See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/368396869

Optimal Impedance Control of A 2R Planar Robot Manipulator

Conference Paper · January 2023

CITATIONS READS

16 203

2 authors:

Abdullah Erdemir Mete Kalyoncu

14 PUBLICATIONS 87 CITATIONS
Konya Technical University
131 PUBLICATIONS 1,151 CITATIONS
SEE PROFILE
SEE PROFILE

All content following this page was uploaded by Abdullah Erdemir on 07 May 2023.

The user has requested enhancement of the downloaded file.

 MAS 17th
INTERNATIONAL EUROPEAN CONFERENCE ON MATHEMATICS, ENGINEERING,
NATURAL & MEDICAL SCIENCES
January 26-27, 2023
Cairo, Egypt

OPTIMAL IMPEDANCE CONTROL OF A 2R PLANAR ROBOT MANIPULATOR

Abdullah ERDEMİR1
1
MPG Machinery Production Group Inc. Co., Konya, Türkiye.
1
ORCID ID: https://orcid.org/0000-0002-7267-3111
Mete KALYONCU 2
2
Prof. Dr., Konya Technical University, Department of Mechanical Engineering, Konya, Türkiye.
2
ORCID ID: https://orcid.org/0000-0002-2214-7631

ABSTRACT
In this study, optimum impedance control of a 2R planar robot manipulator was performed. Industrial
applications of robot manipulators are generally related to manipulation tasks such as painting that only
require arm position control. However, there are other robotic tasks such as pushing, polishing, cleaning,
and grinding that require interaction between the robot manipulator and a contact surface or
environment. This fact makes it imperative to control the interaction between the robot and the
environment. The impedance controller aims to control the dynamic relationship between the robot and
the environment. The force applied by the robot to the environment depends on the position of the robot
manipulator endpoint and the corresponding impedance. The impedance controller forces the robot to
follow the desired reference or target impedance. Force task for force/position hybrid controller in
literature; It is divided into two subspaces, the force control subspace, and the position control subspace.
Then, two independent controllers are designed for each subspace. Conversely, the impedance controller
proposed in this study does not attempt to explicitly control the force. Instead, it tries to control the
relationship between the force and the position of the end effector in contact with the environment. It is
also possible to plan a virtual trajectory such that a desired force profile is obtained when the
environment has a rigid structure with known properties. The Bees Algorithm was used to optimize the
proposed impedance controller and a numerical application was made to evaluate its performance. As a
result of the optimization, the objective function was reduced by 57%. The obtained results are presented
numerically and graphically. Thanks to the proposed impedance controller, the robot manipulator
endpoint precisely follows both the desired force profile and the desired position.
Keywords: 2R planar robot manipulator, optimum impedance control, controller design, The Bees
Algorithm.

INTRODUCTION
Industrial applications of robot manipulators are generally related to manipulation tasks such as painting
that only require arm position control [1-3]. The painting process can be done by spraying from a
distance without the need for contact. Therefore, there is usually no need for force control. However,
there are other robotic tasks such as pushing, polishing, cleaning, and grinding that require interaction
between the robot manipulator and a contact surface or environment [4, 5]. Force based control
techniques are well suited to this type of operation. One of these control techniques is impedance control.
The impedance controller aims to control the dynamic relationship between the robot and the
environment [6-8]. The force applied by the robot to the environment depends on the position of the
robot manipulator endpoint and the corresponding impedance. The impedance controller forces the robot
to follow the desired reference or target impedance.
Force task for force/position hybrid controller in literature; It is divided into two subspaces, the force
control subspace, and the position control subspace [9, 10]. Then, two independent controllers are
designed for each subspace. Conversely, the impedance controller proposed in this study does not
attempt to explicitly control the force. Instead, it tries to control the relationship between the force and

