Professional Documents
Culture Documents
Research Article
Keywords: Bald eagle swarm intelligent optimization algorithm, Multi-stage optimization strategy, Inverse
Kinematics, Redundant robotic arm
DOI: https://doi.org/10.21203/rs.3.rs-2333928/v1
License: This work is licensed under a Creative Commons Attribution 4.0 International License.
Read Full License
A 7DOF redundant robotic arm inverse kinematic solution algorithm
based on Bald Eagle swarm intelligent optimization algorithm
Guojun Zhaoa,d, Ying Suna,d*, Du Jianga,c*, Xin Liua,d*, Bo Taoa,d,
Guozhang Jiangc, Jianyi Konga,b,c, Juntong Yunb,d, Ying Liub,d, Gongfa Lia,b,c
a
Key Laboratory of Metallurgical Equipment and Control Technology of Ministry of Education, Wuhan University
of Science and Technology, Wuhan 430081, China
b
Research Center for Biomimetic Robot and Intelligent Measurement and Control, Wuhan University of science
and Technology, Wuhan 430081, China
c
Hubei Key Laboratory of Mechanical Transmission and Manufacturing Engineering, Wuhan University of science
and Technology, Wuhan 430081, China
d
Precision Manufacturing Research Institute, Wuhan University of Science and Technology, Wuhan 430081,
China.
*Correspondence: Sunying@wust.edu.cn; Jiangdu@wust.edu.cn; Liuxin@wust.edu.cn
Abstract: Solving the robot inverse kinematic solution is the key to the subsequent path planning and trajectory
tracking control of the robot, which is of great research significance. The inverse kinematic solution of the redundant
robotic arm is a great challenge because the parsing solution cannot be obtained by the conventional inverse kinematic
solution method. The swarm intelligent optimization algorithm is widely used in the inverse kinematic solution
problem of redundant robotic arms by converting the inverse kinematic solution problem of the robotic arm into the
minimum value optimization problem of the fitness function, avoiding the tedious process of the traditional inverse
kinematic solution. This paper innovatively applies the bald eagle swarm intelligent optimization algorithm (BES
algorithm) to the inverse motion solution problem of a 7DOF redundant robotic arm for the first time. The BES
algorithm simulates the process of prey hunting by bald eagles in nature and consists of three main phases: selection
phase, search phase, and dive phase. In these three phases, the algorithm updates the joint angles to be sought by
using different optimization strategies, and obtains high accuracy position values by bringing the obtained joint angles
into the positive kinematic expression of the robot arm. The article takes the YuMi 14000 ABB 7DOF industrial
robotic arm and the S-R-S humanoid 7DOF robotic arm as the research objects, and the BES algorithm is
experimentally compared with the traditional swarm intelligence optimization algorithms DE algorithm, FA
algorithm, FOA algorithm, GA algorithm and PSO algorithm in terms of position solving accuracy. The experimental
results show that the BES algorithm has higher position solution accuracy and solution stability compared with other
algorithms.
Keywords: Bald eagle swarm intelligent optimization algorithm, Multi-stage optimization strategy, Inverse
Kinematics, Redundant robotic arm
1. Introduction of the robot arm by observing the geometric relationship of
the robot arm. This method is more intuitive compared to
the analytical method, but the application scenario of the
The inverse kinematic solution problem of the robotic
geometric method is narrower than that of the analytical
arm is of great research significance as the basis of robotic
method. Unlike the parsing method, the numerical solution
arm trajectory planning as well as control (Deng et al. 2021,
method transforms the inverse kinematic solution problem
Xiao et al. 2021). The robotic arm can be classified into
of the robotic arm into a minimum value optimization
redundant robotic arm and non-redundant robotic arm
problem of the objective function by constructing an
according to the number of degrees of freedom of the
adaptation function containing the solved joint angles and
robotic arm. Generally, robotic arms with six degrees of
optimizing it to finally obtain a set of approximate
freedom and less than six degrees of freedom are called
solutions that satisfy the requirements. The current
non-redundant robotic arms, and robotic arms with more
mainstream numerical methods contain intelligent
than six degrees of freedom are called redundant robotic
algorithms, neural networks, and numerical iterative
arms. Compared with non-redundant robotic arms,
methods based on Jacobi matrices (Sherbiny et al. 2017).
