Professional Documents
Culture Documents
Abstract- Coordination between a specific mobile robot types has tracking or path following for omni wheel mobile robots can be
been widely investigated, e.g coordination between unicycles. To found for example in [7]- [9].
extend the applicability of the system, coordinated trajectory
In [7], a kinematic control is built for the omnidirectional
tracking of mixed type of mobile robots is considered. We extend
and prove the concept that if a certain type of wheeled mobile robot. A detailed kinematic model is implemented to solve a
robot is able to individually track its own references, then trajectory tracking problem. The controller takes into account
coordination in tracking with other type of robots can be achieved kinematic model of the wheels to reduce slip effects. In [8], a
simply by sharing individual tracking errors. Using two type of combination of kinematic and torque control is used to track
wheeled mobile robots namely unicycle type (a nonholonomic reference trajectories, while in [9], an omnidirectional robot is
mobile robot) and omni wheels type (a holonomic mobile robot), a used to help patients during rehabilitation process. All
coordinated control algorithm can achieve a global asymptotically
stable condition of the error dynamics of the systems. Although
algorithms are designed for a single robot.
communication between robots has to be done bidirectionally, the From the results in [1]-[4], coordination between robots
overall group is able to maintain individual tracking while can be achieved simply via exchanging states of the robots. In
coordinating the movements with other robots regardless the the unicycle case, the individual tracking errors are the states
occurring perturbations in the system. Simulation results suggest that are shared between the robots. The stability proof suggests
that information sharing between the robots increase the that proper choice of control structure can simplify the
robustness in coordinating individual trajectories.
mathematical procedure, as well as extending the system with
Keywords: coordinated trajectory tracking, unicycles, omni wheels
robots, Lyapunov function different type of robot.
On the other hand, studies of coordinating trajectories of
I. INTRODUCTION omni wheels in a simple way are rare subjects. From the results
in unicycle case, it can be said that applying similar concept for
The ability of a group of wheeled mobile robots in omni wheels mobile robots will be a valuable result.
coordinating hard and difficult tasks has drawn attention from Motivating by the fact, this paper addresses the problem of
scientist, hobbyist and practitioner. A group of mobile robots is coordinating a group of omni wheels mobile robots by means
able to handle more difficult and complex tasks. The group can of exchanging individual tracking errors. Furthermore, the
achieve a certain formation as well as coordinating movement group is extended to be a mixed of unicycle and omni wheel
between the members so that certain tasks can be mobile robots. The stability of the system is investigated using
accomplished. Lyapunov theorem.
One problem in coordination between robots is the problem The contributions of this paper are as follow: i) a simple
of coordinating individual trajectories so that the overall trajectory tracking controller for a single omni wheels mobile
trajectories form a certain spatial patterns, see e.g. [1]-[4]. In robot, ii) a trajectory tracking controller for omni wheels that
this particular case, a group of unicycle mobile robots is able to achieve coordination with other omni wheels, as well as
coordinating its member’s trajectories such that the with unicycle mobile robots, iii) globally asymptotically stable
coordination forms spatial patterns. The coordination is error dynamics of the overall mixed wheeled mobile robot
achieved by exchanging individual errors within the group. systems, iv) simulation validation of the overall systems.
Due to the nonholomomic constraints that exist in the The rest of the paper is organized as follows. Section 2
systems, unicycle mobile robots is less flexible compared to gives the kinematic model of unicycles and omni wheels as
holonomic mobile robots, e.g. omni wheels mobile robots. In a well as theories for stability. Section 3 presents the control
compact space, the ability of omni wheels to move sideways design process, start with the design of a trajectory tracking for
becomes very important a single omni wheels mobile robot, followed by the extension
Omni wheels mobile robots have been widely used in to m-mixed-unicycle-omni wheels mobile robot systems.
football competition [5],[6]. Although it easily suffers from Stability analysis is presented in this chapter. Section 4 gives
slip, a good control and mechanical design can optimize the the simulation validation and performance analysis of the
ability of an omni wheels mobile robot. Example of trajectory mixed systems. Finally, in section 5 the conclusions of this
work are given and suggestions for further research are One may argue that the kinematic models of the
presented. robots are built using two different approaches. The unicycle is
built using forward and steering velocities as inputs that still
II. MATHEMATICAL PRELIMINARIES need to be translated into left and right wheel speeds. On the
other hand, the omni wheels model is represented directly as a
A. Kinematic model of unicycle mobile robots
function of individual wheel.
