | sgn(Si ) + ςi
2 tS 0
agent i. All the agents will have similar condition like (29).
Remark 5: (a) Equation (29) presents a dynamic condition where Si (tSf ) is the value (at t = tSf ) when agent i reaches the
as the parameters on right-hand side (RHS) change value at sliding surface.
different time instants. (b) For smooth implementation of (29) Remark 7: The S(tS0 ) and S(tSf ) are values at time instants
in real-time scenario, the “>” symbol is replaced by “≥.”  tS0 and tSf when describing point of ith agent initially starts
toward and finally reaches the sliding surface, respectively. 
B. Finite-Time Consensus Control On simplifying expression (37)

For real-time consensus applications, the sliding phase should 2  1
Si (tS )
 f

start in finite time, or, in other words, reaching time to the sliding |Si | 2 sgn(Si )
δ Si (tS0 )
surface should be finite. The describing point of an agent then  Si (tS )  
moves along its sliding surface. Theorem 2 guarantees finite- ςi 1  f 
− ln δ|Si | 2 sgn(Si ) + ςi   = tSf − tS0 . (38)
time consensus control by STSMC. δ Si (tS0 )

Authorized licensed use limited to: Indian Institute Of Technology (IIT) Mandi. Downloaded on July 12,2023 at 11:22:02 UTC from IEEE Xplore. Restrictions apply.
NANDANWAR et al.: FINITE-TIME ROBUST ADMISSIBLE CONSENSUS CONTROL OF MULTIROBOT SYSTEM UNDER DYNAMIC EVENTS 785

As the agent reaches sliding surface, the value of Si , i.e., Using (42) and (46) is written in terms of event-triggered error
S(tSf ) = 0. Equation (38) is further expressed as as

2  1
d(ei ) 1
 
 − |Si (tS0 )| 2 sgn(Si (tS0 )) ėi = i − 12
≤ |Si (tk )| i 12 i
δ(ei +|Si (tk )| sgn(Si (tk )))+ςi
δ dt 2
ςi    ⎡  ⎤
1 
− ln −δ|Si (tS0 )| sgn(Si (tS0 ))  = (tSf − tS0 ) (39)
2 1 i i − 12 i i i 12 i i
δ ⎢ 2 |s1 (tk )| δ(e1 + |s1 (tk )| sgn(s1 (tk ))) ⎥
⎢  ⎥
where tSf − tS0 = Tr . On rearranging (39) ⎢ ⎥
⎢ i
+ ς1 ⎥
 ⎢ ⎥
2  1 d(ei ) ⎢⎢
..
⎥
⎥
Tr =  − |Si (tS0 )| 2 sgn(Si (tS0 )) ⇒ ≤⎢ . ⎥.
δ dt ⎢  ⎥
⎢ ⎥
ςi    ⎢ 1 |si (ti )|− 12 δ(ei + |si (ti )| 12 sgn(si (ti )))⎥
1  ⎢2 n k n n k n k ⎥
− ln −δ|Si (tS0 )| 2 sgn(Si (tS0 ))  . (40) ⎢  ⎥
δ ⎣ ⎦
i
+ ςn
The Tr in (40) is the maximum reaching time to the sliding
surface. (47)

C. Admissibility of Event-Based Control Using the vector norm in (47), we can write
The admissibility of event-triggered control is necessary to  
d ei  1 i − 12 i 12 i
avoid the piling of triggering instants (i.e., Zeno effect) which ≤ |Si (tk )| δ(ei +|Si (tk )| sgn(Si (tk )))+ςi
can force the system to instability. Theorem 3 provides a bound dt 2
on interevent execution duration to avoid such scenario.  
1 1 1
Theorem 3: For the system in (7), event-based control is ≤ |Si (tik )|− 2 δ ei  + δ|Si (tik )| 2 sgn(Si (tik ))+ςi .
updated when the condition in (15) is satisfied. The admissibility 2
of such control update is guaranteed if the interevent execution (48)
time Te (Te = tik+1 − tik ) is lower bounded by
  From (48)
i 12 δetr
Te ≥ 2 |Si (tk )| ln 1 + .  etr
1
δ|Si (tik )| 2 sgn(Si (tik )) + ςi d ei 
1
(41) 0 δ ei  + δ|Si (tik )| 2 sgn(Si (tik )) + ςi
Proof: The event-triggered error for agent i is
1 1  tik+1
ei = |Si (t)| 2 sgn(Si (t)) − |Si (tik )| 2 sgn(Si (tik )). (42) ≤
1 1
|Si (tik )|− 2 dt
tik 2
The time derivative of event-triggered error in (42) is
  ⎛ ⎞
d 1
i 12 i 1
δ ei  + δ|Si (tik )| 2 sgn(Si (tik )) + ςi etr
ė1 = |Si (t)| sgn(Si (t)) − |Si (tk )| sgn(Si (tk ))
2