redundant robotic arms have more degrees of freedom and
The traditional parsing methods cannot be applied to the
can be applied to more flexible scenarios. The introduction
redundant robotic arm because the parsing solution is not
of redundant degrees of freedom, on the one hand, makes
available for the redundant robotic arm. The current
the robotic arm more flexible and provides a broader
inverse kinematic solution methods for redundant robotic
scenario for robotic arm applications, but on the other hand,
arms mainly use numerical methods based on intelligent
it also makes the inverse kinematics solution of the robotic
algorithms.
arm difficult. Through the efforts of researchers, the non-
Some related scholars have studied the method of
redundant robotic arm satisfying the "piper" criterion can
solving the inverse kinematic solution of redundant robotic
now decouple the joints of the intersecting axes and finally
arms based on intelligent algorithms. Wang et al. (2008)
obtain the parsing solution of the non-redundant robotic
applied the DE algorithm to the inverse kinematic solution
arm (Tong et al. 2021, Li et al. 2022, Baker and Wampler
problem of a parallel robot arm by transforming the inverse
(1988)). That is, for the non-redundant robotic arm with
kinematic solution problem of the parallel robot arm into
three axes intersecting at one point, the form of the joint
the minimax optimization problem of the objective
angle obtained contains only known parameters. In
function to finally achieve the inverse kinematic solution.
contrast to the non-redundant robotic arm, the inverse
The article solves the minimax optimization problem of the
kinematic parsing solution cannot be solved for the
objective function mainly by using the idea of selecting the
redundant robotic arm, so the inverse kinematic solution
optimal vector in the DE algorithm. According to two
problem becomes quite challenging for the redundant
different random vectors in the population, the third vector
robotic arm (Ananthanarayanan and Ordóñez (2015),
is obtained by mutation and hybridization, and the vector
Shimizu et al. 2008, Tarokh and Kim (2007), Dereli and
is selected by comparing the objective value of the new
Koker (2020)).
vector with that of the old vector, and the minimum value
The current methods for solving the inverse
of the objective function is finally obtained to achieve the
kinematics of robotic arms are mainly divided into two
inverse kinematic solution. Jesus et al. (2021) proposed an
types: parsing and numerical methods (Roth et al. 1973,
improved multimodal FA algorithm for the inverse
Paul et al. 1981, Ananthanarayanan and Ordonez (2013),
kinematic solution problem of redundant robotic arms. The
Singh and Claassens (2010)). Among them, the parsing
traditional firefly algorithm is improved by setting multiple
method is divided into analytical and geometric methods.
joint configurations to improve the inverse kinematic
The analytical method is mainly used to obtain the
solution performance. Parker et al. (1989) proposed a GA
equation with the same number of variables by shifting the
algorithm to solve the inverse kinematic solution of a
terms of the positive kinematic expression of the robot arm,
redundant robotic arm. The end-effector position of the
and finally the inverse kinematic solution is obtained by
robotic arm is positioned by the GA algorithm, while
solving the equation. The analytical method is fast and
ensuring that the joint displacement from the initial
accurate, but it has limited applications because it is
position is minimized. Nearchou (1998) proposed an
generally used for robotic arms that satisfy the "piper"
improved GA algorithm (mGA) for the inverse kinematic
criterion (Roth (1975), Tsai and Morgan A (1984), Lee and
solution problem of a redundant robotic arm in an obstacle
Ziegler (1984), Nakamura and Hanafusa (1986)). The
environment by transforming the inverse kinematic
geometric method obtains the inverse kinematic solution
solution problem of the redundant robotic arm into an algorithm (Shareef et al. 2015), WOA algorithm (Mirjalili
optimization problem under constraints to achieve the et al. 2016), GOA algorithm (Saremi et al. 2017), SSA
inverse kinematic solution. The article achieves the inverse algorithm (Mirjalili et al. 2017), SOA algorithm (Dhiman
kinematic solution by determining the representation et al. 2018), STOA algorithm (Dhiman et al. 2019), MOA
mechanism of the inverse kinematic solution problem algorithm (Zervoudakis et al. 2020), WHO algorithm
encoded as an artificial chromosome, setting the initial (Naruei et al. 2021), and SO algorithm (Hashim et al. 2022),
population for generating chromosomes, selecting the among others.
