Proceedings of the 2015

IEEE Conference on Robotics and Biomimetics

Zhuhai, China, December 6-9, 2015

A Robotic Off-line Programming System Based on SolidWorks

Hongmin Wu, Huajian Deng, Cao Yang, Yisheng Guan∗ , Hong Zhang and Hao Li

Abstract— It is well known that off-line programming (OLP) and research institutions. The OLP systems from the first
is an efficient control mode for industrial robots. However, two sources are usually commercial and very expensive.
OLP is not yet commonly and widely used in applications, Therefore, many institutions developed their own systems
since commercial OLP systems are very expensive, with prices
even much higher than those of robotic systems themselves. based on CAD software or toolkits, such as AutoCAD,
Simple but effective OLP systems are desired, and such systems SolidWorks, OpenGL and VRML. Typical OLP softwares
may be made based on other commonly used CAD systems. include Tecnomatix Robcad, DELMIA D5/V5, Robotworks
We have developed a robotic OLP system, which we called and so on.
RobSim, based on SolidWorks and with Microsoft Visual Tecnomatix Robcad enables design, simulation, optimiza-
Studio 2010. RobSim is developed as an add-on tool for the
commonly used CAD software SolidWorks. With this OLP tion, analysis, off-line programming of multi-devices and
system, an object can be made, imported or modified in the manufacturing process for its production resources [4]. With
conventional mode in SolidWorks environment, and various Robcad, manufacturers are able to achieve desired manufac-
trajectories for a robot can be easily and conveniently defined turing ideas to practical automatic manufacturing process.
and modified in the same environment. Trajectory interposition, DELMIA is a widely used software system combining
kinematic computation, dynamic and graphic simulation can
be conducted. Finally executable codes can be generated for robotics and manufacturing for industrial applications [5].
the robot to perform tasks. In this paper, the development In a 3D simulation environment, DELMIA can test and
of RobSim is demonstrated. Specifically, the architecture of optimize the robot programs. The outstanding features of
RobSim, the method for extracting position and orientation DELMIA are the ”toolboxes”, which are available to pro-
from a pre-defined path on an object for the robot tool, and path gram with numerous functions. These functions are useful in
transformation, are presented. Simulation and experiments are
also conducted to verify the effectiveness of the OLP system. various special areas, such as target definition, reachability
analysis, collisions testing, path planning, etc. RobotWorks
I. I NTRODUCTION is a robotic CAD system running on SolidWorks, especially
useful for workcell design, motion simulation and path pro-
As well known, robots play important roles in modern gramming [6]. These function modules use the SolidWorks
industrial production. Motion planning and control of a robot graphics and internal engine to create a robot path quickly
is one of the basic problems to be solved for industrial appli- along faces, edges and curves of CAD objects.
cations such as welding, polishing, painting, carving and so Other OLP softwares developed by robot manufacturers
on. Compared with teaching-and-playback method, off-Line include RobotStudio (ABB), KUKA.Sim Pro (KUKA) and
Programming (OLP) is a more effective and efficient mode MELFA Works (Mitsubishi) etc. RobotStudio provides vari-
to plan and control motion of robots for many industrial ous tools to increase efficiency of manufacturing tasks such
applications, especially for those tasks on complex surfaces as welding, polishing, teaching and multi-devices coordinate
or along complicated trajectories, as aforementioned. A OLP motion planning [7]. KUKA.Sim Pro is developed for off-line
system allows engineers to simulate production first in a programming of KUKA robots, which is connected to the
visual graphic environment, and then control a real robot virtual KUKA controllers, equipped with cycle time analysis
to implement the physical motion with the control system and the generation of robot codes [8]. It can upload CAD
[1], [2], [3]. models from other CAD softwares.
OLP systems mainly come from three sources, which are MELFA-Works [9] is an add-on tool for SolidWorks
professional robotic software producers, robot manufacturers products, which can automatically generate robot codes from
existing CAD models with little installation and commission-
All authors but H. Li are with the School of Mechanical and Electri-
cal Engineering,Guangdong University of Technology, Guangzhou, China, ing time. Developed by research institutions as educational
510006. Y. Guan is the corresponding author, email: [email protected]. robotics software, OLP systems are usually open source for
H. Zhang is also with the Department of Computing Science, University academic purpose. References [10], [11] presented efficient
of Alberta, Edmonton, AB, Canada, T6G 2E8. H. Li is with Guangdong
Inspection and Quarantine Technology Center, China. robotic OLP methods based on DXF files of 3D modeling
∗ The work in this paper is supported by the National Natural Science softwares AutoCAD. They decodes DXF files that describe
Foundation of China (Grant No. 51375095), the NSFC-Guangdong Joint the 3D position information of graphic elements such as
Fund (Grant No. U1401240), the Natural Science Foundation of Guang-
dong Province (Grant No. S2013020012797, 2015A030308011), the State points, lines, arcs and curves, then transforms into cartesian
International Science and Technology Cooperation Special Items (Grant No. values. Similar approaches were also proposed with Solid-
2015DFA11700), the Frontier and Key Technology Innovation Special Fund- Works API SDK package [12], [3]. A robot simulation system
s of Guangdong Province (Grant No. 2014B090919002, 2015B010917003),
and the State General Administration for Quality Supervision, Inspection based on OpenGL with VC++6.0 was presented in [13].
and Quarantine (Grant No. 2014IK186). Although there are many robotic simulation and OLP

