Professional Documents
Culture Documents
Sciencedirect Sciencedirect Sciencedirect
Sciencedirect Sciencedirect Sciencedirect
Sciencedirect Sciencedirect Sciencedirect
com
Available online at www.sciencedirect.com
ScienceDirect
ScienceDirect
Available online atonline
Available www.sciencedirect.com
at www.sciencedirect.com
Procedia CIRP 00 (2018) 000–000
Procedia CIRP 00 (2018) 000–000
ScienceDirect
ScienceDirect
www.elsevier.com/locate/procedia
www.elsevier.com/locate/procedia
There has been some previous work on stiffness inverse kinematics computation. The nominal DH parameters
optimization, taking advantage of the redundancy of the of the robot can be obtained from the manufacturer, but these
manipulator. For example, the Cartesian stiffness control of a parameters do not accurately represent the position and
redundant manipulator was studied by utilizing Stiffness orientation of the end effector due to dimensional
Feasibility Ellipsoids and reconfiguring the manipulator to inaccuracies, tolerances and assembly processes [8]. For this
acquire desired Cartesian stiffness for an external force [2]. reason, kinematic calibration is crucial to obtain accurate
For a robotic drilling application, posture optimization was representation of the pose of the end effector of the modelled
conducted by taking advantage of redundancy around tool axis manipulator. In pursuit of acquiring better representation of
and optimizing stiffness by implementing Jacobian based the robot kinematics, within the entire DH parameter table,
reconfiguration [3]. The redundancy around tool axis was also only the link length and link offset can be optimized from a
used for stability prediction and optimization in robotic pragmatic perspective. This is because the analytical inverse
milling [4,5] and singularity avoidance [6] as well. However, kinematic solution of industrial manipulators becomes
along with stiffness optimization, application of combination impossible if the assumption of a mechanical arm and
of deflection compensation and stability prediction are spherical wrist combination is broken which hinders the
necessary for meeting better accuracy requirements which decoupling of orientation and position to reach a pose.
hasn’t been addressed yet. Consequently the present study used the Self-Adaptive
In this paper, the Cartesian compliance parameter of an Differential Algorithm (SADE) [9] to identify the true DH
industrial serial arm machining robot was aimed to be parameters from experimental data. The link length and link
optimized by utilizing functional redundancy around the tool offset parameters were optimized by minimizing the
axis in a 5-Dof milling operation with analytical inverse normalized error between different measurements (each
kinematics for error compensation and machining stability obtained with a laser tracker) and modelled poses. The
enhancement purposes. optimized DH parameters are shown in Table 1. As it can be
seen from the table, the optimized parameters can have up to
Nomenclature 22mm difference compared to their nominal values.
T 4 × 4 Homogeneous transformation matrix Table 1: Optimized DH Parameters of ABB IRB6640. Nominal values are
Link twist (°) shown in parenthesis.
a Link length (m) Link Link Twist Link Length Link Offset Joint Angle
d Link offset (m) i i (°) ai (m) di (m) i (°)
n Number of joints/ degrees of freedom in joint space
1 (-90) 0.324, (0.320) 0.768, (0.780) ( 1 )
q Actuating joint variable (°)
Ac Virtual joint variable (°) 2 (0) 1.284, (1.280) (0) ( 2 - 90)
Aggregated joint variable (°), q
i i
Ac
3
4
(-90)
(-90)
0.178, (0.200)
(0)
(0)
1.145, (1.142)
( 3 )
( 4 +180)
X ,Y , Z Global base frame 5 (90) (0) (0) ( 5 )
x, y, z Cartesian coordinates (m) 0.199, (0.200)
6 (0) (0) ( 6 )
r , r , r Rotation around X, Y and Z axes (rad)
2.2. Manipulator Stiffness/ Compliance Model
The relationship between Cartesian, joint and complementary condition number. This quantifies the sensitivity of the
stiffness matrix of the manipulator, K c , K and K f , can be Jacobian with respect to joint variables, indicating the
represented as in [10]; manipulability and the distance of the manipulator from a
singularity. The configurations away from singularities were
Kc JT K K f J1 (4) selected and the Cartesian compliance matrices of the
Current work is still performing the joint stiffness parameter manipulator were computed based on (5). The elements
identification for the ABB IRB6640 robot, based on the within the Cartesian compliance matrix play an important role
combination of methodologies developed by [8,12]. For this on the deflection of the robot and hence the redundant degree
reason, the present study relies on the results obtained for of freedom can be used to select the robot’s configuration that
ABB IRB6660, using model parameters as defined in [13]. will result in an optimum stiffness or compliance.