fitness function for evaluating chromosomes, setting the In this paper, the bald eagle swarm intelligent
genetic operators for reproduction, crossover, and mutation, optimization algorithm (BES) is innovatively used to solve
and setting the control parameters, and finally optimizing the inverse kinematic solution problem of 7D0F redundant
the objective function. Dereli et al. (2018) implemented an robotic arm. By comparing the BES algorithm with the DE
improvement to the traditional PSO algorithm by algorithm, FA algorithm, FOA algorithm, GA algorithm
modifying the fixed weights of the PSO algorithm to and PSO algorithm experimentally on two different 7DOF
random weights, and used the improved algorithm to solve redundant robotic arms, it is found that the inverse
the inverse kinematic solution problem of the redundant kinematic solution obtained by the BES algorithm has
robotic arm, ultimately improving the accuracy of the higher positional accuracy.
solution of the end-effector position pose. Dereli et al. The remainder of this paper is as follows. Section II
(2022) improved the PSO algorithm inspired by golfing examines the YuMi 14000 ABB 7DOF industrial robot arm
and applied it to the 7DOF inverse kinematics solution, and the S-R-S humanoid 7DOF robotic arm for kinematic
which significantly improved the accuracy of the inverse modeling as well as positive kinematic analysis. Section III
kinematics solution. Zhao et al. (2022) proposed an introduces the BES algorithm. Section IV describes the
improved PSO algorithm for the inverse kinematic solution problem of solving the inverse kinematic solution of the
problem of a 7DOF redundant robotic arm. The accuracy redundant robotic arm using the BES algorithm, along with
of the inverse kinematic solution of the redundant robotic the fitness function and the algorithm flow. In Section V,
arm is finally improved by setting a new velocity initial the BES algorithm is experimentally compared with the
approach, adjusting the learning factor, and using adaptive commonly used intelligent algorithms DE algorithm, FA
weights. Mao et al. (2017) proposed a hybrid strategy algorithm, FOA algorithm, GA algorithm, and PSO
combining the PSO algorithm with the DE algorithm for algorithm for redundant robotic arm inverse kinematics
solving the inverse kinematic solution of redundant robotic solution, using YuMi 14000 ABB 7DOF industrial robotic
arms. The algorithm was found to have better convergence arm and S-R-S humanoid 7DOF robotic arm as
with guaranteed accuracy by comparison. Aydilek (2018) experimental objects. This is followed by the discussion
combined the FA algorithm with the PSO algorithm for and conclusion section of the paper.
solving the inverse kinematic solution of the robotic arm,
which eventually significantly improved the solution 2. Positive kinematic analysis
accuracy.
The current intelligent algorithms for solving the 2.1. 7DOF robot arm kinematic model
inverse kinematic solution of the redundant robotic arm
In this section, the YuMi 14000 ABB 7DOF industrial
mainly contain DE algorithm, FA algorithm, FOA
robotic arm and the S-R-S humanoid 7DOF robotic arm
algorithm, GA algorithm, PSO algorithm and the improved
(Shimizu et al. 2008) are used to model the kinematics.
algorithms of the above mentioned algorithms.
Among them, axis 1 and axis 3 of the YuMi 14000 ABB
Considering that the inverse kinematic solution of the
7DOF industrial robotic arm control the overall rotational
redundant robotic arm by intelligent algorithms requires
motion of the robotic arm, axis 2 and axis 4 control the
the optimization of seven or more parameters at the same
overall bending motion of the robotic arm, axis 5 and axis
time, the fitness function has a high complexity, which
6 control the rotational and bending motion of the wrist of
leads to the actual solution accuracy of the above
the robotic arm, respectively, and axis 7 uses the flange for
algorithms is not high. In order to further improve the
rotational motion. The S-R-S humanoid 7DOF robotic arm
solution accuracy of the redundant robotic arm, a new
has a similar structure to a human hand, and the seven
intelligent algorithm needs to be investigated. The current
rotating joints of this robotic arm are arranged into the
mainstream swarm intelligence optimization algorithms
shoulder, elbow and wrist. Among them, the shoulder and
include GWO algorithm (Mirjalili et al. 2014), LSA
wrist joints consist of three joints forming a ball joint, and
the joint axes intersect at a point. The structural sketches Axisi
Axisi −1 ConnectingRod i −1
of the two types of robotic arms are shown in Fig. 1 and
Fig. 2, respectively.
