CAPLOS

Compressed Air Powered Lateral Obstacle Simulator

Marc Schneider, Markus Sieber and Berthold F

arber

Human Factors Institute, Universitaet der Bundeswehr Muenchen, Werner-Heisenberg-Weg 39, Neubiberg, Germany

Keywords:

CAPLOS, Compressed Air Powered Lateral Obstacle Simulator, Driver Behaviour, Test Track.

Abstract:

The development process of advanced driver assistance systems (ADAS) should include user studies in order to

ensure optimal human-machine-interaction. Relevant active safety test scenarios must be based on appropriate

use-cases, i.e. critical trafﬁc situations. In order to avoid endangering subjects’ physical well-being, test

equipment is required that reliably simulates obstacles with realistic appearance that are also safe to crash. For

cross-trafﬁc scenarios and the examination of lateral driving behavior, existing test equipment, such as driving

simulators, exhibit certain deﬁcits. The authors present a compressed air powered lateral obstacle simulator

(CAPLOS), a low-cost system for the creation of incursion-based critical situations. This paper describes the

setup of CAPLOS and demonstrates its suitability for use in active safety ADAS user studies with exemplary

driving study data.

1 INTRODUCTION

Advanced driver assistance systems (ADAS) that pro-

vide active safety functions reduce the danger posed

by critical situations, e.g. by helping prevent col-

lisions. During an intervention, these systems co-

operate or compete with the driver for control over

the vehicle and thus have high requirements regard-

ing the design of the human-machine-interaction. To

ensure that these requirements are met, the ADAS de-

velopment process must include user studies (Koenig,

2012). For these driving experiments, the critical sit-

uation in which the ADAS is designed to intervene

must be reproduced in a way that appears realistic to

the driver in order to elicit realistic driver behavior.

At the same time, the situation must be reproduced in

a manner that does not expose the test subject and test

equipment to any real danger.

2 STATE OF THE ART

Several different approaches have been put forward to

provide ADAS developers with the possibility to test

systems in early development stages.

One such approach is the use of driving simula-

tors. Being virtual test beds, driving simulators nat-

urally provide the possibility to expose test subjects

to critical trafﬁc scenarios in the absence of any real

physical threat. They suffer, however, from serious

limitations regarding the simulation of prolonged or

intense vehicle dynamics and are therefore unable to

provide fully realistic kinesthetic force feedback for

highly dynamic maneuvers that are usually inherent

to active safety scenarios. While having severaly ad-

vantages, driving simulators are therefore not deemed

appropriate for a conclusive evaluation of driver be-

havior, see e.g. (Schmitt et al., 2006).

Another approach is the use of safe to crash ob-

stacles in ﬁeld studies. With test subjects driving a

real car, vehicle dynamics need not be simulated and

are fully realistic. The challenges with this approach

are the reliable reproduction of obstacle placement as

well as the realistic appearance of the obstacle. Sev-

eral studies report the use of foam cuboids or balloon

cars, e.g. (Eckert et al., 2011), but do not mention how

the obstacle was reliably placed in the correct position

at the correct time regardless of external inﬂuences

such as wind, or which measures were employed for

the obstacle to have the appearance of a solid object

with which a crash must be avoided.

A more elaborate and expensive piece of test

equipment for the use in ﬁeld studies is EVITA (Hoff-

mann and Winner, 2008), a system designed for use

on a test track in order to investigate critical scenarios

in longitudinal trafﬁc. It consists of a towed dummy

car that can be released and then perform an intense

braking maneuver, creating a critical situation for the

Schneider, M., Sieber, M. and Färber, B.

CAPLOS - Compressed Air Powered Lateral Obstacle Simulator.

In Proceedings of the International Conference on Vehicle Technology and Intelligent Transport Systems (VEHITS 2016), pages 71-78

ISBN: 978-989-758-185-4

Figure 1: CAPLOS main body.

following car, while being constructed in a way that

cushions a potential collision in order to avoid any

physical harm. EVITA keeps test subjects safe while

simultaneously making a very realistic impression.

Unfortunately, it is not designed for cross-trafﬁc sce-

narios. Test subjects driving behind EVITA are aware

of the possible source of danger in front of them at all

times.