It should be noted that a similar formulation can be
3. Cartesian Compliance Parameter Optimization developed whereby the Cartesian stiffness matrix is
considered instead of the Cartesian compliance matrix. Whilst
A manipulator becomes redundant when it is able to reach this might have advantages in terms of its practical
more degrees of freedom than that of needed for a realized significance, the compliance matrix can be more easily related
task to be executed. This makes the manipulator able to reach to the deflection of the tool in each degree of freedom for an
a particular pose with a theoretically infinite number of input of a force in a particular direction.
configurations. Considering a robotic 5-axis (5-DoF) milling
operation, there are 5-DoFs in operational space compared to 4. Methodology and Preliminary Results
6-DoFs in task space of the robot. This results in functional
redundancy [4]. The redundancy is the rotational DoF around Initially, we investigate the manipulator stiffness along the
tool axis which makes no effect on the execution of the task X-axis, assuming an applied force in the X direction only. The
no matter its value. This is the foundation of Cartesian 2D surface of the machining table at 0.5m height was
Compliance parameter optimization. discretized into evenly spaced (1cm) points in the X & Y-axes
For the sake of simplicity in the manipulator compliance (Figure 2), with tool alignment perpendicular to the surface.
model, the external forces due to cutting process are ignored The redundant DoF around the tool axis, (r ) , was discretized
for the time being, (i.e. K f 0 ), and posture dependent with a resolution of 0.01rad and the cxx parameter was
manipulator compliance is considered. As a result, (4) can be optimized over the machining table.
represented as conventional stiffness formulation or in terms
of Cartesian and joint compliance parameters, Cc and C ; 4.1. Maximized Compliance Parameter, cxx
T
Cc J C J (5)
The manipulator configurations acquiring maximum
Since manipulator deflection under a force lower than its
possible cxx values were obtained and the value of cxx over the
payload is linearly related, the Cartesian compliance matrix
can be used to analyze the effect of each force on the machining table was plotted as shown in Figure 2a.
a) b)
deflection of the manipulator at each DoF by expanding the
Hooke’s Law in terms of compliance parameters. The first
equation describing deflection in X-axis becomes;
x cxx Fx cxy Fy cx F cx F (6)
In order to define the optimization problem, the plane and
the Cartesian coordinates of the 5-DoF poses of a trajectory
need to be defined as in (2). As the redundancy is the
rotational DoF around the tool axis, the rotation around the
local Z-axis is defined as a variable and the other degrees of
freedom are accepted as constant for a 5-DoF pose. Recalling
the fact that 0 r 2 , the rotation around tool axis, r ,
can be discretized between its lower and upper limits into
evenly spaced vector by defining a resolution as r as; Figure 2: (a) Maximised (b) Minimised cxx parameter over machining table
T
r 1r , 2r ,3r , , N r (7) It was observed that as the manipulator moves along the
positive X-axis, the maximum possible cxx parameter tends to
where N represents number of increments and therefore,
decrease. This means that the stiffness of the manipulator
number of 6-DoF poses created from a single 5-DoF pose. For
along the X-axis (when subject to a hypothetical force in X
every 6-DoF pose for the same 5-DoF pose, joint angles are direction) tends to increase. This is due to inevitable
computed with analytical inverse kinematics and the Jacobian alignment of joints along the X-axis. As the tool center point
matrix is computed based on (3). To eliminate the (TCP) deviates from the X-axis, the cxx parameter tends to
configurations close to or at singularity, the condition number
increase, making the manipulator more compliant to a force in
of Jacobian at each configuration was computed based on
X direction due to development of larger torques at joints.
Frobenius norm. The product of Frobenius (Euclidean) norm
of the Jacobian and its inverse was used to obtain the
Huseyin Celikag et al. / Procedia CIRP 77 (2018) 566–569 569
4 Huseyin Celikag/ Procedia CIRP 00 (2018) 000–000
4.2. Minimized Compliance Parameter, cxx (EP/L016257/1) and the European Commission under the
Horizon 2020 COROMA project (723853).
