Portable Safety System using Radar for Flexible

Human-Robot-Collaboration in a Real Semi-automated

Production Line

Christian Bergner, Ferhat Akan, Ronald Schmidt-Vollus

, Peter Heß

and Christian Deuerlein

Institute of Smart Production, Nuremberg Campus of Technology, Fuerther Strasse 246b, Nuremberg, Germany

Keywords: Human-Robot-Collaboration, Safety System, Radar, Flexibility, Portability, Standard-CE.

Abstract: The implementation of a reliable vision system for a human-robot environment is a key issue for the

collaborative production industry. The core challenge of human-robot collaboration is to ensure safety.

Furthermore, a flexible safety system is required for frequently changing applications and work areas. This

paper focuses on the development and application of a workspace monitoring system for safeguarding using

radar sensors. The human-robot collaboration cell is designed to enable a flexible integration regardless of

the work location. This results in higher productivity. Since no separating protective devices are provided for

the cell, safety-oriented monitoring and control by suitable safety sensors is required. The methods to

minimize the size of the necessary safety distance will be presented. The experimental validation shows that

this safety system with radar sensors performs a reliable workspace monitoring system. The high robustness,

reactivity and flexibility of the safety concept makes this system usable for collaborative tasks in a real

industrial environment.

1 INTRODUCTION

The assembly and installation of mechatronic

products in small and midsize companies is mostly

indicated with a high number of variants, which leads

to a low number of order quantities. Consequently,

there are high requirements on the flexibility in a

production line and the assembly, which presupposes

a high number of employees. In contrast, the

automation level in the final assembly of the products

is relatively low. Because of the continuously

increasing shortage of skilled labor, the degree of

automation needs to be increased. The use of

collaborative robot systems that can be flexibly

integrated into an existing production line has the

potential to solve this problem. Most of the work

steps combine filigree work that can be done by the

human and monotonous work that can be done by the

robot. This requires a collaboration, which leads to an

overlapping work area of both parties. Therefore, a

flexible safety system for human-robot collaboration

is indispensable that can be integrated to different

https://orcid.org/0000-0001-8107-2820

workplaces without individual safety considerations.

In our research, we present a safety system that is

completely mounted on a mobile robotic cell so that

no modifications of workplaces is necessary. In

addition, the configuration of the safety system, when

it is moved to a different workplace, is done

automatically. Above all, the system conforms to all

the current international safety standards. In this

paper, the transfer from scientific findings into a real

industrial environment will be outlined.

2 RELATED WORK

With the introduction of collaborative robots in

industry, the field of robot safety has been redefined.

Certain conditions are required for the collaborative

approach. Thus, safety standards such as DIN EN

ISO 10218 part one (DIN Deutsches Institut für

Normung, 2011) and two (DIN Deutsches Institut für

Normung, 2011) have been introduced, which

identify specific applications and criteria. The safety

Bergner, C., Akan, F., Schmidt-Vollus, R., Heß, P. and Deuerlein, C.

Portable Safety System using Radar for Flexible Human-Robot-Collaboration in a Real Semi-automated Production Line.

DOI: 10.5220/0010645800003061

In Proceedings of the 2nd International Conference on Robotics, Computer Vision and Intelligent Systems (ROBOVIS 2021), pages 67-76

ISBN: 978-989-758-537-1

requirements for collaborative robot systems and the

working environment have been extended by the

technical specification ISO/TS 15066 (DIN

Deutsches Institut für Normung, 2016). This

complements the requirements and guidelines for

collaborative robot applications. It is possible for the

robot to move even if the human is working in the

same workspace. For this collaboration, the safety

system is the predominant aspect for the successful

implementation in a real industrial environment. The

state of the art presents many possible solutions.

2.1 Safety Concepts for

Human-Robot-Collaboration

Lasota et al., define four main methods to provide

safety for a human-robot system: motion planning,

prediction, control and consideration. According to

motion planning, the safety system can be subdivided

into collision avoidance and collision recognition

(Lasota, Fong, & Shah, 2020).

The first one is presented by Vogel et al. in their

research to implement a projection- and camera-

based safety system. Depending on the position and

the velocity of the robot, a well-shaped and

dynamically adapted safety space is projected on the

table. If an object disrupts the emitted light rays of the

projector, the robot stops its movement to avoid any

collision with the human (Vogel, Walter, & Elkmann,

2013; Vogel, Walter, & Elkmann, 2017).

