Human-Robot Collaboration (HRC) with Vision Inspection for PCB

Assembly

Mauro Queirós

1,2 a

, João Lobato Pereira

1,2 b

, Nuno M. C. da Costa

, S. Marcelino

José Meireles

, Jaime C. Fonseca

, António H. J. Moreira

4,5 f

and João Borges

2,5 g

University of Minho, Guimarães, Portugal

Algoritmi Center, University of Minho Guimarães, Portugal

MEtRICs Research Center, University of Minho, Guimarães, Portugal

2Ai – School of Technology, IPCA, Barcelos, Portugal

Polytechnic Institute of Cávado and Ave, Barcelos, Portugal

DIB4T, Marinha Grande, Portugal

Keywords: Cobot, PCB Assembly, Work Cell.

Abstract: Flexibility and speed in the development of new industrial machines are essential factors for the success of

capital goods industries. When assembling a printed circuit board (PCB), since all the components are surface-

mounted devices (SMD), the whole process is automatic. However, in many PCBs, it is necessary to place

components that are not SMDs, called pin through-hole components (PTH), having to be inserted manually,

which leads to delays in the production line. This work proposes and validates a prototype work cell based on

a collaborative robot and vision systems whose objective is to insert these components in a completely

autonomous or semi-autonomous way. Different tests were made to validate this work cell, showing the

correct implementation and the possibility of replacing the human worker on this PCB assembly task.

1 INTRODUCTION

In today’s industry, any manual operation in

repetitive production lines that causes delays in the

production process needs to be semi-automated or

fully automated so that the factory remains

competitive in an increasingly aggressive and fast

market. Printed circuit boards (PCBs) are a fine

example of such necessity, as they are in virtually all

electronic components of everyday life. Over the

years, assembling electronic components on PCBs

has undergone several changes, many of which aim to

make this process increasingly automated, efficient,

fast, and economically efficient (Altinkemer, Kazaz,

Köksalan, & Moskowitz, 2000; Andrzejewski,

Cooper, Griffiths, & Giannetti, 2018; Bogner,

Pferschy, Unterberger, & Zeiner, 2018; Crama,

https://orcid.org/0000-0001-5896-7423

https://orcid.org/0000-0002-1053-7405

https://orcid.org/0000-0002-8425-3501

https://orcid.org/0000-0003-0881-2348

https://orcid.org/0000-0001-6703-3278

https://orcid.org/0000-0002-2148-9146

https://orcid.org/0000-0002-5880-033X

Flippo, Van De Klundert, & Spieksma, 1997). This is

how Surface Mount Technology (SMT) arose, with

surface mount devices (Surface Mount Devices)

being mounted directly on the surface of the printed

circuit board, generally in an automatic way using an

SMT machine. Typically, these components are

smaller in size, have better electrical performance,

must withstand higher soldering temperatures, and

must be selected, positioned, and soldered more

carefully to achieve an acceptable manufacturing

yield (Iftikhar et al., 2020).

Despite the advances seen with SMDs, there are

other components (Figure 1) that still need to be

manually assembled on PCBs, such as diodes,

capacitors, and connectors named Pin Through Hole

(PTH) components (Wendy Jane Preston, 2018). This

limitation considerably decreases the production

efficiency of PCBs assembled with this type of

Queirós, M., Pereira, J., C. da Costa, N., Marcelino, S., Meireles, J., Fonseca, J., Moreira, A. and Borges, J.

Human-Robot Collaboration (HRC) with Vision Inspection for PCB Assembly.

DOI: 10.5220/0011526200003329

In Proceedings of the 3rd International Conference on Innovative Intelligent Industrial Production and Logistics (IN4PL 2022), pages 56-65

ISBN: 978-989-758-612-5; ISSN: 2184-9285

components when compared to those that only have

SMD components.

Figure 1: Example of non-SMD components.