Zˆ i
Zˆ i −1 Yˆi −1
ConnectingRod i
Yˆi ai
5 7
ai −1 di Xˆ i
Xˆ i −1 i
3 4 6
i −1
Fig. 3 Sketch of the structure of the adjacent linkage of the
1 robot arm
2 In Fig. 3, ai indicates the length of the connecting
rod, being the distance along axis Xˆ i , moving from Zˆ i
to Zˆ i +1 . i indicates the angle of twist of the connecting
rod, which is the angle of rotation around the Xˆ i axis,
Fig. 1 Sketch of YuMi 14000 ABB 7DOF industrial robot arm
from Zˆ i to Zˆ i +1 . d i denotes the linkage offset, being
structure
the distance along axis Zˆ i , moving from Xˆ i −1 to Xˆ i .
i indicates the angle of rotation around axis Zˆ i , from
4 Xˆ i −1 to Xˆ i . By determining the parameters in Figure 3,
the standard DH parameter table was finally obtained, as
3 5
shown in Table 1.
7
2 Table 1 Table of standard DH parameters
i ai (m) i () di (m) ( )
1 6
1 a1 1 d1 1
2 a2 2 d2 2
3 a3 3 d3 3
4 a4 4 d4 4
Fig. 2 Sketch of S-R-S humanoid 7DOF robotic arm structure
5 a5 5 d5 5
After determining the structure of the robot arm, the
6 a6 6 d6 6
next step is to model the kinematics of the robot arm
7 a7 7 d7 7
according to its structure. Denavit and Hartenberg (1955)
first proposed a DH modeling method, which constructs a
DH parameter table based on the linkage length, linkage 2.2. Kinematic analysis
torsion angle, and linkage deflection of the robotic arm.
Based on the DH parameter table, the position posture The positional attitude of the robotic arm end-effector
transformation matrix between the linkages of the robot in space can be described by the positional attitude
arm is established, and finally the position posture matrix transformation matrix of the coordinate system fixed to the
of the end-effector of the robot arm is obtained based on robotic arm end-effector with respect to the base
the position posture transformation matrix between the coordinate system. In this case, the positional attitude
linkages and the initial position posture matrix of the end- transformation matrix between the adjacent links of the
effector of the robot arm, and the kinematic model of the robot arm can be described by Eq. (1).
robot arm is completed. The sketch of the structure of the r t
i −1T = (1)
i
adjacent linkage of the robot arm is shown in Fig. 3. 0 1
In Eq. (1), i −1iT denotes the position pose
transformation matrix of linkage i-1 with respect to linkage
i. r denotes a 3 × 3 matrix to describe the pose. t denotes a
3 × 1 vector to describe the position. When the structure of 1 0 -90 0.317 −90 1 90
the robotic arm is determined, according to the standard 2 0 90 0 −45 2 45
DH modeling approach (Colson and Perreira (1983), Ruoff 3 0 -90 0.415 −120 3 120
(1981)), the position-pose transformation matrix between 4 0 90 0 0 4 135
adjacent links of the robotic arm can be described by Eq. 5 0 -90 0.48 −90 5 90
(2). 6 0 90 0 −90 6 90
cos i − sin i cos i sin i sin i ai cos i 7 0 0 0.007 −120 7 120
sin cos i cos i − cos i sin i ai sin i
(2)
i-1T =
i i
0 sin i cos i di 1 0 0 a3 + d5 + d 7
0
0 0 0 1 1 0 0
T0 = (4)
In Eq. (2), the 3 × 3 matrix in the upper left corner 0 0 0 d1 + d3 + a4
represents the pose description matrix, and the first three 0 0 1 1
rows of the fourth column represent the position
description vector. ai denotes the length of the 1 0 0 d1 + d3 + d5 + d 7
connecting rod i, i denotes the torsion angle of the 0
1 0 0
connecting rod i, d i denotes the offset of the connecting T0 = (5)
0 0 0 0
rod i, and i is the joint angle i to be found. When the
initial position pose matrix of the end-effector of the robot 0 0 1 1
arm is known, the positive kinematic expression of the
robot arm is determined by the product of the position pose 3. BES algorithm
transformation matrix of each linkage and the initial
position pose matrix, as shown in Eq. (3). In this paper, the BES algorithm is chosen for the first
time to solve the inverse kinematic solution of the seven-
T = 01T 21T 32T 43T 54T 65T 76TT0 (3) degree-of-freedom robotic arm. the BES algorithm
simulates the prey hunting process of the bald eagle in
In Eq. (3), i −1iT (i = 2 ~ 7) represents the position
nature, which consists of three main phases, namely, the
pose transformation matrix of linkage i-1 with respect to
selection phase, the search phase, and the dive phase, as
linkage i. T0 represents the initial position pose matrix of
shown in (a), (b), and (c) in Fig. 4, respectively.