Similarly, other previously existing solutions are

equally expensive designs and/or are based on con-

structions that make test subjects aware of possible

danger, such as rope or cable constructions on or

above the road. The authors desired an easy to con-

struct and inexpensive piece of test equipment for the

use in ﬁeld studies that can be reliably placed, has re-

alistic appearance while being safe to crash, and can

easily be hidden from sight behind a wide variety of

roadside structures in order to elicit realistic driving

behavior with real vehicle dynamics in cross-trafﬁc

scenarios with unsuspecting drivers.

3 CAPLOS

The authors designed CAPLOS as a low-cost solution

to safely, validly and reliably test realistic driver be-

havior to avoid collisions in cross-trafﬁc scenarios.

3.1 Objectives

CAPLOS was designed as a system to help exam-

ine driver behavior in reaction to suddenly appear-

ing obstacles. The objective was to build a system

that can create a critical situation by making an ob-

stacle appear suddenly, while being low-cost, easy to

set up, easy to hide behind cover or camouﬂage, that

is safe for drivers and the test vehicle, provides reli-

able obstacle placement, is able to ﬁt different types

of dummy obstacles, makes a realistic impression on

drivers, and provide different triggering options. It

was a requirement that the length of stroke for the

suddenly appearing dummy obstacle be more than 1m

and that there are multiple safety mechanisms: the

system should provide the ability to pull the dummy

obstacle out of harm’s way in the last possible instant,

while the dummy obstacle should also be risklessly

crashable.

3.2 Solution

The proposed solution named CAPLOS (compressed

air powered lateral obstacle simulator, Fig. 1 + 2) con-

sists of 2 pneumatic pistons, a lever-construction, a

pushing rod with ﬁtting bearings, a frame to direct and

stabilize the translation, an air compressor to provide

power, an associated control unit and a mobile and

crashable dummy at the front. A vehicle approaching

and falling below a certain time-to-collision (TTC) to

CAPLOS triggers the system, causing the movement

of the dummy obstacle into the road. In the last in-

stant before a possible collision or in case of electrical

shutdown the pneumatic mechanism automatically re-

tracts the dummy obstacle for safety reasons.

VEHITS 2016 - International Conference on Vehicle Technology and Intelligent Transport Systems

Figure 2: Construction drawing of CAPLOS (Schreiber, 2015).

3.3 CAPLOS - Structure

3.3.1 CAPLOS - Frame

The frame (Fig. 3) is connected to a steel base-

plate, which provides stability and weight for the sys-

tem while at the same time it can be separated into

more easily manageable parts for transportation. The

frame itself, built from aluminum proﬁles, ensures the

straight and controlled translation of the pushing rod

through the exact positioning of the required bearings.

The rigid construction guarantees functionality even

in the hypothetical case of a clamp in the linear move-

ments or the dummy getting hit by a vehicle, ensuring

the safety of test subjects, bystanders, the vehicle and

CAPLOS itself.

Figure 3: Schematic diagram of the CAPLOS frame.

3.3.2 CAPLOS - Power Unit

An off-the-shelf air compressor with 10 bar maxi-

mum pressure is connected to the pistons through

control valves and serves as the power source. In

order to reach the required speed when pushing out

the dummy or pulling it in again, a work-principle

consisting of linear pneumatic pistons was selected.

Through a conversion of the translation’s length and

speed via the lever construction (Fig. 4 + 5) the re-

quired speed dimensions can be reached even with

low-cost pneumatic machinery. The adjustable con-

nection of the pistons to the lever enables mechanical

regulation of length of stroke and movement speed.

This lever construction is connected to the push-

ing rod with a guiding rod. In combination with the

basic frame and rigid implementation of the joints in

the construction, straight friction-reduced movements

are ensured. At the opposite end of the main pushing

rod a dummy of arbitrary dimensions (car silhouette,

pedestrian, ...) can be equipped.

3.3.3 CAPLOS - Control Unit

CAPLOS is controlled by electronic and pneumatic

components built into a case (Fig. 6). In order to

regulate the position of both pneumatic pistons a 5/3-

way-valve (to switch between pushing and pulling)

and a 3/2-way-valve (to convert an electrical trigger

signal to a pneumatic signal) were used in the control

CAPLOS - Compressed Air Powered Lateral Obstacle Simulator

Figure 4: CAPLOS lever construction.

Figure 5: Schematic diagram of the CAPLOS lever con-

struction.

unit of CAPLOS. An electric signal activating the 3/2-

valve causes the 5/3-valve to switch to push out mode.

