Automated Bicycle Counting System’s Prototype to Evaluate the

Necessity of New Bicycle Lanes in Jelgava City

Armands Kviesis, Aleksejs Zacepins, Vitalijs Komasilovs, Normunds Vetra and Nikolajs Bumanis

Department of Computer Systems, Faculty of Information Technologies,

Latvia University of Life Sciences and Technologies, Jelgava, Latvia

Keywords:

Automatic Bicycle Counting, Pressure Sensor, Pneumatic Tubes, Open Source Technologies.

Abstract:

Every year the number of vehicles on the road is increasing. But there are people that start to choose public

transport or greener transportation options like bicycles or electric cars over typical fuel cars. Regarding

bicycle usage, the problems that arise are related to insufﬁcient bicycle lanes and determination of future

lane locations, so that the resources used for bicycle lane construction would be properly invested. To resolve

mentioned problems municipality ﬁrst need to learn where the suitable bicycle lane location for cyclists should

be. Such a task can be done by applying a cyclist counting system. This paper describes a portable automatic

cyclist counting system’s prototype for bicycle lane location planning and also identiﬁes the limitations for

such a system. Proposed prototype is based on rubber tubes and pressure sensors, Wi-Fi module and open

source electronic platform Arduino. This study is carried out within the ERANet-LAC project RETRACT

(Enabling resilient urban transportation systems in smart cities).

1 INTRODUCTION

Urban air pollution is one of the environmental risks

in modern cities (Elsom, 2014). There are several

factors affecting urban air, like population growth, in-

dustrial expansion, urbanisation and increase in motor

vehicle usage (Bickerstaff and Walker, 2001). Reduc-

tion of vehicle emissions is a major component of sus-

tainable transportation development (Kitthamkesorn

and Chen, 2017).

Many governments promote usage of green trans-

port, like bicycles for citizen daily transportation

(Goodwill, 2015; Kitthamkesorn and Chen, 2017).

Municipalities are investing funds (for instance Dan-

ish capital Copenhagen invested 150 million USD in

cycling infrastructure) to develop new separate bike

lanes or even bridges or trying to improve existing in-

frastructure and adapt it also for bicycle needs (Bao

et al., 2017). Bike along with walking is the greenest

mode of transportation (https://greenseat.nl/en/how-

to-travel-green/). Many municipalities are changing

their development strategies and deﬁne new mobility

principles in cities, where mobility priority focuses on

pedestrians, cyclists and public transport not on pri-

vate car drivers. Traditional ways of planning bike

lanes in a city rely mainly on empirical experience or

citizen surveys (Sustrans, 2014; Hyodo et al., 2000;

Rybarczyk and Wu, 2010). With the growth of smart-

phone and smart GPS applications, more data-driven

approaches on planning bike lanes emerged (Evans

et al., 2012; Evans et al., 2013; Jiang et al., 2016; Bao

et al., 2017).

New bike lane planning requires the municipality

to have some data about the number of bicycles on a

speciﬁc road/walkway. It is highly important to in-

vest in right directions, because building of 1 km bike

road costs approximately 145000 EUR, but each de-

veloped bike road increase the bikers count for 20%

and decrease car number for 10%, according to the

data by the Latvian Cycling Association (www.rdsd.

lv/uploads/media/551901879d5eb.ppt). As stated by

(Weigand et al., 2013) it is not trivial to precisely de-

termine the cost of integrating a bike lane into existing

infrastructure. This is also proved by several sources

as the costs are closely related to project design and

other speciﬁcs, for example, as it was estimated by

City of Portland, the cost for a bike lane per foot (ap-

prox. 0.30 m) is approximately 3 USD; the cost can

be in a range from 5000 USD to 535000 USD (with

average of 130000 USD) per mile (approx. 1.6 km)

for about 1.5 m wide bike lane (Bushell et al., 2013)

or even much higher, like the 12 USD million-a-mile

bike lanes (Alexander, 2018).