On the contrary, Kulic and Croft present a safety

system that is dodging obstacles instead of inducing

an emergency stop. The distance is determined by a

stereo-camera at the bottom of the robot to catch the

human and the trajectory of the robot. Thus, the

system can predict a potential collision and avoid it

(Kulic & Croft, 2005).

In their research Berg et al. present an approach

to integrate safety elements into a task-oriented

programming system to increase the flexibility for

human-robot collaboration. Safety aspects are

considered by a planning, programming and

operation module as well as a safety-check before

operation (Berg, Richter, & Reinhart, 2018).

Antonelli et al., introduce a safety system for a

flexible and safe interactive human-robot

environment in small batch production. The idea is to

integrate a so-called Superior Hierarchical

Controller that is used as interface between the

human and the robot. The controller gathers

information from safety sensors, e.g. laser scanner at

the bottom, as well as from smart cameras that are

located over the working area of the robot (Antonelli,

Astanin, Caporaletti, & Donati, 2014).

A radar-based safety system for estimation of the

distance between the robot and human is presented by

Zlatanski et al. The researchers compared static and

dynamic characteristics of the radar sensor with a

state-of-the-art laser scanner. The experimental set-

ups show that both sensor types are performing

comparable to each other in respect of the field of

view, resolution and reaction time (Zlatanski,

Sommer, Zurfluh, & Madonna, 2018).

Amin et al. are presenting a mixed-perception

approach for safe HRC in industrial automation using

deep learning networks and AI for action recognition

and contact detection. The action is monitored using

a skeleton model of the human inside the workspace.

The physical contact is distinguished between

intentional and accidental interaction. The results

show a high potential for AI-driven solutions for the

safety in HRC (Amin, Rezayati, Venn, & Karimpour,

2020).

A new collaborative robot skin (CoboSkin) for

HRC is presented and investigated by Pang et al. The

skin consists of inflatable and sensing units. The latter

ones are able to measure the force in real-time. By

adjusting the internal air pressure, the stiffness of the

skin can be varied. The results show that the impact

force during a collision of human and robot can be

reduced by adapting the air pressure (Pang et al.,

2021).

Other related safety concepts in the field of HRC

are investigated in(Salmi et al., 2013; Dohi et al.,

2018; Halme et al., 2018; Hoskins, Padayachee, &

Bright, 2019; Matthias et al., 2011).

2.2 Sensor Systems for

Human-Robot-Collaboration in

Real Industrial Environments

In most of the real industrial applications, the safety

system for human-robot-collaboration is realized by

the reduction of speed and force in order to fulfill the

requirements given by the ISO/TS 15066. (KUKA

Systems GmbH, 2018) (Glastechnik Hofmann

GmbH, 2017)

Furthermore, Rexroth developed the so-called

APAS assistant mobile (Rexroth, 2014), which is a

mobile collaborative robotic system that can be

flexibly used at different workplaces. The safety

system consists of a capacitive sensor skin that

detects the presence of a human before a collision

occurs. In this case, the robot is switched to a safety

stop. When no worker is nearby the robot, it is

moving with a reduced speed.

The SafetyEYE is one of the first safe camera

systems for 3D room monitoring (PILZ, 2014). It

ROBOVIS 2021 - 2nd International Conference on Robotics, Computer Vision and Intelligent Systems

offers new possibilities for monitoring and

safeguarding danger zones. The sensor system detects

and reports the intrusion of objects into warning and

detection zones, which can be freely defined. For the

flexible installation, at least four markers have to be

placed on the floor of the supervised area.

The current state of the art presents many possible

solutions that are listed and compared to the proposed

safety system in this paper. Table 1 shows the result:

Table 1: Comparing the proposed system with related

systems.

The comparison makes clear that there is currently

no safety system available on the market that has a

high flexibility according to different workplaces,

allows maximum robot speed, has a CE-Mark, do not

lead to a safety stop, when it is triggered, needs no

individual configuration on new workplaces, is

mobile and needs no modification of the existing

workplaces. Only the portable safety system in this

paper fulfills all those requirements that are

indisputable for the use in a real industrial

environment.

For the methods of (Vogel, Walter, & Elkmann,

2013; Kulic & Croft, 2005; Amin, Rezayati, Venn, &

Karimpour, 2020) the robot cell has to be adapted in

order to integrate their systems. The safety system of

(Antonelli, Astanin, Caporaletti, & Donati, 2014) is

limited to specific workplaces and can not be flexibly

used. The approach of (Amin, Rezayati, Venn, &

Karimpour, 2020) is not conforming to safety

standards. Thus, it can not be integrated into a real

production line. The sensor systems of (KUKA

Systems GmbH, 2018; Glastechnik Hofmann GmbH,

2017; Rexroth, 2014) lead to high cycle times and low

productivity, because the robot is continuously

moving with reduced speed. The system in (PILZ,

2014) has the disadvantage that it has a huge

supervised area. It also has to be configured at every

different workplace. Therefore, more research needs

to be done in order to close this gap.

