Healthy Route Generation and Recommendation

LAZAR PENDOV1, ZHANLIN JI2,3, IVAN GANCHEV1,3,4,*

1Department of Computer Systems,

University of Plovdiv “Paisii Hilendarski”,

236 Bulgaria Blvd., Plovdiv 4027,

BULGARIA

2College of Mathematics and Computer Science,

Zhejiang A&F University,

Hangzhou, Zhejiang 311300,

CHINA

3Telecommunications Research Center (TRC),

University of Limerick,

Limerick V94 T9PX,

IRELAND

4Institute of Mathematics and Informatics,

Bulgarian Academy of Sciences (IMI–BAS),

Sofia 1040,

BULGARIA

*Corresponding Author

Abstract: - This paper presents the utilization of a developed pilot wireless-based Air Quality Index (AQI)

monitoring system, reporting live geo-grid resolved air quality data, for the purposes of healthy route

generation and recommendation to users. The generated routes are visualized on a map and recommended to

users through a specially developed web-based application, as part of the client tier of the supporting IoT

platform EMULSION. A distributed computing architecture is utilized for the generation of healthy (more

precisely, ‘least air pollution exposure’) routes, performed in near real-time using the dynamic Dijkstra

algorithm, based on the interpolated AQI values. In addition, the fastest and shortest routes for each journey,

requested by a user, are generated as well. The importance of the presented work lies within the practical

applicability of the proposed method for healthy route generation, either as a stand-alone version of the

software application developed for the purpose or integrated into the existing popular navigation systems and

applications alike.

Key-Words: - Internet of Things (IoT); IoT platform; EMULSION; Air Quality Index (AQI); healthy route;

route generation; recommendation to users.

5HFHLYHG0DUFK5HYLVHG6HSWHPEHU$FFHSWHG1RYHPEHU3XEOLVKHG'HFHPEHU

1 Introduction

The existing web-based route planners and real-time

navigation systems, accessed by fixed or mobile

personal devices, allow planning and adjusting

journeys by providing comfort and a sense of safety

to users, [1]. By supplying dynamic and integrated

technological support tools to users, along with

interactive planning and navigation features for

various travel modes, these systems mainly find the

shortest, fastest, or cheapest traveling routes.

However, people (especially urban residents) have

begun to pay more attention to their quality of life

(QoL), by considering environmental factors

affecting their health, such as air quality, and this

has become their new focus when traveling, [2].

Travel schemes with relatively low pollutant

exposure can not only improve human health but

can also benefit social stability and sustained

progress. In contrast, path-based long-distance

WSEAS TRANSACTIONS on INFORMATION SCIENCE and APPLICATIONS

DOI: 10.37394/23209.2024.21.52

Lazar Pendov, Zhanlin Ji, Ivan Ganchev

E-ISSN: 2224-3402

558

Volume 21, 2024

outdoor activities, surrounded by poor air quality,

have a negative effect on human health, especially

when cycling, running, jogging, or walking, [3].

Therefore, an alternative approach should be taken

in route planning, including health-related

optimization criteria, based on the concept of

“healthy route” (a.k.a. “green route”, “clean route”)

and related concepts of “safe route” and

“sustainable route”, [1].

A common indicator, used in many countries to

measure air pollution, is the Air Quality Index

(AQI), which was developed by the United States

Environmental Protection Agency (US EPA), based

on the following six major pollutants: ground-level

ozone (O3), particulate matter (PM2.5 and PM10),

carbon monoxide (CO), sulfur dioxide (SO2), and

nitrogen dioxide (NO2). The US EPA AQI values

run from 0 to 500, divided into six levels of concern,

as shown in Table 1, [4]. The higher the AQI value,

the greater the level of air pollution and the greater

the health concern. The green level (AQI ≤ 50)

represents satisfactory air quality, whereby air

pollution poses little or no risk to humans. At the

yellow level (51 ≤ AQI ≤ 100), the air quality is

acceptable, but there may be a risk for some people,

particularly those who are unusually sensitive to air

pollution. At the orange level (101 ≤ AQI ≤ 150),

sensitive groups (people with lung diseases, older

people, and children) may experience air-quality-

related health problems, whereas the general public

is less likely to be affected. At the red level (151 ≤

AQI ≤ 200), part of the general public may

experience health effects, whereas sensitive groups

may experience more serious health effects. The

purple level (201 ≤ AQI ≤ 300), signifies a health

alert, whereby the risk of health effects is increased

for everyone, and most people may experience

increasingly severe adverse health effects. The

maroon level (AQI ≥ 301) represents hazardous air

quality and serves as a health warning of emergency

conditions, so everyone is more likely to be

affected.