dt ⇒ ln ⎝ ⎠
δ 
0
1 1
= |Si |− 2 Ṡi . (43)
2 tik+1
1 
Substituting Ṡi from (32) in (43) gives ≤ i − 12
|Si (tk )| t . (49)
2 ti
1 1 1 k
ėi = |Si |− 2 (qiα (tik ) − K1 |Si (tik )| 2 sgn(Si (tik ))
2
As per Assumption 3, ei  ≤ etr . On simplifying and
+ i (tik ) + di − qiα ). (44) rearranging (49)
Using Assumptions 1 and 2 in (44)
 1 1
1 − 12 1 |Si (tik )|− 2 (tik+1 − tik )
ėi ≤ |Si | Φi − K1 |Si (tik )| 2 sgn(Si (tik )) 2
2 ⎛ 1
⎞
 δetr + δ|Si (tik )| 2 sgn(Si (tik )) + ςi
1
+ i (tik ) + δ|Si | 2 sgn(Si ) . (45) ≥ ln ⎝ 1
⎠
δ|Si (tik )| 2 sgn(Si (tik )) + ςi
1 ⎛ ⎞
Using ςi = Φi − K1 |Si (tik )| 2 sgn(Si (tik )) + i (tik ) the same as
1 δetr
in Section III-B, inequality (45) is written as ⇒ Te ≥ 2 |Si (tik )| 2 ln ⎝1+ ⎠
  1
δ|Si (tk )| 2 sgn(Si (tik ))+ςi
i
1 − 12 1
ėi ≤ |Si | δ|Si | sgn(Si ) + ςi .
2 (46) (50)
2

Authorized licensed use limited to: Indian Institute Of Technology (IIT) Mandi. Downloaded on July 12,2023 at 11:22:02 UTC from IEEE Xplore. Restrictions apply.
786 IEEE SYSTEMS JOURNAL, VOL. 15, NO. 1, MARCH 2021

Fig. 3. MAS digraph for (a) simulation and (b) real-time experiment.

Fig. 4. Trajectory of robots during desired formation.

where Te = tik+1 − tik . The value of Te obtained in (50) provides
lower bound on interexecution time before the next event occurs.
Remark 8: The interevent time is defined using the last (tik )
and forthcoming event instant (tik+1 ). Hence, Si at t = tik and
t = tik+1 are considered. 
Remark 9: Equation (47) has vector components on LHS and
RHS. To proceed further, the vector norm has been used. 
Remark 10: The value of Te varies for different interevent
duration as the parameters involved change their values. 