the end-effector of the robot arm. According to the standard
DH modeling approach, the standard DH parameter tables
for the YuMi 14000 ABB 7DOF industrial robotic arm and
the S-R-S humanoid 7DOF robotic arm are shown in Table
2 and Table 3, respectively. The initial position pose (a) Selection phase (b) Search phase (c) Swoop phase
matrices corresponding to the end-effectors of the two Fig. 4 Bald eagle prey hunting process
robotic arms are shown in Eq. (4) and Eq. (5), respectively.
3.1. Selection phase
Table 2 YuMi 14000 ABB 7 DOF industrial robot arm standard
During the selection phase, the bald eagle selects the
DH parameter table
search space for its prey. Eq. (1) represents this behavior
i ai (m) i () di (m) ( ) mathematically.
1 -0.03 -90 0.166 −168.5 1 168.5
2 0.03 90 0 −143.5 2 43.5 Pi , new = Pbest + r ( Pmean − Pi ) (1)
3 0.0405 -90 0.2515 −123.5 3 80
In Eq. (1), Pi , new denotes the position of the ith bald
4 0.0405 -90 0 −290 4 290
eagle after the selection of the search space. Pbest denotes
5 0.027 -90 0.265 −88 5 138
the best position of the bald eagle in the current search
6 -0.027 90 0 −229 6 229
space. denotes the parameter that controls the change
7 0 0 0.036 −168.5 7 168.5
of the position and takes values in the range of [1.5, 2]. r
denotes a random number with values in the range [0, 1].
Table 3 S-R-S humanoid 7 DOF robotic arm standard DH
Pmean denotes the average of the positions of the bald eagle.
parameter table
Pi denotes the position of the ith bald eagle.
i ai (m) i () di (m) ( )
3.2. Search phase 4. Inverse kinematic solution for robotic arm using BES
During the search phase, the bald eagle searches for
algorithm
prey in a selected search space in a spiral motion. Eq. (2)
represents this behavior mathematically.
4.1. Problem description
Pi , new = Pi + y(i) ( Pi − Pi +1 ) + x(i) ( Pi − Pmean ) (2)
Unlike the conventional inverse kinematic solution,
xr (i) yr (i) the inverse kinematic solution of the 7DOF robotic arm is
x(i) = , y(i) = solved by a swarm intelligent optimization algorithm,
max(| xr |) max(| yr |)
which converts the inverse kinematic solution problem into
xr (i ) = r (i) sin( (i)), yr (i) = r (i) cos( (i)) a multi-objective optimization problem for multiple joint
angles. By performing the positive kinematic analysis of
the robotic arm, the position pose matrix of the end-
(i) = a rand , r (i) = (i) + R rand effector of the robotic arm is obtained. Based on the
obtained positional attitude matrix, an adaptation function
In Eq. (2), Pi , new denotes the position of the ith bald
with seven joint angles as variables is established, and the
eagle after the selection of the search space. Pi denotes
group intelligent optimization algorithm obtains a set of
the position of the ith bald eagle. a denotes the angle
inverse kinematic solutions satisfying the conditions by
between the bald eagle and the search prey during the spiral
finding the minimum value of the adaptation function. The
movement, and takes values in the range [5, 10]. R is
problem is described mathematically as shown in Eq. (4).
used to determine the number of search cycles, and takes
values in the range [0.5, 2]. Pi +1 denotes the position of minf (1 ,...,7 ),i [i min ,i max ](i = 1 ~ 7) (4)
the ith + 1 bald eagle. Pmean denotes the the average value
of the location of the bald eagle. In Eq. (4), f denotes the fitness function and i
denotes the i-th joint angle to be solved.