Aiming to mitigate the limitation mentioned above

and take advantage of the rise of collaborative robots

(cobots) (Poór, Broum, & Basl, 2019; Vojić, 2020),

this work proposes a novel industrial work cell for the

automatic insertion of non-SMD components into

PCBs. The proposed cell integrates three main

systems from the literature: (1) a cobot to enable either

collaborative or autonomous work (Robotics, 2018);

(2) a vision system for component validation (David

A. Forsyth, 2002; Kurka & Salazar, 2019); and (3) an

external device for controlling and monitoring the

work cell (Langmann & Rojas-Peña, 2016).

Througout this paper, are going to be presented

the fundamental requirements of this work cell, the

layout and the main workflow of this task. It is also

explained the key aspects of each subsystem

methodology, and to validate this work cell, three

tests were made and demonstrated. Finally the results

and main ideas drawn throughout this work are

discussed and concluded.

2 REQUIREMENTS OF THE

WORK CELL

Taking into consideration the goal of reducing or

eliminating the need for manually inserting different

electronic components into PCBs but maintaining

quality and efficiency during the insertion process, a

set of requirements were defined, of which the most

relevant are the following:

• The cell must have an automatic transport

system for the PCB boards;

• It must have the ability to identify the different

PCBs;

• Adapt the program and the gripper according

to the PCB and components to be assembled;

• It must perform accurate and secure gripping

and insertion for the different components;

• The system needs to validate each component

prior to insertion, using, for example, vision

software to read the number of pins and make

sure it is in good condition;

• Needs to assure that the PCB is correctly and

fully assembled within 30 seconds;

• It also requires collecting information relevant

to the operation of the work cell and making it

available in a database for future analysis and

study.

2.1 Cobot Subsystem

Going into more detail about the robot subsystem and

taking into consideration aspects such as the

components under study and the PCBs in question,

the requirements of the robot are: speed and

flexibility, as well as the ability to operate in

conjunction with a human being; at least 700 mm

range; ate least 6 degrees of freedom (DOF); accuracy

and repeatability of less than 0.1 mm; multiple I/O

interfaces; and support for various communication

protocols.

2.2 Vision System

To validate the quality of each component, one of the

most frequently used approaches in the industry is the

use of a vision system. In this case, for the proper

functioning of the work cell, a minimum sensor

resolution of 2M pixels is required to cover the entire

insertion area of the component (500x500 mm2) and

still allow accurate hole inspection and pin integrity

verification. In addition, this system must permit the

sharing of data with other subsystems.

The system must also be able to read the barcodes

in the PCBs and inspect their correct position to

reference the robot with respect to it doing the correct

insertion of all components. For these tasks, the

camera used may have a lower resolution.

2.3 Programmable Logic Control

(PLC) System

Monitoring and recording the activity of an

automated process is paramount, and thus this

subsystem will be responsible for managing the entire

work cell (command, control, and monitoring). For

this to be possible, the external device must comply

with the following requirements: ethernet, RS232,

and EtherCAT communication facilities; digital

inputs and outputs; database iteration facilities, have

OPC-UA; and be modular. As main characteristics,

the same PLC must have database interaction

capabilities and the capacity to make process-relevant

variables available through the OPC-UA protocol.

Human-Robot Collaboration (HRC) with Vision Inspection for PCB Assembly

3 LAYOUT AND WORKFLOW

To meet the requirements mentioned above, one can

propose the use of: (1) the cobot TM5-700, a 6 DOF

collaborative robot capable of using different grippers

and that integrates its own vision system; (2) the

FH1050 vision system by Omron, which has various

image acquisition and processing functionalities; and

(3) the NX102-9020 Omron, a PLC acting as an

external device capable of monitoring and controlling

all tasks and functionalities of the work cell. These

proposed systems were used and are connected, using

the TCP/IP protocol for data exchange. Figure 2

illustrates how the subsystems interact throughout the

task. In short, the PLC receives a signal from the

assembly line to start the flow, and, after that,

different signals are exchanged between the cobot,

the PLC, and the vision system to proceed with the

task. All the essential information to be seen by the

human worker is visible in an interface and then sent

to a database.

Figure 2: Data exchanged between the three subsystems.