The health-driven measurement of air quality

with reasonable geo-grid resolution is in a growing

demand across the world. As the current geo-grid

resolved AQI is informative of environmental

conditions related to personal health, in many

countries, especially in urban areas, there are people

needing to be able to daily check the current AQI

value before going to work or doing outdoor

activities. This can be integrated into mobile apps

and such. To enable millions of simultaneous AQI

requests, a server-side AQI monitoring and

publishing system is required, operating with high

throughput and high availability. It is easy to

visualize such a system, established on a

corresponding Internet of Things (IoT) platform,

e.g., functioning as an integral part of a smart city.

Given the sizeable deployment and maintenance

expense involved, the general public service thrust

is to build a low-resolution AQI geo-grid network,

e.g., with AQI monitoring points a kilometer apart.

Figure 1 shows a sample low-resolution AQI

network, established in Hebei Province (China) with

just 62 AQI monitoring stations. The Chinese

government’s national requirement would be for

around 10,735 AQI monitoring stations deployed in

such an area of 10,735.78 km². Hence there is a

good motivation and market for the development of

low-cost sensor-based AQI monitoring stations.

Table 1. The AQI levels of concern

AQI range

Level of concern

Color

0–50

good

green

51–100

moderate

yellow

101–150

unhealthy for sensitive

groups

orange

151–200

unhealthy

red

201–300

very unhealthy

purple

301–500

hazardous

maroon

Fig. 1: The utilized low-resolution AQI network

The AQI monitoring stations in the established

network function as part of the developed pilot AQI

monitoring system, operating on top of the IoT

platform EMULSION [5], which was successfully

implemented and tested. EMULSION is a horizontal

IoT platform of a combined type (hardware and

software), built with low-cost electronics and open-

source software, and consisting of seven tiers. In the

sensor & actuator tier, different types of sensors,

environment monitoring stations, location trackers,

etc., operate to capture the changes occurring in the

WSEAS TRANSACTIONS on INFORMATION SCIENCE and APPLICATIONS

DOI: 10.37394/23209.2024.21.52

Lazar Pendov, Zhanlin Ji, Ivan Ganchev

E-ISSN: 2224-3402

559

Volume 21, 2024

physical world and send the corresponding

information to the cloud tier through data/remote

transfer units (D/RTUs) and smart communication

gateways, through different wireless access

networks. After analyzing the data sent by the

sensor & actuator tier, the cloud tier makes

appropriate decisions, generates suitable

recommendations for users, and sends the necessary

configuration information and/or commands to the

controllers, actuators, guards, etc., located in the

sensor & actuator tier, for enforcing the required

actions needed for the realization of the imposed

changes in the physical world.

Each AQI monitoring station includes a variety

of pollution sensors connected with an ultra-low-

power geo-grid identified D/RTU. Sensors

communicate periodically (usually every 5 minutes)

with the corresponding D/RTU. Each D/RTU sends

the collected air-quality data, according to the

scheduling algorithm used, to the cloud tier of

EMULSION, which employs a distributed Redis

database, a Hadoop cluster [6], and a web-based

Geographic Information System (GIS) [7]. The

cloud tier provides scaled GUI services for the

mobile and desktop client applications’ requests.

This paper demonstrates the use of the developed

pilot AQI monitoring system for the generation of

healthy routes for outdoor activity planning by

users. The difference between the presented work

with those published in the literature is that most of

the published research suggests a specific

application that does some specific things. In

contrast, we offer here a more generic architecture

that is technology-independent (i.e., w.r.t. hardware,

operating system, programming language) and

distributed in nature (i.e., hosted on multiple

machines), which provides an opportunity for

extending it almost without a limit. A clear

explanation of the healthy route calculation is

provided in the paper, along with a concrete

implementation in the form of a web-based

application, demonstrating the way it works in

reality. The usefulness of the presented study relates

to the possible integration of the proposed method

for healthy route generation into the existing

navigation systems and applications alike.

2 Related Work

In general, there are two primary types of methods

for route cost calculation, [3], i.e., using: (1) static

cost of paths serving as an input to the standard

Dijkstra algorithm [8]; and (2) dynamic cost of

paths, varying over space and time, such as the AQI

value and the travel time which depends on the path

infrastructure’s condition and traffic flow. As more

advanced, multiple methods of the second type have

been proposed, e.g., based on adaptive decision

rules [9], genetic algorithms [10], [11], [12],

probabilistic models [13], uncertainty [14], etc.