978-1-4673-9675-2/15/$31.00 © 2015 IEEE 1711

systems, most of them are not accessible. Those developed SldWorks
CustomPropertyManager
by professional software producers or robotic manufactures MathPoint
are usual of high prices, and those by institutions are Environment
MathTransform
not open to other organizations. And some systems lack Frame
MathUtility
of sufficiently excellent 3D modeling and path extracting AttributeDef
MathVector
functions. In this paper, we investigate how to develop an Modeler
effective OLP system based on commonly and widely used ModelDoc2
software tools such as SolidWorks and Microsoft Visual PartDoc AssemblyDoc DrawingDoc
Studio 2010. The system, which we called RobSim, includes
mainly three function modules: which are path processing Other Objects: Feature
(extracting, optimizing and visualizing), motion simulator, Parameter
ModelView
and executable code generator. The SolidWorks API and Entity
Sketch
RobSim architecture are introduced in Section II, and math- EnumLoops2
ematic basis for three sorts of graphic path elements (line, Modeler
MotionStudy
arc and bezier curve) presented in Section III. Section IV SelectionMgr
describes path processing, including configuration extraction
and transformation from pre-defied paths, and singular con- Fig. 1. SolidWorks API interface inheritance hierarchy diagram
figuration avoidance with robotic kinematics. The simulation
and experiment are conducted in Section V, and conclusions
are drawn in the last section. B. RobSim Architecture
There are no existing add-on or toolbox in SolidWorks
II. S YSTEM A RCHITECTURE for robotic kinematic, dynamic, motion planing and so on.
RobSim is built as a robotic add-on to SolidWorks, by adding
For friendly interaction with common users, excellent GUI graphics, kinematics engine and other modules. Solid object
and function modules layout of a robotic OLP system are models can be directly accessed in SolidWorks graphics
required. Unfortunately, many of current OLP systems de- window, so that hidden links of objects are bridged between
veloped with OpenGL, based on AutoCAD or others, are not two software blocks, and objects motion is driven inside
sufficiently convenient and even not capable to achieve these the SolidWorks assembly document. In this way, complex
sound effects. OpenGL is good for circumstance rending and kinematics calculation and motion planning process can be
visual simulation of simple mechanical parts, but difficult for added in RobSim. Fig.2 shows the architecture of RobSim,
complex ones. AutoCAD is an excellent and professional which clarifies interacted relationship between RobSim and
CG modeling software tool for plane drawing, but weak in SolidWorks.
3D modeling. As a OLP system, powerful 3D modelling
RobSim GUI
and simulation are required, and significant features of a Path Extracting Pose Visualization Teach Data Import/Export
robot are included: a) robot configuration can be defined
Robot Switch Tool Settings Point Modify Extral Axis Connection
as a multi-rigid bodies mechanical system; b) different part
are constrained by kinematic pairs; and c) joint motions are
translational or rotational according to kinematic pairs. For Microsoft Visual Studio 2010
Path Elements
these requirements, the CAD software tool SolidWorks is a
good choice for developing a OLP system with its abundant SolidWorks
Planning & Analysis
3D Model
Application Programming Interface (API). Kinematic Solver
Y Error Exsiting ?
Path Generating
N
A. SolidWorks API
Data Generation Configuration Settings
Assembly Mate
As an outstanding CAD software system, SolidWorks has
excellent performance in 3D modeling, visualization and Motion Simulation
& Joint Analysis Joint Optimization
simulation, with rich API containing hundreds of functions
for Visual Basic 6.0, Visual Basic for Application (VBA), Meet All Requirements ? N Reset Parameters
Microsoft Visual studio⃝ R
2010. The functions belong to
Y
some predefined object models and provide direct access to Robot Executiving File Robot Controller Realistic Robot
SolidWorks functions, such as creating a plane, extruding a
body, inserting an existing part into a assembly document Error Calibration
and verifying the parameters of a surface. Partial graphical
hierarchy diagram of the interface inheritance is shown Fig. 2. Relationships between SolidWorks and RobSim
in Fig.1. SolidWorks⃝ R
2012 SP0 and Microsoft Visual
⃝R
studio 2010 are chosen to develop our system. The open The following functional modules are shown in the archi-
source API of SolidWorks is employed in RobSim. tecture diagram.