There are many methods and systems (piezoelec-

420

Kviesis, A., Zacepins, A., Komasilovs, V., Vetra, N. and Bumanis, N.

Automated Bicycle Counting System’s Prototype to Evaluate the Necessity of New Bicycle Lanes in Jelgava City.

DOI: 10.5220/0007714104200425

In Proceedings of the 5th International Conference on Vehicle Technology and Intelligent Transport Systems (VEHITS 2019), pages 420-425

ISBN: 978-989-758-374-2

tric strips, radio beams, active infrared, pneumatic

tubes etc. (Ryus et al., 2014)) available for bike

counting. Over the last decade, there has been in-

creasing interest in counting bicycles and establishing

non-motorized counting programs (Nordback et al.,

2016). Three types of commercially available pneu-

matic tube counters (dual and single tube conﬁgu-

ration - bicycle-speciﬁc, classiﬁcation and volume

counters) were studied by (Nordback et al., 2016),

where counting accuracy decreased with high trafﬁc

volume and longer tubes. The undercounting problem

is also highlighted by (Ryus et al., 2014) where two

products were compared. Counting can be done man-

ually or in automated way. Automated counting make

it possible to collect data for long time period and to

monitor temporal data variety in volumes more effec-

tively than manual counts (Proulx et al., 2016). Auto-

mated counting can be done using different technolo-

gies, like infrared, radio beam, pneumatic tubes, in-

ductive loops and also image processing (Zangeneh-

pour et al., 2015; Komasilovs et al., 2018). As well

publicly showing the number of bikes going on the

speciﬁc roads or comparing speed of bike and car on

the speciﬁc part of the road can increase individual

motivation to use the bike next time.

By the data of Latvian Central Sta-

tistical Bureau, 17% of Latvian citizens

almost twice per week are using bikes

(https://www.csb.gov.lv/lv/statistika/statistikas-

temas/socialie-procesi/veseliba/meklet-tema/2141-

17-latvijas-iedzivotaju-vismaz-divreiz-nedela), in

capital Riga number of cyclist’s increases approx. by

10% each year. In Latvia bike counting is mainly

done manually by persons standing at the bike lanes

and counting the bicycles, but this approach has

many disadvantages (https://www.diena.lv/raksts/

latvija/zinas/ritenbrauceju-skaita-izmainas-ﬁkse-ar-

masinredzi -un-maksligo-intelektu-14110736). It is

not possible to collect data on the long time period,

to compare data on different weather conditions and

day periods, as well it require more staff time per

hour of data collected. Therefore, local municipality

is searching for better ways on how to collect data

on the number of bikes using speciﬁc roads with

the purpose to build bike lanes on the most popular

roads/sidewalks.

Pneumatic or rubber tubes are commonly used for

gathering short-duration motor-vehicle number (http:

//www.windmill.co.uk/vehicle-sensing.html), but it is

possible to adapt this technology also for bicycle

counting. Equipment for this task consists of two

main elements: rubber tubes to place on the road sur-

face and a data recording and processing unit. As ob-

ject (vehicle or bicycle) pass over the tubes, pulses

of air travel through the tubes to the data recorder,

which detects them due to change in the pressure.

Authors of this research propose a prototype based

on a popular open-source electronic platform Arduino

(https://www.arduino.cc/) for bicycle counting. Main

advantage of such approach could be the price of the

whole system, which gives beneﬁt for local munici-

pality.

2 MATERIALS AND METHODS

Proposed automatic bicycle counting system’s archi-

tecture (see Fig. 1) consists of two pressure sensors

MPX5010DP with rubber tubes, electronics develop-

ment platform Arduino Uno, ESP8266 Wi-Fi module,

remote server with Laravel Framework and MariaDB

in the back end, and HTML, CSS, Javascript, Boot-

strap forming the front end.

Figure 1: Architecture of bicycle counting system’s proto-

type.