IV. CASE STUDY

The proposed methodology has been validated through sim- Fig. 5. (a) Linear velocity and (b) angular velocity of robots.
ulation and real-time experiments.
Remark 11: In the experiments (simulation and real-time),
we have considered random data packet losses and delays in the consensus parameters are
network. The data communication among the mobile robots in ⎡ ⎤ ⎡ ⎤
real-time experiment occurs through wireless network following 1 0 −1 0 0 1 0 0 0 0
⎢ ⎥ ⎢ ⎥
IEEE 802.11 protocol which is susceptible to the above network ⎢0 0 0 0 0⎥ ⎢0 1 0 0 0⎥
⎢ ⎥ ⎢ ⎥
 ⎢
L = ⎢ 0 −1 1 0 0⎥ ⎢ 0⎥
uncertainties. ⎥ B = ⎢0 0 0 0 ⎥.
Remark 12: There are some limitations during implemen- ⎢ ⎥ ⎢ ⎥
⎣0 0 −1 1 0⎦ ⎣0 0 0 0 0⎦
tations. In the experiments, only time-varying matched distur-
−1 0 0 0 1 0 0 0 0 0
bances are considered. Only random packet losses and delays
are considered as wireless network (following IEEE 802.11 The leader is free to move in any direction. The initial position
protocol) is used, which is susceptible to network uncertainties. of leader is x0 = [0; 0], whereas followers are initially at
There is no separate fault identification and isolation module in x1 = [−2; 1], x2 = [−3; 2], x3 = [−2; −1], x4 = [−3; −2],
the proposed model.  and x5 = [−4; 0]. The system parameters are as follows: l =
0.21 m, K1 = diag{0.5, 0.5}, K2 = diag{0.001, 0.001}, Δ1 =
(−0.8 m, −0.8 m), Δ2 = (0.8 m, −0.8 m), Δ3 = (−0.8 m,
A. Simulation Experiment
0.8 m), Δ4 = (0.8 m, 0.8 m), Δ5 = (0.8 m, 0 m), and α = 5/7.
The approach has been validated through simulation in The parameter δ in disturbance bound as per Assumption 1 is
Gazebo environment with Pioneer P3-DX robot model. In the 0.01.
MAS framework shown in Fig. 3 (a), five robots act as fol- The task is to achieve the shape of Gerono lemniscate (hori-
lower and one acts as leader. All robot models are identical in zontally oriented eight-shaped figure). The leader and follower
performance and control. Each robot system is equipped with robots’ trajectories are shown in Fig. 4. The leader pursues the
an onboard computer with Ubuntu 14 platform having robot trajectory-shaped Gerono lemniscate whereas followers follow
operating system (ROS), which facilitates an ultimate control the leader trajectory maintaining a desired distance. Initially,
over all sensors and actuators. A Wi-Fi router (Tinda) is used the robots are not in consensus but after few seconds, they
to maintain a sufficient network strength throughout the experi- gradually attain the desired trajectory. Fig. 5 shows the linear
ment. Each robot is provided a unique IP address for exchange and angular velocities of followers with respect to the leader. It
of information with each other. The odometer sensor provides can be observed that the velocities of followers gradually get in
current position and orientation of mobile robot. The sensor data sync with the leader velocities. Fig. 6(a) ensures that the desired
is available at the rate of 10 Hz. Hence, 0.01 s is used as sampling separation is gradually maintained throughout the task even in
interval. the presence of disturbances. The sliding surfaces of all the
The weight for the communication among agents is assumed follower agents are shown in Fig. 6(b). All the agents attain their
to be 1. As per the communication graph G in Fig. 3(a), the respective desired sliding surface in finite-time. The dynamic

Authorized licensed use limited to: Indian Institute Of Technology (IIT) Mandi. Downloaded on July 12,2023 at 11:22:02 UTC from IEEE Xplore. Restrictions apply.
NANDANWAR et al.: FINITE-TIME ROBUST ADMISSIBLE CONSENSUS CONTROL OF MULTIROBOT SYSTEM UNDER DYNAMIC EVENTS 787

Fig. 6. (a) Relative distance of followers from leader and (b) sliding surface Fig. 8. (a) Trajectory of robots during desired formation. (b) Relative distance
of followers. of followers from leader.

Fig. 7. Event-triggered error and threshold of (a) F-1 and (b) F-3. Fig. 9. (a) Linear velocity and (b) angular velocity of robots.