1712
 • 3D Model Builder: This module builds the 3D CAD A. Line
models of parts, robots and their environments with A line segment can be defined by two points (the start
SolidWorks. CAD models can also be made with other point and the end point). In order to discretize the line,
3D CG tools, like Pro/E and UG. Many CAD models of we can commonly parameterize the linear equation with a
commercial industrial robots can be found in websites. −
→
point (x0 , y0 , z0 ), a direction vector d = (m, n, p) and the
The CAD models are collected in a robotic library of parameter ti . Therefore, An arbitrary point pi (xi , yi , zi ) on
RobSim. End-effectors or other accessories need to be the line can be described by the following equations
designed or imported by users, and assembled with the
robot models in the SolidWorks assembly document xi = x0 + ti m, xi = y0 + ti n, zi = z0 + ti p (1)
environment. Discrete interposition points are thus simply obtained with
• Path Generator: Path generating includes path drawing the above equations. However, orientation information for
and path extracting from entities. Line elements in dif- a regular robotic path is needed. Therefore the orientation
ferent styles, like component edges, sketches, or spline information along the line is to be extracted. According to
curves, can be created with SolidWorks UI commands. geometry, spatial edge of a body lies on two adjacent faces
As robot paths, the line elements will be planned A1 and A2 , that is, the intersection of them. The normal
into motion sequence by functions of RobSim. In this vectors −→ = (u , v , w ) and −
n 1 1 1 1
→ = (u , v , w ) of the
n 2 2 2 2
process, the graphic models always play important role two surfaces are obtained, respectively. Then, an appropriate
between SolidWorks and RobSim. coordinate system with the line direction vector and one of
the surface normal vectors can be built up, as shown in Fig.3.
With one of the surface normal vectors − →, the configura-
• Kinematic Solver: This module defines the mapping
n 1
between the joint space of a robot and the Cartesian s-
tion Ti at the i-th point pi on the line can be defined by the
pace of the end-effector. Forward kinematics and inverse
following homogeneous matrix.
kinematics of robots are established by this module.
• Configuration Visualizer: After path calculation, con- n
figurations (positions and orientations) at discrete path A1
points can be visualized with a series of short coordinate (x0, y0, z0) Ti d
arrows. Visualization is achieved using the SolidWorks
OpenGL graphical interface. A2

• Motion Simulator: There are two kinds of motion de-

veloped in RobSim, which are motion simulation of Fig. 3. Spatial line expression
the end-effector and that of the manipulator. With the
help of Motion Analysis in SolidWorks, the profile
charts of the joint displacement, joint velocity, and joint [ −
→ −→×− → p ]
→ −
Ti = d n 1 d n 1 i
acceleration can be rapidly exported. These profiles 0 0 0 1
hence make motion simulation available and convenient.  
• Error Calibrater: This module is to find out the errors m pv1 − nw1 u1 xi
 n mw1 − pu1 v1 yi 
between the simulated environment and the real world. =