PROCEEDINGS BOOK 82
 MAS 17th
INTERNATIONAL EUROPEAN CONFERENCE ON MATHEMATICS, ENGINEERING,
NATURAL & MEDICAL SCIENCES
January 26-27, 2023
Cairo, Egypt

the position of the end effector in contact with the environment. It is also possible to plan a virtual
trajectory such that a desired force profile is obtained when the environment has a rigid structure with
known properties.
The parameters of the impedance controller need to be optimized. Optimization algorithms are divided
into two as global optimization algorithms and local search algorithms [11].
Global optimization algorithms such as Genetic Algorithm [12, 13], Particle Swarm Optimization [14]
and The Bees Algorithm [15, 16] focus on searching across the entire space. Local search algorithms
such as Hooke-Jeeves [17] and Newton Raphson [18] focus on searching in local areas. The Bees
Algorithm was used to optimize the proposed impedance controller and a numerical application was
made to evaluate its performance [19]. The obtained results are presented numerically and graphically.
Thanks to the proposed impedance controller, the robot manipulator endpoint precisely follows both the
desired force profile and the desired position.
MATHEMATICAL MODEL
The robot has two degrees of freedom and consists of revolute joints. The system consisting of 2 linkages
with masses m1 and m2 is shown in Figure 1.

Figure 1. 2 dof robotic system

In order to derive the mathematical model of the system, the position of the centers of mass with respect
to the joint angles must first be obtained. The center of mass position of the linkages with masses m1
and m2 according to the joint angles is shown in Equation (1).

𝐿1
𝑥̅1 = cos θ1
2
𝐿1
𝑦̅1 = cos⁡θ2
2 (1)
𝐿2
𝑥̅2 = 𝐿1 cos θ1 + cos(𝜃1 + 𝜃2 )
2
𝐿2
𝑦̅2 = 𝐿1 sin θ1 + sin(𝜃1 + 𝜃2 )
2

The mass moments of inertia of the rotation of the linkages around the center of mass are taken as 𝐼 =
1
mL2 . The kinetic energy T of the system is given in Equation (2). The potential energy V is given in
12
Equation (3).

1 1 1 1 2
𝑇= 𝑚1 (𝑥̅̇12 + 𝑦̅̇12 ) + 𝐼1 𝜃̇12 + 𝑚2 (𝑥̅̇22 + 𝑦̅̇22 ) + 𝐼2 (𝜃̇1 + 𝜃̇2 ) (2)
2 2 2 2

PROCEEDINGS BOOK 83
 MAS 17th
INTERNATIONAL EUROPEAN CONFERENCE ON MATHEMATICS, ENGINEERING,
NATURAL & MEDICAL SCIENCES
January 26-27, 2023
Cairo, Egypt

𝑉 = 𝑚1 𝑔𝑦̅1 + 𝑚2 𝑔𝑦̅2 (3)

Euler – Lagrangian equation of motion is used to obtain the equation of motion of the system. The
Lagrangian expression for this is given in Equation (4).

𝐿 =𝑇−𝑉 (4)

The Euler – Lagrangian equation of motion applied to obtain the torque in the joints is given in Equation
(5).

𝑑 𝜕𝐿 𝜕𝐿
𝜏𝑖 = ( )− (5)
𝑑𝑡 𝜕𝜃̇𝑖 𝜕𝜃𝑖

The expansion of torques 𝜏1 and 𝜏2 is given in Equation (6).

1 1 2
𝜏1 = ( 𝑚1 𝐿21 + 𝑚2 (𝐿21 + 𝐿22 + 𝐿1 𝐿2 cos(𝜃2 ))) 𝜃̈1 + 𝑚2 𝐿2 (𝐿1 cos(𝜃2 ) + 𝐿2 ) 𝜃̈2
3 3 3
1 2 1
− 𝑚2 𝐿1 𝐿2 sin(𝜃2 ) (𝜃̇1 𝜃̇2 + 𝜃̇2 ) + (𝑚1 + 2𝑚2 )𝑔𝐿1 cos(𝜃1 )
2 2
1 (6)
+ 𝑚2 𝑔𝐿2 cos(𝜃1 + 𝜃2 )
2
1 2 1 1
𝜏2 = 𝑚2 𝐿2 (𝐿1 cos(𝜃2 ) + 𝐿2 ) 𝜃̈1 + 𝑚2 𝐿1 𝐿2 sin(𝜃2 ) 𝜃̇12 + 𝑚2 𝐿22 𝜃̇22
2 3 2 3
1
+ 𝑚2 𝑔𝐿2 cos(𝜃1 + 𝜃2 )
2