The work cell further includes one box for the

damaged components, the PCB allocation zone, and

other zones for the boxes of the different types of

components. Figure 3 shows a sketch of the work cell,

illustrating all the components mentioned above

distribution.

Figure 3: Sketch of the work cell.

With this distribution, the cobot can reach every

position needed, independently of the velocity chosen

for the task, without too much torque effort.

In a simplified way, the general task flow of this

work cell is shown in Figure 4.

Figure 4: Flow of the work cell.

In short, the cobot will start by reading the

barcode on the PCB to know what components must

be inserted and do the referencing task to know where

the PCB is. After that, it is time for the external vision

system to read the number of pins each component

has to validate its quality. Depending on the

information received, the cobot will discard the

component or place it in the proper PCB position

upon the validation task. After placing the

component, if the PCB is not yet fully assembled, the

cobot will repeat the operation for another

component. When finished, it will wait for the arrival

of a new PCB.

4 METHODOLOGY

To execute the workflow explained above, one should

consider each subsystem individually and how to

integrate them to focus on precision and speed,

aiming to respect the following key goal: replace the

worker in the assembly of PCBs with this type of

components, trying to speed up the process and

reduce costs without ever losing quality.

To understand the entire process of this work cell,

each subsystem will be explained in a more detailed

way.

4.1 Cobot Subsystem

The cobot will have two major tasks: first, it is the

subsystem responsible for the movement in the main

task, including picking up a component, taking it to

the vision subsystem, and, finally, placing it in the

right location on the PCB. The other part is regarding

IN4PL 2022 - 3rd International Conference on Innovative Intelligent Industrial Production and Logistics

the internal vision system that this cobot has. Here,

the cobot will perform two analyses: read the PCB's

barcode to know which component to insert and

reference itself with respect to the assembly board

before inserting the first component (Figure 5). Both

tasks will be solved using the TMFlow software,

whereas, for the vision tasks, the TMvision software

is employed.

Figure 5: Workflow of the cobot subsystem.

4.1.1 Motion

Regarding this part of the task, the robot will have to

move around the work cell to reach the desired

positions, with the highest possible precision and

speed, without compromising the safety of the

environment and operator.

First, it will start in its home position, and as soon

as the PCB sensor is activated, the robot moves to the

zone where it will do the barcode reading and the

referencing task. Upon gathering the information

needed, the cobot picks the specific PCB component

and moves it to the zone where the external camera

will check the integrity of the pins. After validating

the integrity of the component, the robot proceeds to

the insertion of the same.

4.1.2 Cobot Vision System

To identify the PCB that just arrived and place the

components in the correct position, even if the PCB

comes tilted or just crooked, the cobot's internal

vision system has two types of functionalities (Figure

6) (Omron, 2020).

Figure 6: Functionalities available in the cobot’s internal

vision system: a) identify functions; b) find functions.

Every PCB comes with an associated barcode to

know precisely which type (and amount) of

component that PCB takes. To this end, the

Barcode/QR code identify function is employed.

Here, it is necessary to specify the region where the

barcode can appear, and once the camera reads a

barcode, it returns the specific number of that PCB.

Knowing the group to which that particular number

belongs, it is possible to tell which components must

be inserted (Figure 7).

Figure 7: Example of a barcode in a PCB.

For PCB referencing, the Fiducial mark find

function is employed. Fiducial marks are small

targets in an assembly board placed on the top copper

layer (and bottom if it is 2-layers) and allow the vision

system to recognize where the PCB is (Figure 8).

Human-Robot Collaboration (HRC) with Vision Inspection for PCB Assembly

Figure 8: Examples of fiducial marks in a PCB.

The fiducial marks, the target study area, and the

similarity threshold need to be selected to use this

function properly. Once the cobot reaches the

position to do the referencing, it will find the fiducial

marks and create a new coordinate frame, where the

middle point between the two fiducial marks is the

origin. From this point on, regardless of the position

of the PCB, the cobot will always be able to make the

same movement to the insertion positions, ensuring

the correct filling of the board.