To switch the state of the 3/2-way-valve a voltage of

12V to 24V is provided by the built-in power supply

unit, which also supplies the additional light barriers.

With the electric signal dropping out, caused either by

CAPLOS receiving the signal to pull back or by the

failing of the electrical power source, CAPLOS auto-

matically returns to pull-in mode by default, thereby

preventing possible harm from system breakdown.

The electrical trigger signal is provided by a Rasp-

berryPi minicomputer’s GPIO-pins or USB-port. The

control unit continually monitors whether the trigger

conditions are met. The required data can be transmit-

ted to the RaspberryPi in several ways, two of which

have so far been implemented for CAPLOS:

• via a WiFi signal sent by the test vehicle (Fig. 7)

• via light barriers and speed-calculation within the

RaspberryPi (Fig. 8)

The system’s activation via WiFi signal is realized

with high-range antennas. The light barriers can be

connected to the RaspberryPi directly with a 5 wire

cable. The experimenter deﬁnes the desired trigger

criteria, for example a certain time-to-collision (TTC)

or distance of the ego vehicle to CAPLOS.

Figure 6: Control case of CAPLOS.

Figure 7: Schematic diagram of activation via WiFi.

Figure 8: Schematic diagram of activation via light barriers.

3.3.4 CAPLOS - Dummy

The dummy obstacle, which is positioned at the front

end of CAPLOS’s pushing rod, enables the machine

to carry multiple types of dummies like car silhou-

ettes (Fig. 9) or pedestrian puppets (Fig. 10). In order

to prevent damage to the test vehicle or CAPLOS in

case of the vehicle hitting the dummy, it is comprised

of two distinctive parts linked by hinges, enabling the

obstacle to fold away when hit, thereby signiﬁcantly

reducing the force on the dummy as well as the test

vehicle. Hook-and-loop fasteners ensure stable cohe-

sion of the dummy parts until a collision occurs.

Guidance of the dummy obstacle in lateral and

vertical direction is realized through three attached

wheels and the use of a hinge-construction (Fig. 11).

The wheels provide the required stability while the

VEHITS 2016 - International Conference on Vehicle Technology and Intelligent Transport Systems

Figure 9: Exemplary CAPLOS dummy obstacle: passenger

car.

Figure 10: Exemplary CAPLOS dummy obstacle: pedes-

trian.

Figure 11: Schematic diagram of the hinge-construction be-

tween pushing rod and dummy obstacle.

hinge-construction compensates for unevenness of

the ground.

Table 1 shows mechanical property data based on

static and dynamic simulations of the facility.

Table 1: Performance data from simulation.

mean speed 3 m/s

maximum speed 10 m/s

minimum distance 0.8 m

maximum distance 1.3 m

minimum piston power 1120 N (2 bar)

maximum piston power 5600 N (10 bar)

4 SUITABILITY OF CAPLOS

Suitability of CAPLOS for the simulation of lateral

incursion-based critical scenarios was evaluated using

data collected in a driving study that was conducted

on the test track of the Universitaet der Bundeswehr

Muenchen.

4.1 Driving Experiment

60 test subjects were instructed to drive along a

straight rural road with two lanes at 50 km/h. They

were told that the object of the study was to examine

differences in lane keeping between right-hand and

left-hand trafﬁc with a left-handed vehicle. CAPLOS

was hidden from sight behind a stationary car parking

perpendicularly at the side of the road approximately

100m from the exit of the double lane change traf-

ﬁc cone corridor. The subject car was instrumented

with an inertial measurement unit fed with GPS as

well as dGPS correction data from a nearby base sta-

tion, ensuring an accuracy of at least 0.02m in posi-

tional tracking. When the combination of the car’s

distance to the known position of CAPLOS and its

current speed yielded a TTC of 1.6s or less, CAPLOS

was activated via a WiFi connection. Once CAPLOS

was activated, a fotorealistic silhouette of a passenger

car was pushed out into the road, partly blocking the

subject car’s lane and forcing the driver to react to the

suddenly appearing obstacle in order to avoid colli-

sion (Fig. 12). Four groups of 12 test subjects each

experienced different system-induced warnings in re-

action to the obstacle, the remaining 12 test subjects

were left unassisted.

Figure 12: CAPLOS in action.