3.3. Swoop phase
4.2. Fitness function
During the swoop phase, the Bald Eagle moves from
the optimal search space to the target prey at a very fast After clarifying the problem to be solved, the form of
speed. Eq. (3) represents this behavior mathematically. the fitness function is further given in this paper, as shown
Pi , new = rand Pbest + x1 (i) ( Pi − c1 Pmean ) + y1 (i) ( Pi − c2 Pbest ) (3) in Eq. (5).
xr (i) yr (i)
x1 (i) = , y1 (i) = f = ( px − p 'x )2 + ( py − p ' y )2 + ( pz − p 'z )2 (5)
max(| xr |) max(| yr |)
In Eq. (5), f denotes the fitness function,
xr (i ) = r (i ) sinh[ (i)], yr (i) = r (i) cosh[ (i)] px,p y,pz represents the position value in the x, y, and z
direction obtained by the swarm intelligence optimization
(i) = a rand , r (i ) = (i ) algorithm, and p 'x ,p ' y ,p 'z denotes the position value
in the x, y, and z direction expected to be reached by the
In Eq. (3), Pi , new denotes the position of the ith bald end-effector of the robot arm, respectively.
eagle after moving toward the target prey. Pbest denotes
4.3. Algorithm flowchart and pseudo-code
the best position of the bald eagle in the current search
space. a denotes the angle between the bald eagle and To further illustrate the BES algorithm, the flowchart
the searched prey during the spiral movement, taking and pseudo-code for solving the 7DOF inverse kinematic
values in the range [5, 10]. Pi denotes the position of the solution by the Bald Eagle swarm intelligent optimization
ith bald eagle. c1 , c2 is a parameter taking values in the algorithm are given in this paper, as shown in Fig. 5 and
range [1,2]. Pmean denotes the average value of the bald Table 4, respectively.
eagle's position.
4. For (Each set of joint angles of n sets of joint angles)
Start
5. Calculate the fitness values of joint angle: f ( )
Calculate the fitness of Joint angle 10. Update the fitness values according to Eq. (2)
11. End for
Execute the selection phase Swoop phase
according to Eq. (1) 12. For (Each set of joint angles of n sets of joint angles)
5. Experiment
Calculate the fitness of Joint angle
In this section, the BES algorithm is compared with
the conventional swarm intelligence optimization
Execute the Swoop phase algorithms DE algorithm, FA algorithm, FOA algorithm,
according to Eq. (3) GA algorithm, and PSO algorithm to solve the inverse
kinematics of YuMi 14000 ABB 7DOF industrial robot
Obtain the global optimal solution arm and S-R-S humanoid 7DOF robotic arm in terms of
for the joint angle position accuracy experimentally. The experiments firstly
obtained the working space of the two robotic arms based
on the range of values of the joint angles of the two robotic
End arms. They are shown in Fig. 6 to Fig. 9 and Fig. 21 to Fig.
24, respectively. Then, a continuous trajectory containing
Fig. 5 BESAlgorithm flow chart 100 points was selected based on the working space of the
two robotic arms. They are shown in Fig. 10 and Fig. 25,
respectively. The robotic arm end-effector position pose
Table 4 BES algorithm pseudo-code matrices corresponding to 100 points on the two
Algorithm: Solving 7DOF robotic arm inverse continuous trajectories were brought into the swarm
kinematic solutions based on BES algorithm intelligence optimization algorithm to obtain the new
continuous trajectories as shown in Fig. 11 to Fig. 16 and
1. Initialization of the joint angle according to the
Fig. 26 to Fig. 31, respectively. The position solution
range of values of the joint angle, accuracy of the algorithm is initially compared by
2. Calculate the fitness values of initial joint angle: comparing the degree of agreement between the trajectory
obtained by the algorithm and the original trajectory. In
f ( ) ,
order to further compare the solution accuracy of the
algorithms, experiments are presented for the
3. For (Iterate within a set number of iterations)
transformation of the fitness function values and the
Selection phase position errors in X, Y, and Z directions during the
iterations of the two robotic arms, as shown in Fig. 17 to industrial robot arm workspace workspace in the XY plane
Fig. 20 and Fig. 32 to Fig. 35, respectively. Also
experiments were conducted to compare the mean, median,
minimum and maximum values of position error values in
X, Y and Z directions for both robotic arms, the
comparison results are shown in Table 6 to Table 11 and
Table 12 to Table 17, respectively. The experiments on
both robotic arms were conducted on a computer
configured with "Intel(R) Core(TM) i9 CPU, 3.19 GHz". Fig. 8 Projection of the Fig. 9 Projection of the
The specific parameter settings for each of the algorithms workspace in the XZ plane workspace in the YZ plane
are shown in Table 5.