This paper considers two mobile robot types: unicycle and
Although the choices seem contradictive, in fact they
omni wheels. The unicycle type is represented in Figure 1.
help in proving the concept that coordination can be achieved
simply by exchanging individual state information regardless
the model of the robots.
𝑉 = 𝒆!! 𝒆!
= 𝒆!! (𝒒!" − 𝒒! )
= 𝒆!! (𝒒!" − 𝐴𝒗)
= 𝒆!! (𝒒!" − 𝐴(𝐴!! (𝒒!" + 𝒌𝒆! )))
= 𝒆!! (𝒒!" − 𝐼𝒒!" − 𝐼𝒌𝒆! )
= −𝒌𝒆!! 𝒆! ≤ 0 (4)
The result in (3) show that is 𝑉 negative definite. Since the
Figure 2. Kinematic model of an omni wheels mobile robot. Lyapunov candidate is a tracking error function of the omni
wheels, this means that the controller given in (3) renders the formation pattern. If the control parameters are chosen so that
origin of tracking errors dynamics of an omni !" !" !" !"
𝑘! = 𝑘! , 𝑘! = 𝑘! ≥ 0, and 𝑘!" , 𝑘!" , 𝑘!" , 𝑘!" , 𝑘!" > 0 and
wheels 𝒆! globally asymptotically stable (GAS). it is assumed that messages are received by the corresponding
robots without delay, then the controller given in (5) and (6)
B. Extension to m-mixed mobile robot systems renders origin of the 𝒆𝒔𝒚𝒔 = 𝒒!! − 𝒒! , 𝒒!! − 𝒒! , … , 𝒒!" −
Up to this point, individual trajectory tracking controllers for 𝒒! globally asymptotically stable (GAS).
either unicycle or omni wheels are available. The extension to -
m-mixed mobile robot systems is formulized as the ability of Proof of Theorem 2.
the group to coordinate individual trajectories so that the The stability is proven using Lyapunov functions. To simplify
coordinated movements create a certain formation type. The the analysis, the system is categorized into three subsystems.
member of the group can be either unicycles or omni wheels. The subsystems represent the types of robots exist in the
To achieve coordination, using the similar principal as in system. Using the theory that states if a subsystem is GAS
[1] and [2], each robot has to share individual tracking errors. using Lyapunov function, then the overall system is also GAS
Using the shared information, a robot that off the trajectories if there is no switching in systems, the error dynamics are
can be compensated by the movements of other robots. As analyzed. The subsystem are as follow:
mentioned in [1] and [2], coordination still can be achieved 1. A subsystem S1, where ∀𝑖, 𝑗 ∈ 𝑚 is a unicycle type. In
only by means of trajectory tracking, although in the weakest this condition the system becomes a homogeny unicycle
possible way. In some cases this condition may not be enough systems, which is similar as in [2] and its error dynamics
to keep the coordination. Thus, exchanging information is still is GAS. All stability proof follows the one given in [2].
needed.
2. A subsystem S2, where ∀𝑖, 𝑗 ∈ 𝑚 is an omni wheels type.
Mathematically, coordination is formulated as the
In this condition the system becomes a homogeny omni
problem of tracking collection of individual trajectories 𝐹 = wheels system.
𝒒!! , 𝒒!! , … , 𝒒!" , so that the overall tracking error systems
3. A subsystem S3, where 𝑖, 𝑗 ∈ 𝑚 is either unicycle or omni
𝒆𝒔𝒚𝒔 = 𝒒!! − 𝒒! , 𝒒!! − 𝒒! , … , 𝒒!" − 𝒒! = 𝒆𝟏 , 𝒆𝟐, … . , 𝒆𝒎 wheels., i.e. the case of mixed mobile robot systems.
goes to zero as time goes to infinity.
To achieve the goal, the controller is equipped with Stability proof for S2
coupling terms that are other robot individual tracking errors. Consider a Lyapunof function 𝑽𝟐 = 𝒆𝑻𝒔𝒚𝒔 𝒆𝒔𝒚𝒔 . For S2, the
For the unicycle robot, the controller is [2]: errors are all coming from omni wheels mobile robots. Thus,
the derivative of 𝑽𝟐 can be expressed as:
𝑣!"# cos 𝑒!"# + 𝑘!" 𝑒!"# 𝒎
𝑣!