This calibration processes relative accuracy and absolute p nu1 − mv1 w1 zi 
accuracy. The workpiece frame and the tool frame with 0 0 0 1
respect to the robot base frame need to be calibrated. B. Arc
Implementation algorithms are integrated in RobSim. Arc edge of 3D model in SolidWorks can be extracted in
• Robot Controller: This module is to download exe- a similar manner. The start point s(sx , sy , sz ), the end point
cutable codes generated by RobSim into the realistic e(ex , ey , ez ), the circular center o(ox , oy , oz ), the normal
robot controller. Reading the executable codes, the vector −→n = (u, v, w) and the radius r are accessible from
realistic robot conduct the desired motion. the SolidWorks API function, as shown in Fig.4. Then, if
incremental value ∆i is set, any point pi (xi , yi , zi ) on the
arc can be calculated by the following equations
v1 = sx − ox + ∆i ∗ (ex − sx )
III. M ATH BASIS FOR G RAPHIC E LEMENTS v2 = sy − oy + ∆i ∗ (ey − sy )
v3 = sz − oz + ∆i ∗ (ez − sz )
To ensure that the robot end-effector can smoothly move a- R
long the paths extracted from the SolidWorks, transformation kt = √ 2 (2)
v1 + v22 + v32
and interposition must be executed [14], [15]. In this section,
xi = ox + kt ∗ v1
robotic rules for three graphic elements (line, arc and Bezier
curve) are introduced. According to the rules, mathematical yi = oy + kt ∗ v2
models of these graphic elements are established. zi = oz + kt ∗ v3

1713
where v1 , v2 , v3 and kt are intermediate variables. with respect to u:
( )
dp(u) ∑
n
n
n = i ui−1 (1 − u)n−i pi −
du i
i=0
s O d e ∑n ( )
r n
(n − i) ui (1 − u)n−i−1 pi
i
pi i=0
∑n ( )
n−1
=n ui (1 − u)n−i−1 pi (pi+1 − pi )
i
i=0
(4)
Fig. 4. Spatial arc expression
−→
The tangent vector tan = (tx , ty , tz ) at an arbitrary tessel-
For the orientation, the coordinate system can be obtained lation point can be obtained. To set up a coordinate system,
according to the geometrical relationship between the point another vector containing the corresponding point must be
pi and the circular center O (refer to Fig.4), which is defined found. For this end, we define a vector − →
n = (nx , ny , nz )
→
− pointing from the current tessellation point pi (xi , yi , zi ) to
as d = (m, n, p). Hence, the configuration Ti at the i- −→
th point pi on the arc can be defined by the following the next one pi+1 (xi+1 , yi+1 , zi+1 ). Since the vectors tan
−
→
and n intersect at point pi , the cross product result is
homogeneous matrix.
→
− −
→
a =− →n × t
From above discussion, a Bezier curve is specifically
[ −
→ − −
→ → ]
Ti = d →
n × d −
n pi shown in Fig.5. With the position and orientation at a
0 0 0 1 tessellation point pi of Bezier curve, we can derive the
following homogeneous matrix

  a
m pv − nw u xi
 n mw − pu v c4
yi  c3 o
=
 p nu − mv w
 c2
zi  pi t c7
pi+1
0 0 0 1 c1 (p1 ) n
c6
c5
C. Bezier Curve

Bezier curve is defined by the vertices of a polygon that Fig. 5. Spatial Bezier curve expression
encloses the resulting curve [16]. The effects of the ver-
tices are weighted by the corresponding blending functions
[ −
→ ]
n −
→
n ×−→
a −
→
and blended as in the Hermite curve. We can derive the
a pi
controlling vertices cj of the controllable polygon by the Ti =
0 0 0 1
SolidWorks API functions. The equation of the Bezier curve  
can be obtained as nx ny az − nz ay ax xi
 ny nz ax − nx az ay yi 
n ( ) =
 nz