Arduino platform acts as a data collector and

at the same time is capable of calculating bicycle

speed, direction and count at the given time. Men-

tioned pressure sensor is a piezo resistive trans-

ducer that outputs analogue signal proportional to

the pressure applied (https://www.nxp.com/docs/en/

data-sheet/MPX5010.pdf). The data collection part is

similar as posted in https://hackaday.io/project/4567-

trafﬁc-counter-road-tube. For testing purposes, a Wi-

Fi technology for wireless data transfer was used. In

remote locations, the Wi-Fi can still be used, when,

for example, a mobile phone is turned into a portable

hotspot and sharing cellular data plan. Other solu-

tion would be to choose a GPRS module. Since

Arduino Uno board itself lacks Wi-Fi functionality,

data were transferred to the remote server by us-

ing a Wi-Fi module (in this case Adafruit HUZZAH

ESP8266), which is connected to the Arduino via Uni-

versal Asynchronous Receiver-Transmitter (UART).

After a certain period of time (deﬁned in the soft-

ware), Arduino communicates with ESP8266 module

and sends the necessary data (direction, count, speed).

Automated Bicycle Counting System’s Prototype to Evaluate the Necessity of New Bicycle Lanes in Jelgava City

421

The Wi-Fi module then prepares HTTP message and

executes POST request to the remote server (example

is shown below).

POST /api/data HTTP/1.1

Host: example.host.com:port

Content-Type: application/json

{

"count":5,

"direction":1,

"speed":12

}

Remote server processes the incoming requests

and inserts data into the database. The front end can

provide user with information about cyclist directions

during selected day and cyclist count by hours as well

(see Fig. 2).

Figure 2: Bicycle count chart representation.

The MPX5010DP pressure sensor can measure up

to 10 kPa pressure with a response time of 1 ms and

in combination with rubber tubes is suitable for the

task. Tests were performed to determine the pres-

sure applied to the sensor tubes by bicycle. As it

was found, pressure did not exceed 2 kPa (as mea-

sured and calculated by Arduino using the given trans-

fer function (1) (https://www.nxp.com/docs/en/data-

sheet/MPX5010.pdf):

out

= V

∗(0.09P+0.04)±(P

error

∗T

∗0.09V

) (1)

where:

out

– output voltage from sensor, V;

– supply voltage (5.0 V ± 0.25 Vdc), V;

P – pressure, kPa;

error

– pressure error, kPa;

– temperature factor,

◦

As it was concluded, such values fall within the

sensor minimum and maximum range and the pres-

sure applied by bicycle is enough to detect changes.

However, precise pressure values (in kPa) are not

very important, since this sensor application mostly

requires the detection of pressure being applied (in

other words: True or False). The analogue value of

the sensor is required in terms of calibration - it helps

to adjust the threshold when the tube is pressed.

Two sets of tubes (different material and diam-

eter) were used for testing, those were of material

polyurethane/rubber (further in text ”Tube 1”) and

rubber (”Tube 2”). The length of the tubes were 3

m (Tube 1) and 2 m (Tube 2); outer diameter 6 mm

(Tube 1) and 10 mm (Tube 2); thickness 1.2 mm

(Tube 1) and 2mm (Tube 2).

In order to set up the counting system for proper

operation, it was critical to place the tubes so that the

distance between them is constant. This distance is

hard-coded in the system for precise speed (if needed)

calculations. During testing, tubes were placed on a

concrete surface, since in Jelgava city we mostly have

walkways of such material. Setup of the system is

shown in Fig. 3. As it can be seen in the ﬁgure, the Ar-

duino is not in an enclosed box, since the testing phase

was aimed to test the working principle and identify

the limitations. The cost of the tested prototype was

approximately 100 EUR.

Figure 3: Counting system’s setup for testing.

3 RESULTS AND DISCUSSION

As mentioned previously, two different rubber tubes

were selected to attach to pressure sensors. Tube se-

lection is very important, since they are made of dif-

ferent materials. Material choice adds extra factors

that can have impact on system’s correct operation

and thus needs to be taken into account. Factors, such

as size and ﬂexibility of the tubes, can be mentioned.