= sin 𝑒!"# = 𝒆𝑻𝒔𝒚𝒔 𝒆𝒔𝒚𝒔 𝒆!!" (𝒒!"# − 𝒒!" )
𝜔! 𝜔!"# + 𝑘!" 𝑣!" 𝑒!"# + 𝑘!" 𝑒!"# 𝑽𝟐
𝑒!"# 𝒊!𝟏
!
!" 𝑒!"# − 𝑒!"
𝑘! 𝑒!"
! ! (5)
!!!,!!! 1 + 𝑒!"# + 𝑒!" 𝒎
+ !
= 𝒆!!" 𝒒!"# − 𝐴 𝐴!! (𝒒!" + 𝒌𝒆!
!" 𝑒!"# − 𝑒!" 𝒊!𝟏
𝑘! 𝑒!"
! !
!!!,!!! 1 + 𝑒!"# + 𝑒!"
!
𝑒!!! − 𝑒!"
As for the omni wheels robot, the modified controller is !"
𝑘! 𝑒!"
formulated as follows: !!!,!!! !
1 + 𝑒!"# !
+ 𝑒!"
!
!" 𝑒!!! − 𝑒!" + !
𝑒!"# − 𝑒!"
𝑘! 𝑒!" !"
𝑘! 𝑒!"
! !
!!!,!!! 1+ 𝑒!"# + 𝑒!" !!!,!!! !
1 + 𝑒!"# !
+ 𝑒!"
!! !
𝒗𝒊 = 𝐴 (𝒒!"# + 𝒌𝒆!" ) + 𝑒!"# − 𝑒!" (6) 0
!"
𝑘! 𝑒!" 𝒎
!!!,!!! !
1 + 𝑒!"# !
+ 𝑒!" = −𝒌𝒆!!" 𝒆!" + 𝜑
𝒊!𝟏
0
For unicycle i, j-th robot can be either unicycle or omni wheels !
!" 𝑒!!! − 𝑒!"
and vice versa for the omni wheels mobile robot. Given the 𝑘! 𝑒!"
! !
!!!,!!! 1 + 𝑒!"# + 𝑒!"
controller in (5) and (6), for a mixed group of unicycle and
omni wheels mobile robots, the following theorem holds. 𝜑 = −𝒆!!" !
!" 𝑒!"# − 𝑒!"
𝑘! 𝑒!"
! !
Theorem 2. There exist m-mobile robots, which can be either !!!,!!! 1 + 𝑒!"# + 𝑒!"
unicycle or omni wheels mobile robots. The robots are tracking 0
the given references trajectories that in overall create a spatial = 0, because of the coupling gain choices
Thus, errors in keeping the relative time varying distance between the
𝒎
robots.
𝑽𝟐 = −𝒌𝒆!!" 𝒆!" ≤ 0 (7) If 𝑃!
!"#$
=0, it means that all robots maintain the desired
𝒊!𝟏
relative distances, i.e. the formation is kept. It is to be noted
which is negative definite. This shows that the controller in (6) !"#$
renders the origin of the error dynamics of subsystem S2, i.e. a that 𝑃! can indicate a good formation shape but the one that
group of omni wheels mobile robots, globally asymptotically equals to a rotation mirror of the desired formation shape.
!" !" !" !"
stable (GAS) because of the choice of 𝑘! = 𝑘! , 𝑘! = 𝑘! ≥ 0.
1
For S3, errors are coming from either unicycle or omni wheels. 0.6
y [m]
2
0
derivative of the Lyapunov function contains the unicycle part -0.2
3
4
and the omni wheels part. For the omni wheels, the results are
-0.4
similar to the results in the subsystem S2 due to the specific
structure of the controller and the coupling gain choices. -0.6
in [2], the addition of omni wheels does will not affect the -1
-1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1
Lyapunov condition due to the specific controller and coupling x [m]
gain choices. Thus, the derivative of the Lyapunov function is Fig.3. The reference trajectories of the robots.
simply:
𝑽𝟑 = −𝒌𝒆!!" 𝒆!" −𝒌𝒆!!" 𝒆!" ≤ 0 (8) Table 1. Simulation scenarios
Formation shape (FS)
which is negative definite. This implies that the controller
given in (5) and (6) renders the origin of the error dynamics of
the mixed system globally asymptotically stable (GAS).
0
sharing does not always means the worst performance.