IMPLEMENTATION OF IMPEDANCE CONTROL

The impedance control schematic is shown in Figure 2. Impedance control requires forward kinematics
and inverse dynamics equations of the point where the robot is desired to interact, using the joint angles.
The distance between the position obtained with forward kinematics and the desired position is
converted into interaction force by multiplying the spring and by multiplying the velocity with the
damping coefficients. It is then multiplied by the transpose of the Jacobian matrix and converted into
torques at the joints. These torques are added to the torques found by inverse dynamics to obtain a final
torque. This torque is sent to the robot and control is provided.

Figure 2. Impedance control

The PID-based interaction force is given in Equation (7).

PROCEEDINGS BOOK 84
 MAS 17th
INTERNATIONAL EUROPEAN CONFERENCE ON MATHEMATICS, ENGINEERING,
NATURAL & MEDICAL SCIENCES
January 26-27, 2023
Cairo, Egypt

𝑥2𝑑 − 𝑥2 𝑥̇ − 𝑥̇ 2 𝑥2𝑑 − 𝑥2
𝐹𝑖𝑛𝑡 = 𝑘 [𝑦 − 𝑦 ] + 𝑏 [ 2𝑑 ] + 𝑖 ∫ [𝑦 − 𝑦 ] 𝑑𝑡 (7)
2𝑑 2 𝑦̇ 2𝑑 − 𝑦̇ 2 2𝑑 2

The spring coefficient of the impedance controller is defined as k, the damping coefficient b, and the
integral gain i. 𝑥2𝑑 and 𝑦2𝑑 are the targeted location of the robot's endpoint. 𝑥2 and 𝑦2 are the actual
position of the robot's endpoint. The position of the end point of the robot is given in Equation (8).

𝑥2 = 𝐿1 cos θ1 + 𝐿2 cos(𝜃1 + 𝜃2 )
(8)
𝑦2 = 𝐿1 sin θ1 + 𝐿2 sin(𝜃1 + 𝜃2 )

In order for the interaction force in Equation (7) to be converted into torques, it is necessary to multiply
it by the Jacobian matrix of the robot endpoint. The Jacobian matrix of the system is given in Equation
(9).

𝑑𝑥2 𝑑𝑥2
𝑑𝜃1 𝑑𝜃2
𝐽=
𝑑𝑦2 𝑑𝑦2
(9)
[𝑑𝜃1 𝑑𝜃2 ]
−𝐿1 sin(𝜃1 ) − 𝐿2 sin(𝜃1 + 𝜃2 ) −𝐿2 sin(𝜃1 + 𝜃2 )
𝐽=[ ]
𝐿1 cos(𝜃1 ) + 𝐿2 cos(𝜃1 + 𝜃2 ) 𝐿2 cos(𝜃1 + 𝜃2 )

The impedance torques obtained by multiplying the interaction forces with the transpose of the Jacobian
matrix are given in Equation (10).

𝜏𝑒 = 𝐽𝑇 𝐹𝑖𝑛𝑡
−𝐹𝑥 (𝐿1 sin(𝜃1 ) + 𝐿2 sin(𝜃1 + 𝜃2 )) + 𝐹𝑦 (𝐿1 cos(𝜃1 ) + 𝐿2 cos(𝜃1 + 𝜃2 )) (10)
𝜏𝑒 = [ ]
−𝐹𝑥 𝐿2 sin(𝜃1 + 𝜃2 ) + 𝐹𝑦 𝐿2 cos(𝜃1 + 𝜃2 )

In Equation (10), the x component of the 𝐹𝑖𝑛𝑡 force is 𝐹𝑥 and the y component is 𝐹𝑦 . The final torques
are the sum of dynamic torques in Equation (6) and the impedance torques in Equation (10).
NUMERICAL APPLICATION
L1 = L2 = 1 m and m1 = m2 = 5 kg. Figure 3 shows the path followed by the scenario. The BC plane is 1
m away from the first joint of the robot and is at a 65° inclination with the ground. The endpoint of the
robot first moves from point A to point B. It then proceeds from point B to point C by applying a force
of 50 N to the BC plane.