The smaller and harder are the tubes, the less pressure

can the bicycle tire apply to them. Another factor is

VEHITS 2019 - 5th International Conference on Vehicle Technology and Intelligent Transport Systems

422

temperature, that has impact on pressure, as this is de-

ﬁned by Gay-Lussac’s law (Pressure Law). In colder

environment, pressure decreases (if the volume and

mass is constant). Results section is divided in two

subsections - tests with both material tubes.

3.1 Tests with Tube 1

At the beginning the system was tested by simply

rolling over a bicycle, to see how the air pressure

changes. In this ﬁeld test with Tube 1 (made of

polyurethane) it was concluded that pressure values

changed minimally or not even at all. Tempera-

ture also affected the tubes – the material became

much harder and the pressure decreased. As this was

observed by bringing the tubes from warmer room

(23

◦

C) to the outside environment (5

◦

C). It was con-

cluded that in this test case bicycle tire did not have

any effect on the tube, thus no changes in sensor read-

ings were noticed. Therefore it was impossible to use

such tubes for reliable system operation.

3.2 Tests with Tube 2

Single Rubber Tube (bicycle count only)

If information about bicycle speed and direction is

not important, then a single sensor with one tube is

enough. During ﬁeld tests all single bicycle riding

cases were counted correctly, but some speciﬁc situ-

ations, conditions where recognized, that causes in-

terruptions of the systems operation. Thus following

limitations were identiﬁed:

• only count can be determined. Single tube limits

the operation of the system regarding the details

that can be obtained from data. It is not possible

to detect direction the cyclist is driving from and

speed as well. Although it is possible to know the

time that takes both tires to cross the tube, it is

not enough to calculate the speed, as the distances

between tires vary almost with each bicycle;

• simultaneous crossing of multiple cyclists. Sys-

tem counts cyclists with error when several bicy-

cles are crossing the tube simultaneously. Sensor

reading in such situation will reach the threshold,

but in order to distinguish the difference between

one or two cyclists it would mean to hard-code

sensor values. And as tests showed, there are no

distinguishable differences when two cyclists are

crossing the tube simultaneously, because when

the tube is pressed by one cyclist, the other cy-

clists pressure does not show any impact on the

sensors value.

Two Rubber Tubes (direction, count, speed)

In order to detect also speed, two rubber tubes were

placed on a concrete surface with distance of 50 cm

between them. Since bicycles are in various lengths

(also the reason why speed can not be detected when

using only one sensor), such a distance should ensure,

that both tires are not on the sensor tubes simultane-

ously (in this scenario the system will fail to count

such case). Total of 50 bicycle riding tests were car-

ried out. Tests included driving with various speeds –

from 6 to 20 km/h. Speed was calculated within the

Arduino by formula (2):

v =

(2)

where:

v – speed, m/s;

s – distance between rubber tubes, m;

t – time it took for the ﬁrst tire to cross both

tubes, s.

All 50 testing cases with single cyclist riding over

the tubes were successfully identiﬁed and counted.

During testing phase the following limitations were

identiﬁed:

• threshold adjustments. As it was observed, in

colder weather the pressure decreased after longer

periods of time. It was detected by the analogue

values received by the pressure sensors. This has

an impact on deﬁned threshold – if the pressure

decreases dramatically the threshold could never

be reached, thus cyclists can not be detected;

• pedestrian factor. Since described system is in-

tended to be placed on pedestrian walkway that

also is used by cyclists, pedestrians can intervene

by stepping on tube(-s). There are possible solu-

tions to ﬁlter these cases out. If pedestrian only

steps on one tube leaving the other untouched, the

system should check time till the other tube is ac-

tivated. If the time exceeds some deﬁned limit,

this case should be ignored;

• cyclist behavior. There may be cases when cy-

clists behave differently when they see tubes on

the road/walkway. They could try to avoid cross-

ing them or drive them over with one tire. But

these also could only be some rare, speciﬁc cases

(enthusiasts of ”extreme” bike riding).