TABLE I
EVENTS TRIGGERED FOR FOLLOWER ROBOTS

F: Follower.

event-triggered errors and their thresholds for F-1 and F-3 are
Fig. 10. Sliding surfaces of two followers.
shown in Fig. 7. The other agents too provide the similar results.
The events occur when errors cross their respective dynamic
thresholds. It can be observed that events are more when there
are turns or change in orientations of the trajectory. Table I shows sampling time considered for the experiments is 0.01 s. The
the number and percentage of events triggered for all the five value of δ in disturbance bound is 0.01.
follower robots during consensus in 100 s. The trajectories of all the robots are shown in Fig. 8(a). The
desired formation is achieved for leader-follower framework as
the desired deviation Δ is obtained in Fig. 8(b). The distance
B. Real-Time Experiments
between each follower and leader is large initially but after
Three Pioneer P3-DX (two-wheeled differentially driven) some time event-based controllers drive the agents to a desired
robots have been used for real-time experiments which form a formation and then maintain required separations for rest of the
MAS framework. The two robots are followers while one acts as trajectory. The linear and angular velocities of the robots are
a leader. The robots are equipped with sonar sensor and position shown in Fig. 9(a) and (b), respectively. The proposed controller
encoder. They communicate through wireless network having has been tested for various velocities and turn maneuvers. The
unique IP addresses assigned to them. sliding surfaces (S) of the followers are presented in Fig. 10.
The consensus parameters of the graph are L = [0 0; −1 1] The result suggests that the describing point of agents reach the
and B = [1 0; 0 1]. The leader robot is free to roam in sliding surfaces approximately in 12.3 s. The expression of upper
any direction. To achieve consensus-based formation, bound on finite reaching time in (40) evaluates to 115.94 s. Thus,
the followers should keep track of leader trajectory. the reaching time during real-time experiment is well below the
The initial position of leader is x0 = [0; 0], whereas upper bound. The event-triggered error and the respective error
followers are initially at x1 = [−4; 0.5] and x2 = [−4; threshold as per condition in (15) for both the followers are
−0.5]. The system parameters are as follows: l = shown in Fig. 11. The event-triggered error crosses the threshold
0.21 m, K1 = diag{0.5, 0.5}, K2 = diag{0.001, 0.001}, Δ1 = very frequently initially which leads to triggering of events for
(0.7 m, −0.7 m), Δ2 = (0.7 m, 0.7 m), and α = 5/7. The control update. As the consensus is gradually achieved, the

Authorized licensed use limited to: Indian Institute Of Technology (IIT) Mandi. Downloaded on July 12,2023 at 11:22:02 UTC from IEEE Xplore. Restrictions apply.
788 IEEE SYSTEMS JOURNAL, VOL. 15, NO. 1, MARCH 2021

Fig. 11. Event-triggered error and threshold of (a) Follower-1 and Fig. 14. Results based on [18]. (a) Trajectory of robots. (b) Relative distance
(b) Follower-2. of followers from leader.

Fig. 12. (a) Trajectory of robots during circle formation. (b) Relative distance Fig. 15. Velocities based on approach in [18]. (a) Linear velocity and
of followers from leader. (b) angular velocity.