Similarly, the manipulator configurations acquiring
minimum possible cxx were obtained and value of cxx over the References
machining table was plotted in Figure 2b. In this case, a
[1] U. Schneider et al., "Improving robotic machining
similar trend to maximized cxx parameter was observed except
accuracy through experimental error investigation and
the region close to the manipulator base. In this region, the modular compensation," Springer, April 2014.
manipulator configures itself to lower the torque at joints 3
and 5, resulting in the manipulator being stiffer in the [2] A. Ajoudani, N. G. Tsagarakis, and A. Bicchi, "On the
direction of the force. Role of Robot Configuration in Cartesian Stiffness
Control," in IEEE International Conference on Robotics
and Automation (ICRA), Seattle, Washington, May 2015.
4.3. Ratio of Maximized to Minimized Compliance Parameter,
cxx [3] Y. Guo, H. Dong, and Y. Ke, "Stiffness-oriented posture
optimization in robotic machining applications,"
In order to estimate the improvability of the stiffness of the Elsevier, Robotics and Computer-Integrated
manipulator along the X-axis to a force applied parallel to X- Manufacturing, vol. 35, pp. 69-76, February 2015.
axis, the ratio of maximized to minimized cxx parameter over [4] S. Mousavi, V. Gagnol, B. C. Bouzgarro, and P. Ray,
the machining table was taken as shown in Figure 3. "Stability optimization in robotic milling through the
control of functional redundancies," Elsevier, Robotics
and Computer–Integrated Manufacturing (2017), pp. 1-
12, September 2017.
[5] S. Mousavi, V. Gagnol, B. C. Bouzgarrou, and P. Ray,
"Dynamic modeling and stability prediction in robotic
machining," Springer, May 2016.
[6] W. Xiao, H. Strauß, T. Loohß, H. W. Hoffmeister, and J.
Hesselbach, "Closed-form inverse kinematics of 6R
milling robot with singularity avoidance," Springer, vol.
5, pp. 103-110, November 2010.
[7] R. N. Jazar, Theory of Applied Robotics, 2nd ed.
Melbourne, Australia: Springer New York Dordrecht
Heidelberg London, 2010.
Figure 3: Ratio of maximised to minimised cxx parameter
[8] N. R. Slavkovic, D. S. Milutinovic, and M. M. Glavonjic,
It was observed that manipulator cxx parameter can be "A method for off-line compensation of cutting force-
optimized to a maximum of 161% compared to its minimum induced errors in robotic machining by tool path
value around the region close to its base. The parameter can modification," Springer, vol. 70, pp. 2083–96, 2013.
change by a minimum of 58% at (1.2, 0, 0.5)m coordinates in [9] K. Worden and G. Manson, "On the identification of
Figure 3 which indicates the lowest stiffness improvability hysteretic systems. PartI: Fitness landscapes and
over the table. evolutionary identification," Elsevier, Mechanical
Systems and Signal Processing, vol. 29, pp. 201-212,
5. Conclusion January 2012.
[10] G. Alici and B. Shirinzadeh, "Enhanced Stiffness
In conclusion, the manipulator stiffness along X-axis for a Modeling, Identification and Characterization for Robot
force applied in X-direction can be improved by optimizing Manipulators," IEEE Transactions on Robotics, vol. 21,
Cartesian compliance parameter, cxx . For the scenarios pp. 554-564, 2015.
considered it has been shown that the compliance can vary by [11] A. Klimchik, B. Furet, S. Caro, and A. Pashkevich,
between 58% and 161% simply by utilizing the redundant "Identification of the manipulator stiffness model
degree of freedom of the robot. In this way, stiffness parameters in industrial environment," Elsevier,
improvements can be made in order to acquire better accuracy Mechanism and Machine Theory, vol. 90, pp. 1-22, 2015.
in robotic machining operations. However, further work is
needed to consider the true forces that arise during milling, to [12] C. Dumas, S. Caro, S. Garnier, and B. Furet, "Joint
see how the performance can be illustrated using the stiffness stiffnessidentificationofsix-
and compliance matrix approach, and to combine this with a revoluteindustrialserialrobots," Elsevier, Robotics and
stability estimation for chatter avoidance. Computer-Integrated Manufacturing, vol. 27, pp. 881-
888, February 2011.
6. Acknowledgements [13] K. Subrin, L. Sabourin, R. Cousturier, G. Gogu, and Y.
Mezouar, "New redundant architectures in machining:
The authors acknowledge funding provided by EPSRC for serial and parallel robots," in The Manufacturing
Industrial Doctorate Centre in Machining Science Engineering Society International Conference, MESIC
2013, 2013, pp. 158-166.