3 MOBILE ROBOTIC CELL

During the project AdhocMRK we wanted to define

and develop a safety system for human-robot-

collaboration that can be flexibly moved and

integrated into a real semi-automated production line.

The main research question is, how a sensor system

for a movable and portable robotic application that

also confirms to the international safety standards

could be designed.

To put this system into operation in a real

industrial environment, there are many requirements

that have to be fulfilled. The safety system must not

transfer the robot to a safety stop, when the sensors

are triggered to increase the productivity. The

existing workplaces must not be modified or

remodeled. The entire sensor technology has to be

mounted on the mobile robotic cell. The safety system

has to be maximum adaptable to new workplaces so

that no individual reconfiguration of the sensors is

necessary. Finally, no individual safety assessment is

supposed to be performed on a new workplace.

Current safety concepts are not able to fulfill all

the requirements that are provided to the robotic

system. The main reasons for this is that the robotic

cell has to be movable and deployable on different

applications. The existing safety concepts are not

portable and thus limited to an individual workplace.

To cope with these challenges, we defined a

sensor system for a mobile robotic cell that supervises

the space in front and to the both sides of the robotic

application including the considered workplace. To

the front a safe laser scanner is used that is configured

for hand detection. To the both sides, safe radar

scanner are detecting the presence of a human. These

sensors are configured with person detection. The

sensors do not supervise the access from behind the

workplace. Thus, e.g. a safety fence has to prevent a

human from entering the robotic system. This so-

called external safety system is completely mounted

on the mobile robotic cell and detects the presence of

a human nearby the working area of the robot. We

also defined an internal safety system that is

supervising the movement of the robot by the usage

of safety planes.

According to the safety system, the robot can run

in normal or reduced mode. In normal mode, the robot

can move with maximum speed and force. The

reduced mode can be initiated either by the external

or by the internal safety system. In this mode the

robot´s velocity and force is strictly limited, but not

stopped. This leads to a reduction of cycle time,

because the robot is still moving. Only when the robot

collides with the human, a protective stop is

initialized and the movement stops.

For the realization of the safety concept, we

constructed and built a robotic cell that consists of the

following elements (Figure 1). First, the collaborative

Safety S top

When Triggered

Modi fy

Workplaces

Mobile

System

Indi vidual

Configuration

Flexibility

Allows Maximum

Robot S peed

Mar k

Vo gel e t al.

no yes no no low yes no

Kulic & Croft

no yes no no high no no

Anto nelli et al.

no yes no yes low yes no

Amin et al.

no no no yes low no no

KUKA Sys tems

yes yes no yes very low no yes

Glastechnik Hofmann

yes yes no yes very low yes yes

Rexrot h

yes no yes no high no yes

PILZ

yes yes no yes very low yes yes

Portable Safety System

no no yes no very high yes yes

Portable Safety System using Radar for Flexible Human-Robot-Collaboration in a Real Semi-automated Production Line

robot (UR5e) including the teach-panel and the robot

controller, which are mounted on a mobile platform.

This platform can be moved to different workplaces

manually via guide rolls. Second, a framework was

constructed, on which the external sensor system, a

signal tower as well as some pushbuttons and

switches are installed. Third, a control cabinet, which

involves the safe programmable logic controller

(PLC), safe digital I/O modules, the power-supply

unit for 24 VDC and the controller of the radar

sensors.

Figure 1: Flexible and mobile robotic cell.

For the communication between the robot cell and

an industrial workplace in a real scenario, a specific

and standardized plug system is used. This plug

contains of the power supply, compressed-air supply

as well as digital I/Os for the controlling of the motion

sequence of the robot. A toggle fastener realizes the

firm connection towards the workplace. The plug and

the toggle fastener make the mobile robotic cell

applicable to different workstations in a real industrial

environment.

4 EXTERNAL SENSOR SYSTEM

The sensor system for the mobile robotic cell consists

of one safe Lidar Laser scanner from SICK, the

S3000 Standard and four safe radar scanner from

Inxpect, the LBK system. A PLC is used to unite the

sensor data. The entire sensor system and all the

described components are configured with two

channels to conform the international standards. For

each of the sensors specific safety distances, which

are defined as the distance from the beginning of the

supervised area by the sensor to the working area of

the robot, have to be calculated.