in desired formation shape and maintains a desired separation

distance throughout the process with respect to the leader. It is
evident from the plots that formation in MAS is achieved in
finite time in the presence of time-varying matched disturbance,
which infers the robustness of controller. The above experiment
confirms the efficacy of the proposed approach. The consensus
is achieved within 5–7 s. A YouTube link of the video for
Fig. 13. (a) X-position trajectories and (b) Y-position trajectories of robots
during circle formation with desired separation distance.
experiments with three other sets of initial positions is given
in [36].
crossings by event-triggered error is largely reduced. The least
interevent time is 0.01s. C. Comparative Analysis With Other Approaches
For the validation of the proposed approach, another real-time To compare the efficiency of the proposed approach,
experiment is performed with leader and follower robots trying we have repeated the experiments for approach presented
to achieve the desired formation in the shape of “circle.” The in [18]. The control law used there is defined as ui (t) =
initial position of leader is x0 = [0; 0]. The followers are initially γi qiα (tik ). The triggering condition used in [18] is ei (t) ≤
at x1 = [−1; −0.7] and x2 = [1; −0.7]. The desired distance (η/n(1−β)/2 )1/β qi (ti )α/β , where event-triggered error mag-
separation between leader and followers 1 and 2 should be nitude is ei (t) = (qiα (tik ) − qiα (ti ))1/β . The other param-
Δ1 = (0.9 m, 0 m) and Δ2 = (−0.9 m, 0 m), respectively. The eters are α ∈ (0, 1) and β ∈ (0, 1]. The initial position of
leader robot’s linear velocity (v) and angular velocity (ω) is leader is x0 = [0 ; 0]T , whereas followers are initially at x1 =
0.2 m/s and 0.15 rad/s. The consensus parameter, system pa- [−4 ; 0.5]T and x2 = [−4 ; −0.5]T .
rameter, and disturbance are similar to the previous experiment. On comparing the trajectories during desired formation
The leader and follower robots trajectories are shown in in Fig. 14, the proposed approach provides better trajectory
Fig. 12(a). The leader pursues a circle-shaped trajectory with the than [18]. One can notice in Fig. 15 the linear and angular
followers following the leader trajectory maintaining a desired velocities of robots change abruptly at times for [18], whereas the
distance. Initially, the robots are not in consensus but after few change in velocities are smooth for the proposed approach. For
seconds, they gradually attain the desired trajectory. Fig. 12(b) the generalized consensus-based problem, the approach in [18]
ensures that the desired separation is gradually maintained works well but not suitable for MAS with disturbance.
throughout the task even in the presence of disturbances. A comparison has been drawn out with conventional SMC
Fig. 13 shows that the followers are achieving consensus with approach presented in [32]. The event-triggering condition and
respect to the leader in finite time with the desired separation. the control law have been used as given in [32]. The robot
The square block in Fig. 13(a) shows that the follower comes parameters are kept the same as in this article. The trajectories

Authorized licensed use limited to: Indian Institute Of Technology (IIT) Mandi. Downloaded on July 12,2023 at 11:22:02 UTC from IEEE Xplore. Restrictions apply.
NANDANWAR et al.: FINITE-TIME ROBUST ADMISSIBLE CONSENSUS CONTROL OF MULTIROBOT SYSTEM UNDER DYNAMIC EVENTS 789

than other existing event-based approaches. The less number

of events significantly decrease the count of control updates
which further reduce actuator actions, computations, and data
communication required to address the problem. This ensures
saving of resources available during multirobot system opera-
tion. However, this work does not consider network uncertainties
among the agents. These uncertainties may be in the form of
transmission delay, communication delay, packet loss, channel
noise, etc. They may deteriorate the system performance from
the desired objective and may even cause the system to be un-
Fig. 16. Results based on [32]. (a) Trajectory of robots. (b) Relative distance stable. Another future scope of this article may be to incorporate
of followers from leader.
fault tolerance in the existing model to handle the fault in the
exciting model. The future work, therefore, will consider the
above-mentioned aspects in the proposed approach.