4.2 FH Vision Subsystem

Here, the objective is to ensure that the component

that the cobot will insert is in good condition to be

possible to insert into the PCB without wasting it.

Usually, this is solved by human inspection, but the

goal is to replace it with a vision system to make the

work cell fully automatic and faster.

The FH-1050 vision system is used, taking

advantage of the FZ-PanDA software (Omron).

Figure 9 summarizes the tasks peformed by the vision

system.

Figure 9: a) Workflow; b) vision block.

In brief, upon waiting for the cobot with the

component to reach the point where the camera is

pointing, the system will make one single acquisition

to count the number of pins in a certain area to check

if they are damaged. The idea is to divide the

component into small sections, each with a specific

number of pins or groups/lines of pins (Figure 10).

Figure 10: Component divided into sections and number of

pins of each section.

This procedure is done using the "Shape Search

III" block, where we must select the region of interest

(ROI; big red box) and the subject we want to look

for in that exact area (small red box, for example).

Through an object detection algorithm, this function

block registers a model of an image pattern based on

its contour information, detects parts of inputted

images that most closely match the model, and adds

each similar object to the total count. An experimental

study was performed to select the suitable threshold

value for the degree of similarity between the

template and the new image. Afterwards, the

"Calculation" block is employed to find the total

number of pins that were identified in each "Shape

Search III" block. If the number equals those

expected for the component in question (44 in the

example in figure 10), the component is in good

condition, and therefore the cobot will proceed with

its insertion. If not, the cobot will discard this

component, placing it in a specific box for damaged

components.

Of note, the other two blocks present in figure 10,

the "Camera Image Input FH", are used to calibrate

the camera and enhance the image between analyses

4.3 PLC Subsystem

This subsystem aims to establish a connection

between the other two subsystems, being responsible

for the control and monitoring of the work cell and

registering the relevant data during the execution of

the task, to be later stored in a database. The kind of

Red – 2 groups

Yellow – 2 lines

Green – 2 lines

Blue – 20 pins

Purple – 14 pins

Orange – 4 pins

TOTAL: 44

IN4PL 2022 - 3rd International Conference on Innovative Intelligent Industrial Production and Logistics

data to keep track of includes, for example, the

number of damaged components, the number of

PCBs filled, the time it takes to fill each board type,

and so on. Figure 4 summarizes the information

exchanged between subsystems, and as shown, the

whole process of this task is divided into three main

parts.

First, the cobot waits for the PLC to tell it that a

new PCB has arrived. Then, as mentioned before, the

cobot will use its internal vision system to do the

referencing task and to read the barcode, sending the

reading signal to the control system. Here, the PLC

will then decide what type of PCB it is and then tells

the cobot what type of components to insert.

Upon picking up the component, the cobot will

take it to the external vision system area. Here, the

PLC will send two commands: (1) first the command

measurement, where the system will run the code to

count the number of pins to assess the component’s

condition; and (2) the get data command, where the

PLC will receive from the external vision system the

number of pins that the software detected. Then, the

PLC compares this value with the expected value, and

if the result is positive, it informs the cobot that it can

proceed with the insertion. If it is not, the PLC

informs the cobot to discard this component and pick

a new one.

Finally, the cobot will proceed with the insertion

of the good component, repeating these steps until the

PCB is filled. Meanwhile, the PLC will store relevant

data during these steps, such as the number of PCBs

filled, the number of damaged components, and the

time to fill one PCB, among others.

Table 1 explains how the information will flow

between the three subsystems.

Table 1: Information exchange.

PLC takes PLC sends

PCB sensor - PCB

arrived at the assembly

area

Signal/Message for

cobot to start

PCB barcode from

the cobot reading.

Type of PCB for

cobot to know which

type of components it

takes

External vision

system reached

Measure command

for vision system

------------------ Get data command

for vision system

------------------ OK/NG component

for cobot

PCB assemble

------------------

5 TESTS

This section demonstrates some practical examples of

this work cell, showing the correct implementation of

the main systems and the compliance with the

requirements. The tests were:

• Test A: How much time does it takes for the

cobot to fill a PCB;

• Test B: Reliability of the external vision system;

• Test C: Monitoring and data acquisition.