4.1 Safe Lidar Laser Scanner with

Hand Detection

To supervise the access of a human to the front of the

robotic cell, the safe laser scanner is used (Figure 2).

The sensor is working with the method light detection

and ranging (Lidar).

Figure 2: Laser scanner to the front.

By measuring the time between sending and receiving

of laser impulses, the distance can be calculated or

respectively the presence of a human. According to

the international standard DIN ISO13855 (DIN

Deutsches Institut für Normung, 2010) the minimum

safety distance S

min, Laser

to the front side, can be

calculated by (1). The sensor is configured for hand

detection which means that the sensor detection level

d = 40 mm. This value represents the distance

between two laser beams that are emitted by the

sensor. The stopping time T is composed of the

stopping time of the PLC, the laser scanner and the

UR5e Cobot. The approach speed K is a constant

value and is set to 1.600 mm/s according to the

standard, which results with (1) in a minimum safety

distance of S

min,Laser

= 640 mm.

𝑆

,

𝐾∙𝑇8∙



𝑑14



(1)

The supervised area by the laser scanner is configured

rectangular. Figure 6 shows the supervised area (A).

4.2 Safe Radar Scanner with Person

Detection

The safe radar scanner are used to detect the access of

a human to both sides of the robotic system (Figure

3).

Figure 3: Radar sensor for the detection to the sides.

ROBOVIS 2021 - 2nd International Conference on Robotics, Computer Vision and Intelligent Systems

The LBK system is based on a 24 GHz radar

algorithm that filters out disturbances, e.g. smoke,

dust, splashes or machining waste. This leads to a

reduction of false alarms and thus increases the

productivity. The sensor transmits the radio waves

and identifies motion information by analyzing the

returned signals reflected from both static and moving

objects in the operating area. The sensor only detects

the movement of objects, not the presence of an

object itself, which is the biggest unique feature

compared to a laser scanner. It is also automatically

reconfigured on a new workplace with different

environment, which makes the entire external safety

system portable. The portability is given, when the

robotic cell is added to a new workplace, where the

surrounding always changes. The supervised area of

the radar sensor can be adjusted with two variants

concerning the two axes horizontal and vertical:

• wide protective area: 110° horizontal, 30° vertical

• narrow protective area: 50° horizontal, 15°

vertical

According to DIN 13855 (DIN Deutsches Institut für

Normung, 2010) the minimum safety distance S

min,

Radar

to both sides of the robotic cell can be calculated

by (2):

𝑆

,

𝐾∙𝑇𝐶

(2)

These sensors are configured for person detection,

which implies a sensor detection level d = 70 mm.

The stopping time T is composed of the PLC, the

radar scanner and the UR5e Cobot. With the equal

approach speed K as for the laser scanner and

equation (2) the minimum safety distance S

min, Radar

1.242 mm is determined. Figure 6 shows the

supervised are by the radar scanner to both sides (B).

For the automatic restart of the robot with maximum

velocity in normal mode, two more radar sensors are

used to supervise the area that is not covered by the

laser scanner and the radar sensors to both sides. In

case a human worker enters the working area of the

robot the safety system is triggered and the robot is

set to reduced mode. Two additional radar sensors

make sure that no human is inside the robotic cell.

After a timeout of 10 s, the robot is set back to normal

mode. Figure 4 shows the supervised areas for the

automatic restart. To conform the standards, the radar

sensors for the automatic restart have to be mounted

in a specific height over the ground. According to

DIN 13855 (DIN Deutsches Institut für Normung,

2010) the minimum height of the safety field H

min

can

be calculated by (3):

𝐻



15∙



𝑑14



(3)

With a sensor detection level of d = 70 mm and (3),

the minimum height H

min

is calculated with 300 mm.

Figure 4: Supervised area for automatic restart.

4.3 Programmable Logic Control for

Unification

The safe PLC is used for the communication between

the robot controller and the external sensor system.

Figure 5 shows the configuration of the PLC. If one

of the two sensor types (radar or laser) are triggered,

because a worker is entering the supervised area, the

two-channeled digital outputs of the PLC to the robot

controller are switched to FALSE. This transfers the

robot into the reduced mode.

Figure 5: Configuration of the external sensor system by the

PLC.

The robot controller is also communicating with

the PLC, when the robot is either running in the

reduced mode or stands still, because of a safety stop.