REFERENCES
[1] W. Ren and R. W. Beard, Distributed Consensus in Multi-Vehicle Coop-
erative Control. London, U.K.: Springer-Verlag, 2008.
[2] Y. Toda and N. Kubota, “Self-localization based on multiresolution map
for remote control of multiple mobile robots,” IEEE Trans. Ind. Inform.,
vol. 9, no. 3, pp. 1772–1781, Aug. 2013.
[3] D. Zhang, Z. Xu, H. R. Karimi, Q.-G. Wang, and L. Yu, “Distributed H∞
output-feedback control for consensus of heterogeneous linear multiagent
Fig. 17. Velocities based on approach in [32]. (a) Linear velocity and systems with aperiodic sampled-data communications,” IEEE Trans. Ind.
(b) angular velocity. Electron., vol. 65, no. 5, pp. 4145–4155, May 2018.
[4] H. Sayyaadi and M. Doostmohammadian, “Finite-time consensus in di-
rected switching network topologies and time-delayed communications,”
TABLE II Scientia Iranica, vol. 18, no. 1, pp. 75–85, 2011.
COMPARISON OF DIFFERENT APPROACHES [5] A. Mondal, C. Bhowmick, L. Behera, and M. Jamshidi, “Trajectory
tracking by multiple agents in formation with collision avoidance and
connectivity assurance,” IEEE Syst. J., vol. 12, no. 3, pp. 2449–2460,
Sep. 2018.
[6] A. Jenabzadeh and B. Safarinejadian, “Distributed tracking of nonholo-
nomic targets over multiagent systems,” IEEE Syst. J., vol. 13, no. 2,
pp. 1678–1681, Jun. 2019.
[7] Y. Koren and J. Borenstein, “Potential field methods and their inherent
F-1: Follower-1 and F-2: Follower-2. limitations for mobile robot navigation,” in Proc. IEEE Int. Conf. Robot.
Autom., Apr. 1991, pp. 1398–1404.
[8] N. K. Dhar, N. K. Verma, and L. Behera, “Adaptive critic-based event-
during formation in Fig. 16 are shaky as with the linear and triggered control for HVAC system,” IEEE Trans. Ind. Informat., vol. 14,
no. 1, pp. 178–188, Jan. 2018.
angular velocities in Fig. 17 too showing chattering. Though [9] A. K. Kar, N. K. Dhar, and N. K. Verma, “Event-triggered adaptive
the minimum interevent time is the same for all the approaches neural network controller in a cyber-physical framework,” IEEE Trans.
in Table II, the average triggering is lesser for the proposed Ind. Inform., vol. 15, no. 4, pp. 2101–2111, Apr. 2019.
[10] T. Henningsson, E. Johannesson, and A. Cervin, “Sporadic event-based
approach than the approaches in [18] and [32]. control of first-order linear stochastic systems,” Automatica, vol. 44, no. 11,
Based on the theoretical developments followed by real-time pp. 2890–2895, 2008.
experiments, we can conclude that the proposed event-based [11] A. Anta and P. Tabuada, “To sample or not to sample: Self-triggered
control for nonlinear systems,” IEEE Trans. Autom. Control, vol. 55, no. 9,
STSMC approach is one of the best choices to handle consensus- pp. 2030–2042, Sep. 2010.
based formation problem in the presence of disturbance. [12] D. V. Dimarogonas, E. Frazzoli, and K. H. Johansson, “Distributed event-
triggered control for multi-agent systems,” IEEE Trans. Autom. Control,
vol. 57, no. 5, pp. 1291–1297, May 2012.
V. CONCLUSION [13] A. Wang, B. Mu, and Y. Shi, “Event-triggered consensus control for
multiagent systems with time-varying communication and event-detecting
This article designs a finite-time event-based STSMC for delays,” IEEE Trans. Control Syst. Technol., vol. 27, no. 99, pp. 507–515,
achieving consensus in a leader–follower-based MAS frame- Mar. 2019.
work in the presence of external bounded disturbance. The pro- [14] B. Mu, K. Zhang, F. Xiao, and Y. Shi, “Event-based rendezvous control
for a group of robots with asynchronous periodic detection and commu-
posed approach guarantees finite-time consensus of each agent nication time delays,” IEEE Trans. Cybern., no. 99, pp. 1–10, Jul. 2018.
along with its stability. The admissibility criteria satisfied by [15] H. Li, X. Liao, T. Huang, and W. Zhu, “Event-triggering sampling based
each agent’s control update helps in avoiding the Zeno behavior leader-following consensus in second-order multi-agent systems,” IEEE
Trans. Autom. Control, vol. 60, no. 7, pp. 1998–2003, Jul. 2015.
during the consensus. The simulation and real-time experiments [16] W. Zhu and Z.-P. Jiang, “Event-based leader-following consensus of multi-
using six and three robots, respectively, for desired consensus agent systems with input time delay,” IEEE Trans. Autom. Control, vol.
validate the theoretical developments in the article. The results 60, no. 5, pp. 1362–1367, May 2015.
[17] M. Yu, C. Yan, D. Xie, and G. Xie, “Event-triggered tracking consensus
obtained justify the performance of the proposed approach to with packet losses and time-varying delays,” IEEE/CAA J. Autom. Sinica,
be at par with traditional time-triggered approaches and better vol. 3, no. 2, pp. 165–173, Apr. 2016.