These tests were employed in the following

environment, where it is possible to see: (A) the

external vision system for inspection; (B) the cobot

with the internal vision system for barcode reading

and referencing; (C) the box of components; (D) the

PCB (Figure 11).

Figure 11: Environment of the work cell.

5.1 How Much Time Does It Take for

the Cobot to Fill a PCB

In this test, the idea was to run the full task of this

work cell and see how long it takes to fill a PCB,

repeating this trial for different velocities (Figure 12).

Figure 12: Workflow of the main task for PCB type 4.

Human-Robot Collaboration (HRC) with Vision Inspection for PCB Assembly

The max velocity was set to 1.5 m/s, and the trials

were done for 20%, 40%, 60%, 80%, and 100% of

that speed. Of note, it was possible to see that the

cobot never reaches the max speed because the

movements that it makes during this task do not have

the range needed to reach that velocity. Table 2 shows

the time recorded for each speed setting.

Table 2: Time to fill a PCB with different speeds.

Velocity Time (m/s)

20% 0:59,260

40% 0:38,672

60% 0:31,313

80% 0:29,241

100% 0:28,039

It is possible to conclude that the cobot needs less

than 30 seconds to fill a PCB of type 4 (one that takes

three components).

So far, one has shown that the robot can indeed

meet the mentioned requirements, both in terms of

speed and efficiency, but we have not observed its

added value over the human operator. Thus, some

additional tests were performed, where an operator

tried to insert the components manually without

validation of any kind. Despite initially taking around

20 seconds to fill a PCB, this time was reduced to an

average of 15 seconds after a few attempts. The same

type of test was executed using the cobot, observing

a time of 16.280 seconds. Overall, it is possible to

conclude that the cobot, besides being able to

maintain this pace continuously, does not present

significant delays over an operator, as it takes only 12

seconds more to fill the PCB while also inspecting it

as well as the components.

5.2 Reliability of the External Vision

System

This last test aims to confirm if the external vision

system has the capacity and reliability to validate

different types of damaged components. For that,

different components were structurally compromised

in different areas, and the task was run ten times to

inspect how many true positives the vision system

could detect. Additionally, the same test was run for

one good component to validate if the system could

validate its integrity either (Figure 13).

During the ten trials of each test, the external

vision system showed its reliability when validating

the components, verifying that it is in bad condition

in all of the examples mentioned above, except for the

case of a "good” component. Moreover, it also has the

capability of showing in which area the damage is,

presenting a result of 10 out of 10 successful

validations. From the moment the cobot reaches the

inspection zone, and the moment the main system

(PLC) receives the answer, the camera only needs

around 0,8-1s to process all this.

Figure 13: Examples of the components used.

5.3 Monitoring and Data Acquisition

The goal here is to test whether the PLC can retrieve

the necessary information during the execution of the

task mentioned above to be later able to visualize and

document this data. The number of components and

the number of PCBs filled with each type are

examples of data that needed to be retrieved to be

easier for the co-worker to know, for example, when

to change the empty box of components, to know the

percentage of damaged components or the number of

PCBs of type 3 filled in one day.

To check the retrieved information, an interface

was built using the CX-Designer software (Figure

14).

Figure 14: CX-Designer interface.

Figure 14 illustrates the five main parts of the

interface:

IN4PL 2022 - 3rd International Conference on Innovative Intelligent Industrial Production and Logistics

• Section A notifies about the steps being

performed during the routine;

• Section B informs the type of PCB detected

during the barcode read function and the number

of components it takes;

• Section C presents the number of components

inserted, the number of damaged components,

and the total number of picked components;

• Section D informs the number of components

left in the box, showing a yellow LED when it

is almost empty, and a red LED when it is

empty;

• Section E lets know the number of pins read,

showing the result of the external vision system.

Figure 15 shows a possible sequencing of what

can be observed in the interface over several

iterations.