PROCEEDINGS BOOK 85
 MAS 17th
INTERNATIONAL EUROPEAN CONFERENCE ON MATHEMATICS, ENGINEERING,
NATURAL & MEDICAL SCIENCES
January 26-27, 2023
Cairo, Egypt

Figure 3. Numeric application view

In order for the robot to apply 50 N perpendicular to the surface while its tip point moves in the BC
plane, the 50 N force at a 90° - 65° = 25° inclination from the robot's tip point to the ground is added to
the interaction force seen in equation (11).

cos(25°)
𝐹𝑖𝑛𝑡,𝑚𝑜𝑑𝑖𝑓𝑖𝑒𝑑 = 𝐹𝑖𝑛𝑡 + 50 [ ] (11)
sin(25°)

The objective function used to optimize the k, b and i coefficients in Equation (7) is defined in Equation
(12).
𝑡2

𝑓𝑜𝑏𝑗 = ∫ (𝑥2𝑑 − 𝑥2 )2 + (𝑦2𝑑 − 𝑦2 )2 + (𝐹𝑛 − 50𝑁)2 (12)

𝑡1

𝐹𝑛 is the reaction force of the BC plane against the robot endpoint. 𝑡1 is the first contact time of the
robot's endpoint to point B. 𝑡2 is the time when the robot endpoint reaches point C.
The objective function is obtained by squaring the distance error and the difference of the surface
reaction force from 50 N. By minimizing this objective function, both the distance error will be reduced,
and the surface reaction force will be approached to 50 N.
THE BEES ALGORITHM
The algorithm starts with n scout bees being placed randomly in the search space. The fitness of the
population is calculated; this means the error of all the scout bees are calculated. The array of the scout
bees is reordered from minimum error to maximum error. The best m sites are selected to be search for
neighborhood search. The next bees will search those sites within the radius of patch size which is ngh.
But more scout bees will be sent to elite sites which are shown as number e. The remaining scout bees
are less than the number of elite bees. Each site is reordered from minimum error to maximum error. In
addition, the fittest bee is selected for that site. The remaining (n-m) bees are replaced with the new
randomly created bees. The fitness of the new population is recalculated, and the loop continues until

PROCEEDINGS BOOK 86
 MAS 17th
INTERNATIONAL EUROPEAN CONFERENCE ON MATHEMATICS, ENGINEERING,
NATURAL & MEDICAL SCIENCES
January 26-27, 2023
Cairo, Egypt

the stop condition occurs. The pseudo code for The Bees Algorithm in its simplest form is:
1. Initialize population with random solutions.
2. Evaluate fitness of the population.
3. While (stopping criterion not met)
4. Select sites for neighborhood search.
5. Recruit bees for selected sites (more bees for best e sites) and evaluate fitness.
6. Select the fittest bee from each patch.
7. Assign remaining bees to search randomly and evaluate their fitness.
8. End While
The Bees Algorithm is shown in Figure 4.

Figure 4. The Bees Algorithm

The parameters of The Bees Algorithm are given in Table 1.
Table 1. The Bees Algorithm parameters

n m e nep nsp ngh pmax pmin

10 7 4 5 3 0.1 [30,5,5] [0,0,0]

RESULTS AND DISCUSSION

The result of the optimization made with The Bees Algorithm is given in Table 2. The optimization is
completed with 18243 iterations. As a result of the optimization made from a randomly generated
starting point, the objective function decreased by 57% compared to the initial condition.