Authorized licensed use limited to: Indian Institute Of Technology (IIT) Mandi. Downloaded on July 12,2023 at 11:22:02 UTC from IEEE Xplore. Restrictions apply.
790 IEEE SYSTEMS JOURNAL, VOL. 15, NO. 1, MARCH 2021

[18] Y. Zhu, X. Guan, X. Luo, and S. Li, “Finite-time consensus of multi-agent Narendra Kumar Dhar received the master’s degree
system via nonlinear event-triggered control strategy,” IET Control Theory in control system from IIT Roorkee, Roorkee, India,
Appl., vol. 9, no. 17, pp. 2548–2552, Nov. 2015. in 2013. He is currently working toward the Ph.D. de-
[19] M. Ou, H. Du, and S. Li, “Finite-time formation control of multiple gree with the Department of Electrical Engineering,
nonholonomic mobile robots,” Int. J. Robust Nonlinear Control, vol. 24, IIT Kanpur, Kanpur, India.
no. 1, pp. 140–165, 2014. His research interests include cyber-physical sys-
[20] Q. Liu, Z. Wang, X. He, and D.-H. Zhou, “Event-based H-infinity con- tems, smart home automation, and softcomputing.
sensus control of multi-agent systems with relative Output feedback:
The finite-horizon case,” IEEE Trans. Autom. Control, vol. 60, no. 9,
pp. 2553–2558, Sep. 2015.
[21] A. Levant, “Sliding order and sliding accuracy in sliding mode control,”
Int. J. Control, vol. 58, no. 6, pp. 1247–1263, 1993.
[22] J. A. Moreno and M. Osorio, “A Lyapunov approach to second-order
sliding mode controllers and observers,” in Proc. 47th IEEE Conf. Decision
Control, Dec. 2008, pp. 2856–2861. Dmitry Malyshev graduated from Belgorod State
[23] Y. Zhang, Y. Yang, Y. Zhao, and G. Wen, “Distributed finite-time tracking Technological University named after V.G. Shukhov
control for nonlinear multi-agent systems subject to external disturbances,” in 2015, specializing in mechanical engineering
Int. J. Control, vol. 86, no. 1, pp. 29–40, 2013. technology.
[24] S. Yu and X. Long, “Finite-time consensus for second-order multi-agent Since 2015, he has been working at the Center
systems with disturbances by integral sliding mode,” Automatica, vol. 54, for High Technologies at Belgorod State Technolog-
pp. 158–165, 2015. ical University named after V.G. Shukhov. His re-
[25] N. Liu, R. Ling, Q. Huang, and Z. Zhu, “Second-order super-twisting slid- search interests include a wide range of issues related
ing mode control for finite-time leader-follower consensus with uncertain to numerical optimization and simulation methods
nonlinear multiagent systems,” Math. Problems Eng., vol. 2015, pp. 1–8, as applied to robots and manipulators of various
2015. structures.
[26] H. Yang, Y. Jiang, and S. Yin, “Fault-tolerant control of time-delay markov
jump systems with Itô stochastic process and output disturbance based
on sliding mode observer,” IEEE Trans. Ind. Inform., vol. 14, no. 12,
pp. 5299–5307, Dec. 2018.
[27] H. Yang and S. Yin, “Actuator and sensor fault estimation for time-delay
Markov jump systems with application to wheeled mobile manipulators,” Larisa Rybak graduated with honors from the
IEEE Trans. Ind. Inform., vol. 16, no. 5, pp. 3222–3232, May 2020. Kharkov Aviation Institute, in 1985. She received
[28] H. Yang and S. Yin, “Reduced-order sliding-mode-observer-based fault es- the Ph.D. degree in the theory of mechanisms and
timation for Markov jump systems,” IEEE Trans. Autom. Control, vol. 64, machines from the Moscow Academy of Instrument
no. 11, pp. 4733–4740, Nov. 2019. Engineering, Russia, in 1992 and the Ph.D. Hab.
[29] A. K. Behera and B. Bandyopadhyay, “Event-triggered sliding mode degree in 1998.
control for a class of nonlinear systems,” Int. J. Control, vol. 89, no. 9, From 1985 to 1994, she worked with the Moscow
pp. 1916–1931, 2016. State Open University as an Assistant Professor. From