CAPLOS - Compressed Air Powered Lateral Obstacle Simulator

The test subjects experienced this situation a sec-

ond time after a series of uneventful laps. Data from

both expositions was used to evaluate CAPLOS’s

suitability for the simulation of critical cross-trafﬁc

scenarios in test subject driving studies.

4.2 Suitability Measures and Results

Reliability regarding the positioning of the dummy

obstacle is assumed to be near perfect, as it is held

ﬁrmly in place by the mechanical build of CAPLOS

and the length of stroke is set to a ﬁxed value by the

length of the cylinder pistons and the transformation

ratio within the lever construction.

As a ﬁrst measure of suitability, the machine’s re-

liability was ascertained using video data from a car-

mounted camera capturing the scene in front of the

subject car from 60 experimental trials with different

temperature and weather conditions. Frame-by-frame

analysis was used to determine when the dummy car

ﬁrst started moving and when it reached its further-

most position, yielding the machine’s reaction time

(time between the subject car falling below a TTC

of 1.6s and ﬁrst movement of the CAPLOS dummy

obstacle) and its expansion time (time between ﬁrst

movement of the dummy obstacle and its reaching the

fully expanded position) respectively.

Both reaction time (n=60, M=0.406s, SD=0.040s)

and expansion time (n=60, M=0.726s, SD=0.066s)

were found to be highly reliable despite variance

caused by differences in latencies of the WiFi signal

(which can not be controlled), differences in air pres-

sure (pressure in the pressure tank was only replen-

ished to its nominal value once it fell below a certain

threshold, allowing for some variance), low data fre-

quency (frame rate of the video feed was 25 frames

per second, resulting in 0.04s steps of analysis) and

possible observer imprecision (especially regarding

ﬁrst movement of the CAPLOS dummy obstacle).

No such data were publicly available from similar

existing machines for comparison. Performance was

therefore compared to previous preliminary tests. The

setup used mere positional triggering (not accounting

for the exact subject car speed) and a lab assistant to

tip over a rubber foam obstacle onto the road. Af-

ter an extensive training phase, twenty sets of ﬁve tri-

als each were performed over the course of two days.

Only the third to ﬁfth trials from each set were used

for comparison. Four of these sixty trials failed. In

the remaining, ’reaction time’ (time between subject

car dropping below desired TTC and ﬁrst obstacle

movement) was, unsurprisingly, found to vary to a

much greater extent (n=56, SD=0.284s). Also, ’ex-

pansion time’ (time between ﬁrst movement of the

obstacle and its reaching the ﬁnal position) was found

to be similar on average, but also more varied (n=56,

M=0.753s, SD=0.172s).

As a second measure of suitability, it was ascer-

tained whether the situation induced by the sudden

appearance of the CAPLOS dummy obstacle was per-

ceived as critical by the test subjects. Drivers were

asked to judge the criticality of the experienced situa-

tion using the scale for criticality assessment of driv-

ing and trafﬁc scenarios (Neukum et al., 2008), where

test subjects ﬁrst classify the situation as either imper-

ceptible, harmless, unpleasant, dangerous, or uncon-

trollable and, in a second step, indicate possible ten-

dencies towards a lower or higher category within the

three middle categories, resulting in a 11-point scale

(Fig. 13).

Figure 13: Scale for criticality assessment of driving and

trafﬁc scenarios (Neukum et al., 2008).

In total, less than 5% of the test subjects rated the

experienced situation as harmless. Almost two thirds

(63%) rated the situation as dangerous or uncontrol-

lable. Fig. 14 shows the test subjects’ ratings for each

exposition to CAPLOS in a box plot.

Figure 14: Criticality ratings.

VEHITS 2016 - International Conference on Vehicle Technology and Intelligent Transport Systems

Figure 16: CAD-model of the improved CAPLOSm (Walter, 2015).

As a third measure of suitability it was ascertained

whether the test subjects were surprised by the crit-

ical situation induced by CAPLOS. The Facial Ac-

tion Code (Ekman and Friesen, 1976) is a system

developed for measuring all facial movement using

anatomically based ’action units’. Noticeable raising

of the inner as well as the outer eye-brows, widen-

ing of the eyes by raising of the upper eye-lid, and

dropping of the jaw (corresponding to action units

1, 2, 5, and 26 in the Facial Action Coding System

(Ekman et al., 2002) respectively) were regarded as

signs of surprise (or fear) in subjects’ facial expres-

sions (Fig. 15) in response to the CAPLOS dummy

obstacle’s moving into the road. Facial expressions

were recorded by a camera aimed at the driver with a

frequency of 25 frames per second and were analyzed

in slow motion (50% speed).