Portable Safety System using Radar for Flexible Human-Robot-Collaboration in a Real Semi-automated Production Line

Generally, the reduced mode can be triggered by the

external safety system as well as by the internal safety

system through configured protective levels on the

robot. The safety status of the robot is signalized by

the signal tower, so that the worker gets feedback

even if they are not right next to the robot cell. The

safety switch can be used to switch off the external

safety system so that the robot is continuously

running in the reduced mode.

4.4 Entire External Safety System

As shown, the safety distances need to have a

minimum size to detect a human reliably to the front

and to both sides. Figure 6 shows the entire safety

space of the external sensors.

Figure 6: Supervised area of the entire external safety

system.

Behind the workplace a safety fence prevents a

worker from entering the robotic cell. In our research,

we distinguish between two different constitutions of

the space, where the robotic cell is supposed to be set

up. First, the inappropriate space, when there is not

enough space nearby the real industrial workplace for

the supervised area by the external safety system. In

this case, the safety distances can not be maintained

and the sensors are continuously triggered by humans

working next to the robot. Subsequently, the mobile

robotic cell has to run without the supervision by the

external sensors and is set to the reduced mode by the

safety switch. Second, we considered the so called

sufficient space. In this case, there is enough space for

the supervised area of the external sensors, which

means that workers nearby the robot do not

continuously trigger the sensors. Thus, the robot can

run with maximum speed and force and is only

switched to reduced mode when a human is entering

the robotic cell, for example during a change of the

box for the supply and removal of components.

The external safety system is not limited to one

Cobot size. When a bigger robot is used there are not

more sensors needed. When the reactivity of the

bigger robot is different to the current Cobot, either

the angle of the sensors mounted on the mechanical

structure can be adapted or the safety planes can be

moved to hold the safety distance according to the

standard. There is no need to add more sensors.

5 INTERNAL SAFETY SYSTEM:

SAFETY PLANES AT THE

ROBOT

The safety system also uses the internal sensors of the

robot to transfer the robot into the reduced mode and

to limit the movement area of the robot. In our

research, we configured and tested so-called safety

planes at the robot that are presented in the next sub-

chapters.

5.1 Initiate the Reduced Mode

The reduced mode is initiated, when the robot is

crossing predefined safety planes to the front and to

both sides, as it is illustrated in Figure 7.

Figure 7: Safety planes to initiate the reduced mode.

When the robot moves back inside the curtailed area,

it is set back to the normal mode and moves with

maximum speed. Those planes are necessary to

minimize the supervised area by the external safety

system. The angle of the planes to the side is the same

as the angle of the radar sensors that are supervising

the space next to the robotic cell. Thereby, the safety

distance S

min,Radar

can be maintained.

ROBOVIS 2021 - 2nd International Conference on Robotics, Computer Vision and Intelligent Systems

5.2 Limiting the Movement Area of the

Robot

In order to integrate the flexible safety system in a real

semi-automated production line, the robot´s

movement nearby the head and face of a human has

to be limited. According to the DIN EN ISO 14738

(DIN Deutsches Institut für Normung, 2009) the

standard height of the shoulders including the heel of

the safety shoes is 134 cm. Therefore, the safety plane

at the robot is configured parallel to the ground at this

height. It is not initiating the reduced mode but is

limiting the movement area of the robot to the top.

Respectively, the Tool-Center-Point (TCP) is not able

to cross it.

6 STANDARDIZED CE-MARK

To fulfill the requirement that no individual safety

assessment has to be performed on a new workplace,

we defined criteria to classify a specific component

series. The biggest (95x145x50mm) and heaviest

(536g) part of the series shows Figure 8.

Figure 8: Component from DEHN SE + Co KG.

The rest of the series consists only of three, two or

one chamber. By the definition of the criteria, also

other parts that looks completely different can be

handled without the performance of an individual

safety assessment. The only precondition is that the

considered application is not accessible from behind

the workplace and the new components correspond to

the specification in Table 2.

Table 2: Classification of a component series for a standard

CE-Label.

Criteria Limit

Basic form Cuboi

Sharp forms that are emerging out of

the geometry

Not existing

Size of chambers Max. 32x45mm

Mass Max. 536g

Length Max. 250

Width Max. 250

Heigh

Max. 250

Corner radius Min. 0,5mm

Edge radius Min. 0,5mm

Sharp edges/corners Not existing

Surface condition Rz < 1mm

Next to the definition of the criteria, we also

constructed and 3D-printed a safe vacuum gripper

that is able to handle the components of the series. It

consists of two suction devices that can be flexible

removed and added to the housing of the gripper.