The system was also tested when multiple cyclists en-

tered the scene. It was observed that system does not

operate reliably in such situations: only about 40% of

50 testing cases (riding from the same and opposite

direction) were correctly recognized as two cyclists.

Such poor performance was due to the fact, that ana-

log sensor is being used and detection is highly de-

pendent on a threshold. As a result, if the sensor’s

Automated Bicycle Counting System’s Prototype to Evaluate the Necessity of New Bicycle Lanes in Jelgava City

423

value is being read frequently, e. g., each 10 ms, then

one cyclist’s applied pressure that is greater than the

threshold, could be counted as multiple cyclists, be-

cause the value does not stabilize so fast. Sensor val-

ues, after a single bicycle has driven over the tubes,

is illustrated in Fig. 4. In a single cyclist case, this

can be resolved by ignoring the speciﬁc sensor values

after the threshold is reached, and continue to read

the sensor value after the second sensor threshold has

been reached.

Figure 4: Sensor responses after bicycle impact.

As the ﬁgure (Fig. 4) shows, the sensor is in active

state (threshold is reached) approximately 100 ms or

even more, meaning that to ensure correct systems op-

eration, next value should be read after 100 ms. But

this is not a reliable solution, as in some cases the

value is more than 100 ms above threshold. Also ad-

justing the threshold (increasing its value) is not the

best solution either, as each sensor registers different

values (see Fig. 4).

4 CONCLUSIONS

During this study a bicycle counting system’s proto-

type based on pressure sensors and rubber tubes was

developed and tested, showing promising results (all

bicycle riding cases were counted correctly in a sin-

gle cyclist’s case). Although there are several limita-

tions and cases that causes the system to work poorly.

More than one cyclist counting simultaneously could

be highlighted as the main limitation.

Using digital signal sensors (for example, piezo-

electric swithces) instead of analogue (pressure sen-

sor in our case) should perform better, as there are no

thresholds involved which needs to be adjusted.

Additional tests should be performed on different

type (structure, hardness) of walkway surfaces (for

example, gravel), because there is a possibility, that

on softer surfaces the rubber tubes could be pressed

inside the surface.

Described prototype during testing phase included

Wi-Fi module for data transmission, but usually the

walkways are not close to any internet access points.

Thus, the module can be replaced by a suitable

GPRS module to access mobile internet, for exam-

ple, SIM900. Other cases may include SD card mod-

ule and closed range data transfer technologies, to

store collected data locally. Also the use of an low-

power wide-area network (LPWAN) can be consid-

ered for data transfer, for example, implementation

of LoRaWAN. Although it could be somewhat com-

plex task if the LoRa network should be built from

scratch. Different scenario - if the city already has

a developed LoRaWAN network. Then it is possible

to collaborate with the according telecommunication

company. Jelgava city is one of the cities in Latvia

that has a LoRaWAN network (built by telecommu-

nication company Lattelecom Group) and a collabo-

ration with Lattelecom has already been successful as

described in (Zacepins et al., 2018).

In order to improve (resolve some of the limita-

tions) some additional sensors could be considered to

add to the system, for example, ultrasonic sensors in

order to distinguish patterns between pedestrians and

cyclists or to identify cyclist riding direction in a sin-

gle tube case.

ACKNOWLEDGEMENTS

Scientiﬁc research, publication, and presentation are

supported by the ERANet-LAC Project ”Enabling re-

silient urban transportation systems in smart cities”

(RETRACT, ELAC2015/T10-0761).

REFERENCES

Alexander, G. (2018). Are bike lanes worth the

cost? https://earth911.com/business-policy/are-bike-

lanes-worth-the-cost/. Accessed: 2018-11-01.

Bao, J., He, T., Ruan, S., Li, Y., and Zheng, Y. (2017). Plan-

ning bike lanes based on sharing-bikes’ trajectories. In

Proceedings of the 23rd ACM SIGKDD international

conference on knowledge discovery and data mining,

pages 1377–1386. ACM.

Bickerstaff, K. and Walker, G. (2001). Public understand-

ings of air pollution: the ‘localisation’of environmen-

tal risk. Global Environmental Change, 11(2):133–

145.