Previous studies on “healthy route” generation

can be divided into two main groups [2], i.e., using:

(1) monitoring stations to measure pollutant

exposure on various types of roads, followed by

classification of roads as healthy or unhealthy, based

on the exposure levels; and (2) pollution distribution

data obtained by different means, e.g., by a land use

regression (LUR) [15], [16], an operational street

pollution model (OSPM) [17], an interpolation

method, [3], [18], etc., as an input to the (dynamic)

Dijkstra algorithm for generating healthy routes

[19], using different indicators (traffic volume, AQI,

potential pollutant dose taken, etc.) as road network

weights. If taking full advantage of modern

pollutant retrieval technologies, the trustworthiness

of the generated healthy routes could be

significantly increased. For this, [2] proposes a

short-distance healthy route planning approach,

utilizing fine spatial resolution images, and

meteorological and socioeconomic data to retrieve

the spatial distribution of PM2.5 concentration in

hourly intervals via a back-propagation neural

network. The effectiveness of the approach is

verified by comparing the PM2.5 potential dose

reduction rate between the generated healthy route

and the shortest route, reaching up to 20% reduction

in some cases. As an important factor affecting the

AQI values, PM2.5 concentration can be used also

to predict the AQI, [3].

By utilizing an interpolation method from the

second group, the current paper is focused on the

‘least air pollution exposure’ aspect of the “healthy

route”, computed in near real-time by applying the

dynamic Dijkstra algorithm, whereby the potential

exposure rate is calculated based on the AQI values,

whereby the desired values of the main air

pollutants (PM2.5, PM10, O3, CO, SO2, NO2) serve

as upper boundaries for reducing the number of

possible routes.

3 Utilized Data

3.1 Permanent Data

This type of data rarely or never changes. The open

data of the OpenStreetMap [20] are used as map

data for the study area, shown in Figure 1. This is a

rectangular area enclosed between the GPS

coordinates (36.846000/114.345000) and

(37.655772/115.684369) with a total coverage of

WSEAS TRANSACTIONS on INFORMATION SCIENCE and APPLICATIONS

DOI: 10.37394/23209.2024.21.52

Lazar Pendov, Zhanlin Ji, Ivan Ganchev

E-ISSN: 2224-3402

560

Volume 21, 2024

10735.78 km², which includes most of Hebei

Province (China). We initially focused on the roads.

In the data provided by the OpenStreetMap, the

roads have their own type (motorway, primary road,

etc.) and are described by an ordered list of nodes.

The number of nodes in the study area is equal to

246,686. Each node has its own unique number and

is described with GPS coordinates. Distances

between nodes, and between nodes and the AQI

monitoring stations deployed in the area, can be

calculated, based on their GPS coordinates, using

the haversine formula, as follows:

where  and  denote the latitudes of the two

points (in radians),  denotes the difference

between the latitudes of the two points,  denotes

the difference between the longitudes of the two

points (in radians), and  denotes the radius of the

Earth.

The minimum, maximum, mean, and mode

values of the distances between any two

OpenStreetMap nodes in the study area are

presented in Table 2.

Table 2. Distances between OpenStreetMap nodes

in the study area

Distances (m)

Min

value

Max

value

Mean

value

Mode

value

0.02

2633.79

96.14

11.00

We converted the OpenStreetMap node data into

a graph, suitable for applying the Dijkstra algorithm.

Each edge in the graph represents a connection

between two nodes along a particular road existing

in the area. The indicators used are the edge length

and road type.

3.2 Refreshable Data

These data come from the AQI monitoring system,

operating on top of the IoT platform EMULSION,

which provides data on the main air pollutants

(PM2.5, PM10, O3, CO, SO2, NO2) and the AQI.

There are 62 AQI monitoring stations in the study

area (c.f., Figure 1). These data are updated on

every hour.

3.3 Computable Data

To implement the Dijkstra algorithm, each

OpenStreetMap node is assigned with a particular

air quality value , obtained by interpolation,

using the air quality values of the three nearest AQI

monitoring stations, as follows:

where , , and  denote the air

quality values of the three nearest AQI monitoring

stations, and , , and  denote the distance

from the OpenStreetMap node to each of these three

stations, respectively.

Table 3 presents statistics of the minimum,

maximum, mean, and mode values of the distances

between the OpenStreetMap nodes in the study area

and the corresponding three nearest AQI monitoring

stations (Station 1 is the nearest, and Station 3 is the

farthest).