Finally, the gripper must not be changed at a new

workplace with new parts, so that the safety

assessment and the assignment of a CE-mark is made

much easier.

To verify our safety system, we also performed an

extensive assessment of risk with the biggest and

heaviest part by using the software SafExpert. The

application we considered was from a project partner

of AdhocMRK. In this case, presorted parts from a

box are picked by the robot and inserted into an

automatic test machine. After the successful high

potential test, the parts are removed and sorted into

another box next to the robotic cell.

Most of the risks could be eliminated by an

inherent safe construction. For the rest of the risks we

performed a force and pressure measurement

according to ISO/TS 15066. By evaluating 40

measurements, we were able to define the safe speed

of the robot by 200 mm/s. That is the speed for the

robotic cell, when it is running in the reduced mode.

7 VALIDATION OF THE SAFETY

SYSTEM

In our research, we also performed the validation of

the flexible safety concept with the radar sensors to

check and confirm the functionality of the system as

following:

• Verification of speed regulation and sensor

activity

• Cycle time measurements for the safety planes

• Adaptability for different workplaces

First, we evaluated and confirmed the functionality of

the implemented safety concept with the described

safety functions. Therefore, we first verified the

positions and range of the supervised areas by the

external and internal sensors. We were both entering

the robotic cell from several sides and moving the

TCP of the robot from inside in direction to the

supervised areas and the safety planes. We checked

the signal for the reduced mode with the signal tower

and the safe digital outputs for the reduced mode on

the robot controller for several positions.

Portable Safety System using Radar for Flexible Human-Robot-Collaboration in a Real Semi-automated Production Line

Second, in order to analyze the influence of the

variation of the safety planes position on the cycle

times, a prototypical application was examined.

When the robot is crossing the planes, it is transferred

to the reduced mode. Figure 9 shows the experimental

setup for the time measurements:

Figure 9: Experimental setup for the variation of safety

planes position and time measurements.

The robot is moving cyclic from position (0) of

the left box to position (0) on the table, which

characterizes the workplace, to position (0) of the

right box. Thus, the path would be (0, 0, 0). In the

next cycle, the robot starts again from (0) of the left

box, to (0) of the table and to (1) of the right box. So

the entire sequence would be from (0, 0, 0) to (0, 0,

4), continued with (0, 1, 0) to (0, 1, 4) and so on. The

last cycle is from (4, 2, 3) to (4, 2, 4). Finally, the

robot was driving to every position from the left box,

combined with every position on the table and every

position from the right box during one sequence. All

the positions were sent to the robot via a TCP/IP

socket connection between a computer and the robot

controller. The time was measured for each cycle by

a C#-program and saved to a .csv-file for evaluation.

Overall, we recorded three sequences, which contain

of 75 cycles each. Every sequence is representing one

configurations of the safety planes. In summary, we

evaluated 225 cycles. The longest cycle times

expected to be from point (1) of the left box to all the

other positions of the table and the right box, because

it is the longest path of the robot. Thus, this point is

considered for the evaluation. According to that, the

robot performs five cycles crossing each of the three

points on the table (0, 1 and 2), which results in 15

measurements. In normal mode, the robot drives with

a speed of 400 mm/s, which is performed mostly

nearby the table. When the robot is crossing the

planes, it is only moving with 200 mm/s. Figure 10

shows the result of our cycle time measurement:

Figure 10: Evaluation of the cycle time measurements.

Next to the transition from normal to reduced

mode by the different safety plane configurations, we

were also testing the robot continuously running in

the normal (fast) and the reduced mode (slow). Our

experiment shows that the position of the safety

planes has no significant influence on the cycle times

(variation 1-3). The average value differs only by

1.4 s. The speed of the robot is most essential,

respectively the speed in the reduced mode. This has

the biggest influence.

Third, we recorded and compared the time the

radar systems needs to adapt to new workplaces.

When a new object is added to the supervised area of

the sensors, a new workspace is created. In this case,

the sensors have to be initialized and adapted to the

new environment. This is done automatically by the

sensors and can be monitored by the safe two-

channeled digital outputs of the controller. When the

environment changes, the digital outputs stay FALSE

until they are finally initialized and switched to

TRUE. A timer, which has been programmed on the

robot controller, supervises the digital outputs and

determines the time of initialization.