∑ n nx ay − ny ax az zi 
p(u) = ui (1 − u)n−i pi 0≤u≤1 (3) 0 0 0 1
i
i=0

where IV. PATH T RANSFORMATION

( )
n n! For configuration transformation, Denavit-Hartenberg (D-
= H) method [17] uses the four parameters (a, α, d, θ) asso-
i i!(n − i)!
ciated with a particular convention for attaching reference
frames to the links of a spatial kinematic chain. The matrix
Now we can verify that the Bezier curve defined by the
of configuration transformation from coordinate frame n to
above equation satisfies our target to tessellate the curve by
coordinate frame n − 1 is given by the following equation.
parameter u. Through the tessellation points, we can obtain
the interposition point pi (xi , yi , zi ) of Bezier curve.
 
For the orientation, we need to derive the expression of cθn −sθn 0 an−1
the Bezier curve derivatives. This expression described the  sθn cαn−1 cθn cαn−1 −sαn−1 −dn sαn−1 
n−1
T =
 sθn sαn−1

relation between the derivatives of Bezier curve and their n cθn sαn−1 cαn−1 dn cαn−1 
original control points. We then differentiate the equation 0 0 0 1

1714
 X Axis
importing, path defining and extracting, tool installation,
Y Axis
Z Axis simulation, downloading executable codes to robot controller
F F
TT and running. The performance of the OLP system is satisfied.
Tool
T
B
P T
FT

P
P
T
W

B W
BT W

Fig. 6. Coordination transformation

(a) CAD model of the workpiece (b) Tool pose along the paths

Fig. 7. Workpiece model and pose visualization

where cθn is a short note for cos θn , and sθn for sin θn .
During path planning, path transformation and workpiece
calibration are in need. Fig.6 shows the relationship between
the frames at a discrete path point, the workpiece and the Sucker Path
3D Model Location Generating
robot base, where five transformation matrixes, B F T
FT, T T, PT,
P W B SolidWorks
W T , B T , are to be calculated. F T represents the transfor-
mation from the robotic wrist frame to the base frame, which RobSim
can be gained by D-H method according to the link and joint
Path
parameters. FT T is calculated after a Tool Center Point (TCP) Exertacting
frame is defined on the tool. TP T is the transformation from
the path point frame on the workpiece to the robot end- Tool Robot ...
effector frame. PW T describes the transformation from the Configuration Pose Path Planning
world frame to the the path point frame. And W B T describes Settings Visualization & Calculation
the relationship between the robot base frame and the world
frame. The following equation can be obtained.
B F T −1
FT TT PT = (P
WT
W
B T) (5)
Finally, we can calculate the inverse kinematics from Carte-
sian space (ℜ) to joint space(Θ) for path following,
Kinematic Code Generating
Θ = IK(T ) (6) Motion Analysis
Modeling & Animation
V. S IMULATION AND E XPERIMENT
To verify the effectiveness of the presented method above
and demonstrate the functions of OLP system RobSim, simu- realistic Robot Calibration
lation and experiment are conducted in this section. The task
is to control the robot end-effector to follow curves on the
cover of our suction module as shown in Fig.7(a), to mimic
application of welding along complex paths. The paths are Fig. 8. Key steps of software implementation
specified directly on the CAD model of the workpiece (the
suction module cover), and the poses of the robotic tools
are then defined along the paths and visualized, as shown in VI. C ONCLUSIONS AND F UTURE W ORK
Fig.7(b). The key steps of the software implementation are As presented in this paper, a robotic off-line programming
presented in Fig.8. (OLP) system, RobSim, has been developed. Based on
Motion of the robot is simulated first with the RobSim SolidWorks, the presented OLP system has been built with
system. Once the simulation is successful, executable codes Visual Studio 2010, and can be used as an added-on tool
are generated and downloaded into the robot controller for in SolidWorks. Like other commercial OLP systems, it has
real implementation. Fig.9 (a)-(d) are some snapshots of the basic OLP functions such as 3D object modeling, definition
simulation in the SolidWorks and the experiment with a real and modification of complex trajectories for manipulators,
robot. Fig.10 is shown the main several modules of RobSim position and orientation extraction and interpolation, simula-
in the SolidWorks environment. It takes less than thirty tion, executable code generation. Simulation and experiment
minutes to finish the whole experiment, including 3D model have verified the effectiveness of the built simple OLP