Bushell, M. A., Poole, B. W., Zegeer, C. V., and Rodriguez,

D. A. (2013). Costs for pedestrian and bicyclist infras-

tructure improvements. University of North Carolina

Highway Safety Research Center, University of North

Carolina, Chapel Hill, 45.

Elsom, D. (2014). Smog alert: managing urban air quality.

Routledge.

Evans, M. R., Oliver, D., Shekhar, S., and Harvey, F. (2012).

Summarizing trajectories into k-primary corridors: a

VEHITS 2019 - 5th International Conference on Vehicle Technology and Intelligent Transport Systems

424

summary of results. In Proceedings of the 20th Inter-

national Conference on Advances in Geographic In-

formation Systems, pages 454–457. ACM.

Evans, M. R., Oliver, D., Shekhar, S., and Harvey, F. (2013).

Fast and exact network trajectory similarity computa-

tion: a case-study on bicycle corridor planning. In

Proceedings of the 2nd ACM SIGKDD international

workshop on urban computing, page 9. ACM.

Goodwill, R. (2015). Transport minister encourages

people to get on their bike for cycle to work day.

https://www.gov.uk/government/news/transport-

minister-encourages-people-to-get-on-their-bike-for-

cycle-to-work-day. Accessed: 2018-11-01.

Hyodo, T., Suzuki, N., and Takahashi, K. (2000). Mod-

eling of bicycle route and destination choice behav-

ior for bicycle road network plan. Transportation

Research Record: Journal of the Transportation Re-

search Board, (1705):70–76.

Jiang, Z., Evans, M., Oliver, D., and Shekhar, S. (2016).

Identifying k primary corridors from urban bicycle

gps trajectories on a road network. Information Sys-

tems, 57:142–159.

Kitthamkesorn, S. and Chen, A. (2017). Alternate weibit-

based model for assessing green transport systems

with combined mode and route travel choices. Trans-

portation Research Part B: Methodological, 103:291–

310.

Komasilovs, V., Zacepins, A., Kviesis, A., Pe

na, E., Tejada-

Estay, F., and Estevez, C. (2018). Trafﬁc monitoring

system development in jelgava city, latvia. In VEHITS,

pages 659–665.

Nordback, K., Kothuri, S., Phillips, T., Gorecki, C., and

Figliozzi, M. (2016). Accuracy of bicycle count-

ing with pneumatic tubes in oregon. Transportation

Research Record: Journal of the Transportation Re-

search Board, (2593):8–17.

Proulx, F. R., Schneider, R. J., and Miranda-Moreno, L. F.

(2016). Performance evaluation and correction func-

tions for automated pedestrian and bicycle counting

technologies. Journal of transportation engineering,

142(3):04016002.

Rybarczyk, G. and Wu, C. (2010). Bicycle facility planning

using gis and multi-criteria decision analysis. Applied

Geography, 30(2):282–293.

Ryus, P., Ferguson, E., Laustsen, K. M., Prouix, F. R.,

Schneider, R. J., Hull, T., and Miranda-Moreno, L.

(2014). Methods and technologies for pedestrian and

bicycle volume data collection. Citeseer.

Sustrans (2014). Handbook for cycle-friendly design.

Weigand, L., McNeil, N., and Dill, J. (2013). Cost analysis

of bicycle facilities: Cases from cities in the portland,

or region.

Zacepins, A., Jelinskis, J., Kviesis, A., Dzenis, M., Ko-

masilovs, V., and Komasilova, O. (2018). Applica-

tion of lorawan technology in precision beekeeping.

In Agrosym 2018 Book of proceedings, pages 1759–

1765.

Zangenehpour, S., Romancyshyn, T., Miranda-Moreno,

L. F., and Saunier, N. (2015). Video-based automatic

counting for short-term bicycle data collection in a va-

riety of environments. Technical report.

Automated Bicycle Counting System’s Prototype to Evaluate the Necessity of New Bicycle Lanes in Jelgava City

425