To create new workspaces in the laboratory, four

boxes of different sizes were selected (30x20x150,

40x30x220, 600x400x120, 600x400x320mm) and set

up with varied combinations on a table next to the

robotic cell. The latter is inside the supervised area of

the radar sensors. These boxes are standardized

according to VDA 4500 and are most commonly used

in the automobile industry and for high-automated

production processes. The number and size of the

boxes define the complexity of the combinations. The

higher the number and the bigger the size of the

boxes, the higher the complexity. In our set up, we

distinguish between four grades of complexity: low,

medium, high and very high. Figure 11 illustrates the

results of the initialization measurements for different

workplaces:

ROBOVIS 2021 - 2nd International Conference on Robotics, Computer Vision and Intelligent Systems

Figure 11: Time of initialization with different

complexities.

The median, which is represented by the red

horizontal lines, show that the smaller the change in

the scenario compared to the previous one, the faster

is the time of initialization. The interquartile range

(IQR) differs from 2 s (low), to 2.2 s (medium), 6.3 s

(high) and 15 s (very high). Therefore, the higher the

complexity the higher is the variance for the time of

initialization. Nevertheless, the measurements show

that the highest recorded time was 55 s. When a

worker is manually configuring the sensors at a new

workplace, e.g. a laser scanner to both sides instead

of the radar sensors, the time would be considerably

longer.

8 CONCLUSION

In summary, we presented a safety system that can be

flexibly used for different real industrial applications.

Because of the specific feature of the radar sensors,

no individual configuration is necessary, when the

mobile robotic cell is set up on a new workplace.

Furthermore, all the sensors to supervise the

workspace of the robot are mounted on the mobile

platform. This makes it easy, flexible and fast to be

integrated in an existing production line, because no

workplace has to be adapted. We also presented a first

approach to achieve a standardized CE-mark for a

component series. The laser scanner at the front also

reduces the supervised space in front of the robotic

cell. The advantage is that workers walking by via a

footway do not trigger the safety sensors. The entire

safety system has a high flexibility and mobility,

because it is usable on different workplaces with

different surrounding without an individual safety

consideration and without the adaption of configured

safety planes and safety areas of the radar and laser

sensors. The presented solution is usable in a real

industrial surrounding for the entire e-series of

Universal Robot, but is not limited to those Cobots.

Most of the Cobots have the possibility to program

safety planes that transfer the robot to the reduced

mode. Nevertheless, to improve the efficiency and the

usage of the presented safety system, more testing

with different Cobots is necessary.

For further research, the sensitivity level of the

radar sensors (normal, high, very high) in relation to

the speed and acceleration of the robot is supposed to

be analyzed. When the robot stops too quick or

accelerates too fast, the sensors mounted on the

framework are triggered, because of small vibrations

on the mobile cell. To avoid the undesired triggering

of the radar sensors and to reduce the cycle times

more research is necessary.

ACKNOWLEDGEMENTS

This work was conducted within the research project

AdhocMRK funded by the Kompetenzzentrum

Mittelstand GmbH (KME). We are thankful for the

support and funding from the KME.

REFERENCES

Amin, F., Rezayati, M., Venn, H., & Karimpour, H. (2020).

A Mixed-Perception Approach for Safe Human-Robot

Collaboration in Industrial Automation. In Sensors (S.

1-20). MDPI.

Antonelli, D., Astanin, S., Caporaletti, G., & Donati, F.

(2014). FREE: Flexible and Safe Interactive Human-

Robot Environment for Small Batch Exacting

Applications. In Springer Tracts in Advanced Robotics

(S. 47-62). Springer.

Berg, J., Richter, C., & Reinhart, G. (2018). Integration of

Safety elements into task-oriented programming system

for human-robot-collaboration. In International

Symposium on Robotics (ISR) (S. 281-284). Frankfurt

am Main, Germany: VDE.

DIN Deutsches Institut für Normung. (2009). Safety of

machinery - Anthropometric requirements for the

design of workstations at machinery (ISO 14738:

2009). Berlin: Beuth Verlag GmbH.

DIN Deutsches Institut für Normung. (2010). Safety of

machinery – Positioning of safeguards with respect to

the approach speeds of parts of the human body (ISO

13855:2010). Berlin: Beuth Verlag GmbH.

DIN Deutsches Institut für Normung. (2011). Robots and

robotic devices - Safety requirements for industrial

robots - Part 2: Robot systems and integration (ISO

10218-2:2011). Berlin: Beuth Verlag GmbH.

DIN Deutsches Institut für Normung. (2011). Robots and

robotic devices- Safety requirements for industrial

robots - Part 1: Robots (ISO 10218-1:2011). Berlin:

Beuth Verlag GmbH.