1715
 Fig. 10. Global view of RobSim in SolidWorks software

Proceedings of the 24th Annual Conference of the IEEE Industrial

Electronics Society, vol. 1, pp. 208–213, 1998.
[3] Y. Gan, X. Dai, and D. Li, “Off-line programming techniques for
multirobot cooperation system,” International Journal of Advanced
Robotic Systems, vol. 10, 2013.
[4] S. Wang, J. Zhang, H. Cai, Y. Wan, and Y. Fu, “An advanced robot
simulation system-robcad,” Journal of Harbin Institute of Technology
(in Chinese), vol. 3, p. 021, 1993.
[5] Z. Bzymek, M. Nunez, M. Li, and S. Powers, “Simulation of a
machining sequence using delmia/quest software,” Computer-Aided
Design and Applications, vol. 5, no. 1-4, pp. 401–411, 2008.
[6] A. R. MFF, W. Mansor, W. Muhamad, S. Mohd Zaidi, et al.,
“Design and simulation of robotic spot welding system for automotive
(a) (b) (c) (d) manufacturing application,” 2005.
[7] “Abb robotics: Operating manual robotstudio,” Västerås, Sweden,
2007.
Fig. 9. Simulation and Experiment snapshots [8] K. Vollmann, “A new approach to robot simulation tools with para-
metric components,” in ICIT’02, IEEE International Conference on
Industrial Technology, vol. 2, pp. 881–885, 2002.
[9] V. Bottazzi and J. Fonseca, “Off-line programming industrial robots
system and illustrated its potential application of welding. based in the information extracted from neutral files generated by the
Other potential applications such as painting, deburring, commercial cad tools,” Industrial Robotics: Programming, Simulation
polishing, and carving may be also implemented with this and Applications, p. 349, 2006.
[10] Z. Yin, Y. Guan, et al., “Off-line programming of robotic system based
OLP system. We believe that the architecture of RobSim, the on dxf files of 3d models,” in ICIA’13, IEEE International Conference
presented method for extracting and interpolating positions on Information and Automation, pp. 1296–1301, 2013.
and orientations from the pre-defined trajectories are helpful [11] P. Neto, J. N. Pires, et al., “Robot path simulation: a low cost solution
based on cad,” in RAM’10, IEEE Conference on Robotics Automation
to develop similar OLP systems using other commonly used and Mechatronics, pp. 333–338, IEEE, 2010.
CAD softwares. [12] S. Mitsi, K.-D. Bouzakis, G. MansouG, D. Sagris, and G. Maliaris,
However, RobSim is not yet a complete LOP system. It “Off-line programming of an industrial robot for manufacturing,” The
International Journal of Advanced Manufacturing Technology, vol. 26,
still lacks of many important functions or features such as no. 3, pp. 262–267, 2005.
collision detection in simulation and on-line calibration. In [13] W. Zhang, F. Zou, D. Qu, and F. Xu, “Research of key technologies on
the near future, these functions will be implemented and 3d simulation system of industrial robot,” in The 7th World Congress
on Intelligent Control and Automation, pp. 565–568, 2008.
added to RobSim. Besides, the system functions will also [14] K. Hirasawa, Y. Shimizu, and T. Matsumura, “Time-based interpola-
be extended to integrate coordination of multiple robots. tion control of a robot,” July 28 1987. US Patent 4,683,543.
[15] H. Koyama, T. Asano, F. Noguchi, and S. Fujinaga, “Robot interpo-
R EFERENCES lation control method,” Aug. 25 1987. US Patent 4,689,756.
[16] K. Lee, Principles of CAD/CAM/CAE systems. Addison-Wesley
[1] Y. Yong, J. Gleave, J. Green, and M. Bonney, Off-line programming Longman Publishing Co., Inc., 1999.
of robots. Wiley, 1985. [17] R. M. Murray, Z. Li, S. S. Sastry, and S. S. Sastry, A mathematical
[2] E. Freund, D. Rokossa, and J. Roßmann, “Process-oriented approach to introduction to robotic manipulation. CRC press, 1994.
an efficient off-line programming of industrial robots,” in IECON’98,

1716