[30] R. R. Nair, L. Behera, and S. Kumar, “Event-triggered finite-time inte- 1994 to 2002, she worked with the Mechanical En-
gral sliding mode controller for consensus-based formation of multirobot gineering Research Institute of the Russian Academy
systems with disturbances,” IEEE Trans. Control Syst. Technol., vol. 27, of Sciences. Since 2004, she has been working with
no. 1, pp. 39–47, Jan. 2019. the Belgorod State Technological University, named after V.G. Shukhov, as a
[31] E. Garcia, Y. Cao, and D. W. Casbeer, “An event-triggered control approach Professor and Head of the Scientific Laboratory of Mechatronics and Robotics.
for the leader-tracking problem with heterogeneous agents,” Int. J. Control, Her research interests include mechanics of machines and robots, intelligent
vol. 91, no. 5, pp. 1209–1221, 2018. control systems for machines and robots, mathematical simulation, optimization,
[32] A. Sinha and R. K. Mishra, “Consensus in first order nonlinear heteroge- and computational methods.
neous multi-aagent systems with event-based sliding mode control,” Int.
J. Control, vol. 93, no. 4, pp. 1–14, 2018.
[33] A. Selivanov and E. Fridman, “Event-triggered H∞ control: A switching
approach,” IEEE Trans. Autom. Control, vol. 61, no. 10, pp. 3221–3226,
Oct. 2016.
[34] A. Wang, B. Mu, and Y. Shi, “Consensus control for a multi-agent system
with integral-type event-triggering condition and asynchronous periodic Laxmidhar Behera (Senior Member, IEEE) received
detection,” IEEE Trans. Ind. Electron., vol. 64, no. 7, pp. 5629–5639, the B.Sc. and M.Sc. degrees in engineering from,
Jul. 2017. NIT Rourkela, Rourkela, India, in 1988 and 1990,
[35] W. Ren and N. Sorensen, “Distributed coordination architecture for multi- respectively, and the Ph.D. degree from IIT Delhi,
robot formation control,” Robot. Auton. Syst., vol. 56, no. 4, pp. 324–333, New Delhi, India, in 1996. He pursued the postdoc-
2008. toral studies in the German National Research Center
[36] Finite-Time Robust Admissible Consensus Control of Multi-Robot for Information Technology, GMD, Sank Augustin,
System Under Dynamic Events. [Online]. Available: https://youtu.be/ Germany from 2000–2001.
XY7LrDOX1sg From 1995 to 1999, he has worked as an Assistant
Professor with BITS Pilani and as a Reader with Intel-
ligent Systems Research Center, University of Ulster,
UK, form 2007 to 2009. He has also worked as a Visiting Researcher/Professor
Anuj Nandanwar received the master’s degree
with FHG, Germany, and ETH, Zurich. He is currently working as Poonam
in electronics and telecommunication from the
and Prabhu Goel Chair Professor with IIT Kanpur, Kanpur, India, having
Chhattisgarh Swami Vivekanand Technical Univer- research and teaching experience of more than 24 years. His primary research
sity, Bhilai, India. He is currently working toward
interests include the convergence of machine learning, control theory, robotic
the Ph.D. degree with the Department of Electrical
vision and heterogeneous robotic platforms. His other research interests include
Engineering, IIT Kanpur, Kanpur, India.
intelligent control, semantic signal/music processing, neural networks, control
His research interests include cyber-physical sys- of cyber-physical systems and cognitive modelling.
tems, cooperative control of multirobot systems,
Dr. Behera has established industrial collaboration with TCS, Renault Nissan,
sliding-mode control, and event-trigger control.
BEL, Bangalore and ADNOC, Abu Dhabi. He is a Fellow of INAE.

Authorized licensed use limited to: Indian Institute Of Technology (IIT) Mandi. Downloaded on July 12,2023 at 11:22:02 UTC from IEEE Xplore. Restrictions apply.