Portable Safety System using Radar for Flexible Human-Robot-Collaboration in a Real Semi-automated Production Line

DIN Deutsches Institut für Normung. (2016). Robots and

robotic devices - Collaborative robots (ISO/TS

15066:2016). Berlin: Beuth Verlag GmbH.

Dohi, M., Okada, K., Maeda, I., Fujitani, S., & Fujita, T.

(2018). Proposal of Collaboration Safety in a

Coexistence Environment of Human and Robots. In

IEEE International Conference on Robotics and

Automation (ICRA) (S. 1924-1930). Brisbane,

Australia: IEEE.

Glastechnik Hofmann GmbH. (2017). Chromos Industrial.

Von https://cobots.ch/en/casestudies/hofmann-

glastechnik-handling-of-highly-sensitive-glass-tubes/

abgerufen

Halme, R.-J., Lanz, M., Kämäräinen, J., Pieters, R.,

Latokartano, J., & Hietanen, A. (2018). Review of

vision-based safety systems for human-robot

collaboration. In Proceedings of the 51st CIRP

Conference on Manufacturing Systems (S. 111-116).

Elsevier.

Hoskins, G., Padayachee, J., & Bright, G. (2019). Human-

Robot Interaction: The Safety Challenge (An inegrated

frame work for human safety). In Proceedings of the

South Africa Regional Conference

(SAUPEC/RobMech/PRASA Conference) (S. 74-79).

Bloemfontein, South Africa: IEEE.

KUKA Systems GmbH. (2018). HRC in the production of

the BMW Group. (KUKA) Von https://www.

kuka.com/en-de/industries/solutions-database/2017/06

/solution-systems-bmw-dingolfing abgerufen

Kulic, D., & Croft, E. (2005). Real-Time Safety for Human-

Robot Ineraction. In 12th International Conference on

Advanced Robotics (S. 719-724). Seattle, USA: IEEE.

Lasota, P., Fong, T., & Shah, J. (2020). A survey of

methods for safe human-robot interaction. In

Christensen, H., Siegwart, R. (Eds.), Foundations and

Trends in Robotics (S. 261-349). Boston - Delft: Now

publishers Inc.

Matthias, B., Kock, S., Jerregard, H., Kallman, M.,

Lundberg, I., & Mellander, R. (2011). Safety of

collaborative industrial robots: Certification

possibilities for a collaborative assembly robot concept.

In IEEE International Symposium on Assembly and

Manufacturing (ISAM). Tampere, Finland: IEEE.

Pang, G., Yang, G., Heng, W., Ye, Z., Huang, X., Yang, H.,

& Pang, Z. (2021). CoboSkin: Soft Robot Skin With

Variable Stiffness for Safer Human–Robot

Collaboration. In IEEE Transactions on Industrial

Electronics (S. 3303-3314). IEEE.

PILZ. (2014). Safe camera system SafetyEYE. Von

https://www.pilz.com/en-

INT/eshop/00106002207042/SafetyEYE-Safe-camera-

system abgerufen

Rexroth. (2014). Intelligent systems for the flexible factory.

Von

https://apps.boschrexroth.com/microsites/apas/index.ht

ml abgerufen

Salmi, T., Väätäinen, O., Malm, T., Montonen, J., &

Marstio, I. (2013). Meeting New Challenges and

Possibilities with Modern Robot Safety Technologies.

In 5th International Conference on Changeable, Agile,

Reconfigurable and Virtual Production (S. 183-188).

Munich, Germany: Springer.

Vogel, C., Walter, C., & Elkmann, N. (2013). A projection-

based sensor system for safe physical human-robot

collaboration. In Proceedings of the IEEE/RSJ

International Conference on Intelligent Robots ans

Systems (IROS) (S. 5359-5364). Tokyo, Japan: IEEE.

Vogel, C., Walter, C., & Elkmann, N. (2017). Safeguarding

and Supporting Future Human-robot Cooperative

Manufacturing Processes by a Projection- and Camera-

based Technology. In Proceedings of the 27th

International Conference on Flexible Automation and

Intelligent Manufacturing (FAIM) (S. 39-46). Modena,

Italy: Elsevier.

Zlatanski, M., Sommer, P., Zurfluh, F., & Madonna, G.

(2018). Radar Sensor for Fenceless Machine Guarding

and Collaborative Robotics. In IEEE International

Conference on Intelligence and Safety for Robotics

(ISR) (S. 19-25). Shenyang, China: IEEE.

ROBOVIS 2021 - 2nd International Conference on Robotics, Computer Vision and Intelligent Systems