PROCEEDINGS BOOK 87
 MAS 17th
INTERNATIONAL EUROPEAN CONFERENCE ON MATHEMATICS, ENGINEERING,
NATURAL & MEDICAL SCIENCES
January 26-27, 2023
Cairo, Egypt

Table 2. The Bees Algorithm parameters

Initial Conditions The Bees Algorithm Result Change

𝑘 25 25.10477514 +0.41%

𝑏 3 2.416703157 -19.44%

𝑖 0 0.541278214 +0.54

𝑓𝑜𝑏𝑗 11.09625654 4.771782226 -57%

The convergence graph during the optimization of the objective function is shown in Figure 5.

Figure 5. Convergence graph of the objective function.

The positions of the robot during the scenario are shown in Figure 6.

PROCEEDINGS BOOK 88
 MAS 17th
INTERNATIONAL EUROPEAN CONFERENCE ON MATHEMATICS, ENGINEERING,
NATURAL & MEDICAL SCIENCES
January 26-27, 2023
Cairo, Egypt

Figure 6. Movement of the robot in space and on the plane

The robot touches the BC plane at the 5th second. According to the scenario 50 N force must be applied
on the surface and robot’s endpoint must move from point B to C. From 5 th to 7th second, robot’s end
point pushes on the BC plane until the reaction force is 50 N. From point B to point C, robot’s endpoint
moves while applying 50 N on the plane. The force change is shown in in Figure 7.

Figure 7. The reaction force of the plane during the movement of the robot
The reaction force of the BC plane to the endpoint of the robot ranges from 49.5 N to 50.8 N. The desired
force of 50 N is applied within -1% and +1.6%.
In Figure 8, the desired positions are xd and yd, and the actual positions are seen as x2 and y2. As seen in
the figure, actual positions are very close to desired locations.
The interaction distance at the endpoint of the robot is given in Figure 9.

PROCEEDINGS BOOK 89
 MAS 17th
INTERNATIONAL EUROPEAN CONFERENCE ON MATHEMATICS, ENGINEERING,
NATURAL & MEDICAL SCIENCES
January 26-27, 2023
Cairo, Egypt

Figure 8. Desired and actual locations of the robot's endpoint.

Figure 9. Interaction distances at the robot endpoint

Accordingly, while there are very large interaction distances before the contact with the BC plane, that
is, before the 5th second, it is seen that the interaction distances are less when applying 50 N on the BC
plane, that is, after the 5th second.
CONCLUSIONS
In this study, the position of the endpoint of a 2 dof robot was controlled using the impedance control
technique. In addition, during the movement of the robot's end point on a plane, a change was made on
the impedance force to apply a workload such as 50 N. During this movement, the workload is applied
with an error between -1% and +1.6%.
In addition, the impedance controller parameters are optimized with The Bees Algorithm. As a result of
the optimization made with a randomly determined initial condition, an improvement of 57% was
achieved and the optimization was completed.
In this study, it has been shown that a force can be applied in a desired direction while the end point of
the robot moves on a trajectory using the impedance control technique. In future studies, the change in
desired position can be studied instead of changing the interaction force in this study.

PROCEEDINGS BOOK 90
 MAS 17th
INTERNATIONAL EUROPEAN CONFERENCE ON MATHEMATICS, ENGINEERING,
NATURAL & MEDICAL SCIENCES
January 26-27, 2023
Cairo, Egypt