After the workspace of the robot arm was determined,
Table 5 Specific parameter settings for various algorithms a continuous trajectory was selected based on the
Algorithm BES DE FA FOA GA PSO workspace, and the position accuracy of the inverse
Population 150 150 150 150 150 150 kinematic solutions obtained by the algorithms was
iteration 500 500 500 500 500 500 initially compared by comparing the degree of agreement
c1 / c2 1.5/1.5 - - - - 1.4/1.4 between the trajectories obtained by various algorithms
R/a 1.3/8 - - - - - and the original trajectories. The original trajectory is
1.8 - 0.2 - - - shown in Fig. 10, and the comparison plots of the
0 / - - 2/1 - - - trajectories obtained by BES, DE, FA, FOA, GA and PSO
v / - - - - - 0.5x/0.9 algorithms with the original trajectory are shown in Fig. 11
F0 - 0.4 - - - - to Fig. 16, respectively.
CR - 0.1 - - - -
pc - - - - 0.6 -
pm - - - - 10-4 -
Fig. 13 Comparison of the trajectory obtained by FA algorithm Fig. 16 Comparison of the trajectory obtained by PSO algorithm
with the original trajectory with the original trajectory
Fig. 30 Comparison of the trajectory obtained Fig. 31 Comparison of the trajectory obtained
by GA algorithm with the original trajectory by PSO algorithm with the original trajectory
According to the comparison results shown in Fig. 26 positional accuracy compared to the other algorithms. The
to Fig. 31, it can be found that the trajectories obtained by mean, median, minimum and maximum values of the
the BES algorithm, DE algorithm and PSO algorithm position errors in the x, y and z directions in the process of
match the original trajectories to a greater extent than the solving the inverse kinematic solutions of the trajectory
other three algorithms, so it can be preliminarily stated that points for the BES, DE, FA, FOA, GA and PSO algorithms
the inverse kinematic solutions obtained by the BES are also given in the experiments, as shown in Table 12 to
algorithm, DE algorithm and PSO algorithm have higher Table 17, respectively.
According to the results shown in Table 12 to Table algorithm has the highest position solving accuracy with
17, it can be found that for the average value of position an error of 0. The PSO algorithm has the next highest
errors in the x, y, and z directions, the BES algorithm has position accuracy with an error of 10 -8. The error range of
the highest position solution accuracy with an error range the other four algorithms is 10-4 to 10-2. For the minimum
of 10-18 to 10-16. The PSO algorithm has the next highest value of position error in x, y, and z directions, the BES
position accuracy with an error of 10 -7. The error range of algorithm has the highest position solving accuracy with
the other four algorithms is 10 -3~10-2. For the median of an error range of 0 to 10-16. The PSO algorithm has the next
the position errors in x, y, and z directions, the BES highest position accuracy with an error of 10 -6. The error
range of the other four algorithms is 10 -2 to 10-1. For the
maximum values of position errors in x, y, and z directions,
the BES algorithm has the highest position solution
accuracy with an error range of 10 -16 to 10-15. The PSO
algorithm has the next highest position accuracy with an
error range of 10-7 to 10-5. The error range of the other four
algorithms is 10-2 to 10-1. The experiments also compare
the changes of the fitness function, X-direction position
error, Y-direction position error, and Z-direction position
error during the iterative process of solving the inverse
kinematic solution of the first point of the continuous
trajectory for the BES, DE, FA, FOA, GA, and PSO
algorithms, as shown in Fig. 32 to Fig. 35, respectively.
5.3 Discussion