−0.2 Although adding more information can increase robustness, it
has to done correctly according to the desired coordination and
−0.4
type of robot in the group. Thus, to have a strong coordination,
−0.6 it is suggested to have complete communication between the
−0.8
robots. However, this requires a large communication
bandwidth.
−1
−1 −0.8 −0.6 −0.4 −0.2 0 0.2 0.4 0.6 0.8 1
x [m]
0.7
1 1
2 0.8
0.65
3
0.8
4
0.7
0.6
0.6
0.6 0.55
0.4
pE 0.5 0.5
0.2
0.4 0.45
y [m]
0
0.3 0.4
−0.2
0.2 0.35
−0.4
0.1 0.3
0
−0.6
2 0.25
4 7
−0.8 6 0.2
6 5
4
3
8 2
−1 1
−1 −0.8 −0.6 −0.4 −0.2 0 0.2 0.4 0.6 0.8 1
x [m] robot types com topologies
!"#$
Fig. 4. The resulting movements for platoon group ‘o-u-u-o’. Top: no Fig. 5. 𝑃! from simulation usig “triangle” formation.
information sharing; Down: fully coupled
scenario. The top figure shows the resulting movements where 0.8
0.65
1
The results in Figure 5 and 6 shows that for varying robot
robot types com topologies
type in the groups, the controllers also performs differently
although the effect cannot be justified clearly. The simulation !"#$
Fig. 6. 𝑃! from simulation usig “platoon” formation.
results sugges that mixed robots tends to give a better
performance compares to all unicycle or all omni wheels
robots.
0.7
triange, o−u−o−u
V. CONCLUSIONS AND FUTURE WORKS
0.69
0.68
In this paper we present controllers that achieve globally
asymptotically stable of the tracking error dynamics of the
mixed group of unicycle and omni wheels mobile robots.
0.67
0.66
The coordination between the robots can be achieved by
sharing individual tracking errors between the robots. The
pE
0.65
ACKNOWLEDGMENT
pE
0.3
REFERENCES
0.24
1 2 3 4 5 6 7 [1] S.Adinandra, A-M. Diaz, H. Nijmeijer, “Simultaneous trajectory
com topologies
tracking and formation keeping for a group of unicycle mobile
!"#$ robots”, in IEEE Conference on Control, Systems and Industrial
Fig. 7. Slices of 𝑃! from simulation usig “triangle” formation.
Informatics (ICCSII), pp.89-94,2013.
platoon, o−u−o−u
[2] S.Adinandra, “Coordinated trajectory tracking for a group of
0.71
unicycle mobile robots”, Robionetics, pp.88-93, 2013.
0.7 [3] S.Adinandra, E. Schreurs, H. Nijmeijer, “A practical MPC for a
0.69
group of unicycles”, In IFAC NMPC, pp.472-477, 2012.
[4] A. Sadowska, T. van den Broek, H. Huijberts, D. Kostic, N. van
0.68
de Wouw, and H. Nijmeijer, “A Virtual Structure Approach to
0.67 Formation Control of Unicycle Mobile Robots using Mutual
Coupling”, In International Journal of Control, Vol.84, No.11,
pE
0.66
pp.1886-1902, 2011.
0.65
[5] A.S. Conceicao, A.P. Moreira, P.J.Costa, “Model Identification
0.64 of a four wheeled omnidirectional mobile robot”, 2006.
0.63
[6] X.Li, A. Zell, ”Motion control of an omnidirectional mobile
robot”, in International Conference on Informatics in Control,
Automation, and Robotics (ICINCO), 2007.
0.62
0.61
1 2 3 4 5 6 7 [7] L. Gracia, J. Tornero, “Kinematic control of wheeled mobile
com topologies robots”, in Latin American Applied Research, Vol.38, pp.7-16,
0.35
platoon, u−o−o−o
2008.
[8] D-S. Kim, W.H. Kwon, H.S. Park, “Geometric kinematics and
0.34
applications of a mobile robot”, in International Journal of
0.33 Control, Automation, Vol.1, No.3, pp. 376-384, 2003.
0.32
[9] D. Luo, T. Schauer, M. Roth, J. Raisch, “Position and orientation
control of omni-directional mobile rehabilitation robot”, in IEE
0.31
Multi-conference on Systems and Control, pp.50-56, 2012.
pE
0.3
0.29
0.28
0.27
0.26
1 2 3 4 5 6 7
com topologies
!"#$
Fig. 8. Slices of 𝑃! from simulation usig “platoon” formation.