Two images were recorded during the filling of

the first PCB (Figure 15-1 and Figure 15-2). The first

one shows the pick and place task and the validation

task of the first component, where it is possible to see

the number of pins read by the external vision system

validating the component and resulting in its

insertion. The second shows the expected last

component of that PCB, where it was a damaged

component, showing 42 read pins instead of 44, and

a total of one damaged component out of a total of

three. The last figure (Figure 15-3) illustrates an

example of a possible iteration of the interface during

a day of work.

6 DISCUSSION

During Test A, we confirmed the accomplishment of

the 30 second requirement. We performed some

insertion trials to compare this with a human worker,

concluding that the worker can be faster by 1.280

seconds on average. However, some major

requirements need to be considered, like efficiency,

repeatability, and precision. Humans are susceptible

to tiredness, and after hours of work, their precision

and efficiency will not be the same, leading to delays

in production or, in worst cases, damaged PCBs. In

opposition, the cobot always performs at his highest

level, sustaining these requirements throughout hours

of work. These requirements are also possible with

industrial robots, but they have some disadvantages,

such as: they are usually more expensive, they are

used for heavier payloads, and they also require an

isolated work area.

Moreover, different camera positions were tested

during this test to find the easiest and fastest way to

achieve a more efficient work. When the camera is on

top, the environment lightning variations do not affect

the external vision system precision and fast decision

making. There were still two other possibilities: one

being the cobot’s camera performing the inspection,

however this camera did not present the minimum

requirements for this process; and the other was an

external camera attached to the cobot, which met the

quality and resolution requirements, but increased the

cycle time of the entire process since the robot would

need to place the component in an intermediate

position to take the reading.

Test B confirms the high level of inspection by the

external vision system when validating minor

displacements and defects of the components’ pins. It

thus delivers a sensibility and accuracy that a human

eye cannot achieve.

One of the few tasks in this work cell that a human

worker needs is replacing an empty component box.

In Test C, one concluded that this type of information

cannot be presented as a mere label but also needs to

be displayed using three LEDs, representing the

quantity of components left in the box, giving an

explicit luminous warning to the human worker. In

addition to this task, to reduce the cycle time of this

process, the operator cooperates with the cobot by

informing it of the limited barcode search area since

each type of PCB has its barcode in different areas.

Test C confirms the high level of inspection by the

external vision system when validating minor

displacements and defects of the components’ pins. It

thus delivers a sensibility and accuracy that a human

eye cannot achieve.

As mentioned before, the human worker could

replace some functionalities of this work cell, like the

Human-Robot Collaboration (HRC) with Vision Inspection for PCB Assembly

Figure 15: Interface during sequenced iterations.

barcode read and the referencing functions. Here,

the worker could be responsible for informing the

cobot of what type of components should be inserted.

However, the idea is to minimize the human

interaction, and through the barcode read function is

possible to pass this information in an autonomous

way without wasting much time. Also, when using

the referencing function, we guarantee the correct

placement of each component, as even when using a

mechanical interlock, there is no guarantee of the

correct PCB orientation, which over time can cause

displacements in the order of millimeters.

In addition to the solution proposed throughout

the paper, other two approaches were considered. The

first one considered the collaboration between human

and robot in a way that the robot would perform the

pick task and take it to the vision system, and after the

component validation, it would be inserted by the

human, having the transference of it between the

cobot and the operator. On the other hand, in the

second hypothesis, the operator would perform the

validation task of the component, and after it was

validated, it would be transferred between the

operator and the cobot, and the cobot would perform

the insertion part. Both solutions have disadvantages

since they increase the cycle time of the process, with

the transition of the component between the operator

and the cobot, as well as relying on less accurate and

precise systems.

7 CONCLUSION

In this work, was presented a solution to a prototype

autonomous work cell capable of assembling PCBs

with PTH components. This work cell consisted of

three main systems: (1) the cobot; (2) the vision

system; and (3) the PLC, exchanging information

between them via TCP/IP protocol.