Even in the second exposition to CAPLOS 49 out

of 60 test subjects were found to show perceivable

signs of surprise in facial expression. The relative

amount was even higher (11 out of 12) for the unas-

sisted group, providing evidence that the surprise was

in response to the critical situation elicited by CAP-

LOS and not to the sudden intervention of an assis-

tance system.

Figure 15: Exemplary subject facial expression in reaction

to CAPLOS exposition.

5 CONCLUSION

The presented setup for CAPLOS constitutes a low-

cost test facility that permits reliable sudden position-

ing of different types of safe-to-crash dummy obsta-

cles with different triggering options. The data col-

lected in a driving study featuring CAPLOS further

provide evidence, that CAPLOS is able to create traf-

ﬁc scenarios which test subjects view as critical and

surprising. Bearing in mind that these data are based

CAPLOS - Compressed Air Powered Lateral Obstacle Simulator

on multiple expositions to CAPLOS without active

distraction of the drivers, the authors conclude that

CAPLOS can be regarded as a valuable instrument

suitable for user studies in active safety ADAS devel-

opment and evaluation.

This setup of CAPLOS was, however, built partly

from scrap materials and parts scavenged from other

machinery and was therefore restricted in regard to

optimal design. The authors’ newest work includes

the construction of a new mobile version of CAPLOS

(CAPLOSm Fig. 16) mounted on top of a trailer. This

allows for a more efﬁcient and more robust base de-

sign (Fig. 17) as well as further signiﬁcant improve-

ment concerning ease of deployment. A change in the

leverarm design (Fig. 18) allows for greater variation

in extension length. Moreover, the authors intend to

include a mobile power source with CAPLOSm for

independent operation in any location.

Figure 17: New robust base design of CAPLOSm (Walter,

2015).

Figure 18: CAPLOSm - leverarm and piston mechanism

(Walter, 2015).

REFERENCES

Eckert, A., Hartmann, B., Sevenich, M., and Rieth, P.

(2011). Emergency steer & brake assist: a system-

atic approach for system integration of two comple-

mentary driver assistance systems. Washington DC.

International Technical Conference on the Enhanced

Safety of Vehicles.

Ekman, P. and Friesen, W. V. (1976). Measuring facial

movement. pages 56–75. Environmental Psychology

and Nonverbal Behavior.

Ekman, P., Friesen, W. V., and Hager, J. C. (2002). Facial

action coding system: The manual on cd rom. Salt

Lake City.

Hoffmann, J. and Winner, H. (2008). Evita - Das Unter-

suchungswerkzeug fuer Gefahrensituationen. Garch-

ing. Aktive Sicherheit durch Fahrerassistenz.

Koenig, W. (2012). Nutzergerechte Entwicklung der

Mensch-Maschine-Interaktion von Fahrerassisten-

zsystemen. In Handbuch Fahrerassistenzsysteme,

pages 33–42, Wiesbaden. Vieweg+Teubner Verlag.

Neukum, A., Luebbeke, T., Krueger, H.-P., Mayser, C., and

Steinle, J. (2008). Acc-stop&go: Fahrerverhalten an

funktionalen Systemgrenzen. pages 141–150. Work-

shop Fahrerassistenzsysteme.

Schmitt, J., Faerber, B., Maurer, M., and Breu, A. (2006).

Menschliches und technisches Verhalten an den Sys-

temgrenzen eines FAS. pages 1960:563–579. VDI-

Berichte.

Schreiber, P. (2015). Entwicklung einer mechanischen Ap-

paratur zur Darstellung ploetzlicher Hindernisse im

Rahmen der Fahrerassistenzerprobung (student work).

Neubiberg. Human Factors Institute, Universitaet der

Bundeswehr Muenchen.

Walter, M. (2015). Bau und Dokumentation einer Versuch-

sapparatur zur Fahrerassistenzssystemerprobung (stu-

dent work). Neubiberg. Human Factors Institute, Uni-

versitaet der Bundeswehr Muenchen.

VEHITS 2016 - International Conference on Vehicle Technology and Intelligent Transport Systems