REFERENCES
[1] Chen, H., Fuhlbrigge, T., Li, X., 2008, Automated industrial robot path planning for spray
painting process: a review, Proc. 2008 IEEE International Conference on Automation Science and
Engineering, IEEE, pp. 522-527.
[2] Muzan, I. W., Faisal, T., Al-Assadi, H., Iwan, M., 2012, Implementation of industrial robot for
painting applications, Procedia engineering, 41(pp. 1329-1335.
[3] Asakawa, N., Takeuchi, Y., 1997, Teachingless spray-painting of sculptured surface by an
industrial robot, Proc. Proceedings of international conference on robotics and automation, IEEE, pp.
1875-1879.
[4] Bicchi, A., Kumar, V., 2000, Robotic grasping and contact: A review, Proc. Proceedings 2000
ICRA. Millennium conference. IEEE international conference on robotics and automation. Symposia
proceedings (Cat. No. 00CH37065), IEEE, pp. 348-353.
[5] Pliego-Jiménez, J., Arteaga-Pérez, M. A., 2015, Adaptive position/force control for robot
manipulators in contact with a rigid surface with uncertain parameters, European Journal of Control,
22(pp. 1-12.
[6] Hogan, N., 1984, Impedance control: An approach to manipulation, Proc. 1984 American
control conference, IEEE, pp. 304-313.
[7] Hogan, N., 1984, Impedance control of industrial robots, Robotics and computer-integrated
manufacturing, 1(1), pp. 97-113.
[8] Hogan, N., 1987, Stable execution of contact tasks using impedance control, Proc. Proceedings.
1987 IEEE International Conference on Robotics and Automation, IEEE, pp. 1047-1054.
[9] Hayati, S., 1986, Hybrid position/force control of multi-arm cooperating robots, Proc.
Proceedings. 1986 IEEE International Conference on Robotics and Automation, IEEE, pp. 82-89.
[10] Zeng, G., Hemami, A., 1997, An overview of robot force control, Robotica, 15(5), pp. 473-482.
[11] Erdemir, A., Kalyoncu, M., 2019, Optimization of a Multi-Axle Steered Heavy Vehicle Steering
Mechanism by using the Bees Algorithm and the Hooke-Jeeves Algorithms Simultaneously, Proc. The
1st International Symposium On Automotive Science And Technology (ISASTECH 2019),
Ankara/Turkey, pp. 613-622.
[12] Lau, T. L., 1999, Guided genetic algorithm, pp. v, 166 leaves.
[13] Ortmann, M., Weber, W., 2001, Multi-criterion optimization of robot trajectories with
evolutionary strategies, Proc. Proceedings of the 2001 Genetic and Evolutionary Computation
Conference. Late-Breaking Papers, Morgan Kaufmann San Francisco, CA, pp. 310-316.
[14] Wang, D., Tan, D., Liu, L., 2018, Particle swarm optimization algorithm: an overview, Soft
computing, 22(2), pp. 387-408.
[15] Pham, D., Kalyoncu, M., 2009, Optimisation of a fuzzy logic controller for a flexible single-link
robot arm using the Bees Algorithm, Proc. 2009 7th IEEE International Conference on Industrial
Informatics, IEEE, pp. 475-480.
[16] Erdemir, A., Kalyoncu, M., 2015, Bir Ağır Vasıtanın Çok Akslı Direksiyon Mekanizmasının Arı
Algoritması Kullanılarak Optimizasyonu, Uluslararası Katılımlı 17. Makina Teorisi Sempozyumu –
(UMTS 2015), 17), pp. 421-426.
[17] Hooke, R., Jeeves, T. A., 1961, ``Direct Search''Solution of Numerical and Statistical Problems,
Journal of the ACM (JACM), 8(2), pp. 212-229.
[18] Ypma, T. J., 1995, Historical development of the Newton–Raphson method, SIAM review,
37(4), pp. 531-551.

PROCEEDINGS BOOK 91
 MAS 17th
INTERNATIONAL EUROPEAN CONFERENCE ON MATHEMATICS, ENGINEERING,
NATURAL & MEDICAL SCIENCES
January 26-27, 2023
Cairo, Egypt

[19] Erdemir, A., Alver, V., Kalyoncu, M., 2022, Arı Algoritması Kullanılarak Önden Dümenlemeli
Bir Aracın Dümenleme Mekanizmasının Optimizasyonu, Proc. The 6th International Aegean
Conferences on Innovation Technologies & Engineering, İzmir, Türkiye, pp. 50-59.

PROCEEDINGS BOOK 92

View publication stats