The first one is mainly responsible for the

movement in the main task, including picking up the

components, taking them to the inspection area, and

inserting them into the PCB. The vision subsystem

ensures the precise positioning and identification of

the PCB and verifies the components’ condition. The

last subsystem manages both the cobot and external

vision system’s interactions and is responsible for

monitoring the work cell by registering the relevant

data generated during the execution of the task and

storing it in a database.

Hereto, the three subsystems were demonstrated

and explained, always showing the key aspects of the

creation of this work cell. The performance of the

proposed work cell was demonstrated through three

examples, focusing on the most important aspects,

namely exchange and acquisition of information, the

validation of components, and the gain in precision

and speed during PCB assembly over a human

operator, all of them showing positive and successful

results.

In sum, this work proves the optimization of a

human process. It proves the possible replacement of

the main human tasks with systems that are capable

of maintaining a high level of performance

throughout long periods of time.

ACKNOWLEDGEMENTS

This work is co-funded by the European Regional

Development Fund (ERDF) through the Operational

Competitiveness and Internationalization Programme

(COMPETE 2020) of the Portugal 2020 Program

[Project No. 45070, "FlexASComp”; Funding

Reference: POCI-01-0247-FEDER-045070].

REFERENCES

Altinkemer, K., Kazaz, B., Köksalan, M., & Moskowitz, H.

(2000). Optimization of printed circuit board

IN4PL 2022 - 3rd International Conference on Innovative Intelligent Industrial Production and Logistics

manufacturing: Integrated modeling and algorithms.

European journal of operational research, 124(2), 409-

421.

Andrzejewski, K., Cooper, M., Griffiths, C., & Giannetti,

C. (2018). Optimisation process for robotic assembly of

electronic components. The International Journal of

Advanced Manufacturing Technology, 99(9), 2523-

2535.

Bogner, K., Pferschy, U., Unterberger, R., & Zeiner, H.

(2018). Optimised scheduling in human–robot

collaboration–a use case in the assembly of printed

circuit boards. International Journal of Production

Research, 56(16), 5522-5540.

Crama, Y., Flippo, O. E., Van De Klundert, J., & Spieksma,

F. C. (1997). The assembly of printed circuit boards: A

case with multiple machines and multiple board types.

European journal of operational research, 98(3), 457-

472.

David A. Forsyth, J. P. (2002). Computer Vision - A

Modern Aproach.

Iftikhar, B., Malik, M., Hadi, S., Wajid, O., Farooq, M.,

Rehman, M., & Hassan, A. (2020). Cost-effective,

Reliable, and Precise Surface Mount Device (SMD) on

PCBs. Paper presented at the IOP Conference Series:

Materials Science and Engineering.

Kurka, P. R. G., & Salazar, A. A. D. (2019). Applications

of image processing in robotics and instrumentation.

Mechanical Systems and Signal Processing, 124, 142-

169.

Langmann, R., & Rojas-Peña, L. F. (2016). A PLC as an

Industry 4.0 component. Paper presented at the 2016

13th International Conference on Remote Engineering

and Virtual Instrumentation (REV).

Omron. FH/FHV/FZ5 Series Vision System - Processing

Item Function Reference Manual. Retrieved from

https://www.edata.omron.com.au/eData/manuals.html

Omron. (2020). Software Manual TMvision. Retrieved

from

https://www.edata.omron.com.au/eData/manuals.html

Poór, P., Broum, T., & Basl, J. (2019). Role of collaborative

robots in industry 4.0 with target on education in

industrial engineering. Paper presented at the 2019 4th

International Conference on Control, Robotics and

Cybernetics (CRC).

Robotics, I. F. o. (2018). Demystifying Collaborative

Industrial Robots. Frankfurt, Germany.

Vojić, S. (2020). Applications of collaborative industrial

robots. Machines. Technologies. Materials., 14(3), 96-

99.

Wendy Jane Preston. (2018). The difference between THT

and SMT. Retrieved from https://www.harwin.com/

blog/the-difference-between-tht-and-smt/

Human-Robot Collaboration (HRC) with Vision Inspection for PCB Assembly