Table 3. Distances between OpenStreetMap nodes

and the three nearest AQI monitoring stations

in the study area

Distances (m)

Min

value

Max

value

Mean

value

Mode

value

Station 1

44251

8585

6847

Station 2

507

45406

10269

8756

Station 3

939

47254

11784

10448

As can be seen from Table 3, there are

worryingly large distances between some

OpenStreetMap nodes and the nearest AQI

monitoring stations, but these gaps can be

compensated by deploying more stations in the area

in the future.

4 Distributed Computing Architecture

The elaborated computing architecture is divided

into four subsystems, as presented in Figure 2. Each

subsystem is platform- and program-independent

and can be implemented using different

programming languages, operating systems, and

hardware. Communication between subsystems

adheres to open standard protocols. A major

advantage of this architecture is the use of multiple

computing servers, which makes it easily scalable

depending on the expected number of users and

workload.

WSEAS TRANSACTIONS on INFORMATION SCIENCE and APPLICATIONS

DOI: 10.37394/23209.2024.21.52

Lazar Pendov, Zhanlin Ji, Ivan Ganchev

E-ISSN: 2224-3402

561

Volume 21, 2024

4.1 Data Collection Subsystem

This subsystem consists of multiple AQI monitoring

stations and an information server to which these

stations periodically send their collected data. This

subsystem is apart from the other three subsystems

and is managed separately. The information server

provides the aggregated collected data through an

API in a JSON format.

Fig. 2: The operation of the developed distributed

architecture, utilized for route computation

4.2 Data Update Subsystem

This subsystem consists of an update script and a

database server. Periodically, the script is activated,

making a request to receive updates from the

information server of the first subsystem. The

received data are processed in an appropriate form

and sent to the database server for caching. The

update script then sends an update notification to

each computing server, operating as part of the third

subsystem, described next.

4.3 Computing Subsystem

When a computing server receives an update

notification, it makes a request to the caching

database server to receive the necessary new data,

associated with its activity. Once the data are

received, these are arranged in a form optimized for

fast reading, preferably by caching in RAM.

4.4 Route Generation Subsystem

Through a corresponding client application, the user

makes a request to the API of the proxy server to

calculate a route (of a particular type) between two

points on the map, according to her/his

requirements. The proxy server selects one of the

computing servers suitable for the request and

forwards the request to it. After performing the

required calculations, the computing server returns

the result to the proxy server, which in turn forwards

it to the client application, which renders the result

to the user.

The most recent (updated) data are used for

route calculations. When the time to pass a route

exceeds the updating time of data coming from the

AQI monitoring stations, new requests can be made

during the journey in order to regenerate the route

with greater truthfulness. It is also possible to apply

predictive algorithms based on historical data (such

algorithms have not been implemented yet).

5 Implementation

The computing architecture proposed in the

previous section is independent of the programming

language, operating system, and hardware, but in

order to have some implementation, concrete

technologies must be chosen. In the solution

presented here, the Linux Debian (trixie) operating

system is used for all servers (except for the first

subsystem) and PHP is used as a programming

language (for console and web execution).

5.1 Data Collection Subsystem

This subsystem was developed before starting the

work presented here; thus, it is not discussed further.

5.1.1 Data Update Subsystem

The update script is executed at a certain time

interval (currently, each hour) via cron. The script

makes an HTTP GET request to the information

server of the first subsystem and receives a response

in JSON format. The received data are processed

and sent to the MySQL server for caching.

Notification is then made via parallel GET requests

to the computing servers. Security measures are

taken by applying a unique key for each computing

server so that this call cannot be made by an

unauthorized participant.

5.1.2 Computing Subsystem

A major problem of the current implementation

relates to the natural way PHP scripts work. When a

PHP script is called, a new instance of it is created,

the required resources are allocated, the necessary

operations are performed, and a response is

returned, followed by the release of all allocated

resources. This continuous allocating and releasing

of resources for each call, involving huge data

structures, is extremely inefficient and cumbersome.

To solve this problem, the data are cached in two

stages:

WSEAS TRANSACTIONS on INFORMATION SCIENCE and APPLICATIONS

DOI: 10.37394/23209.2024.21.52

Lazar Pendov, Zhanlin Ji, Ivan Ganchev

E-ISSN: 2224-3402

562

Volume 21, 2024

1) Stage 1: Structuring the data in a jagged

array format in a PHP file, suitable for

importing via include() into the computation

script;

2) Stage 2: Transferring the obtained data

structure to RAM via the APCu extension

for Apache. Thus, once the structure is

allocated in the Apache server's RAM, it is

accessible by all instances of the computing

script without the need for multiple resource

allocations and releases.

When the Apache server is restarted, the cache in

RAM is lost, but the structure located in a file

(created in the first stage) is usable and allows it to

be cached in RAM again.

5.1.3 Route Generation Subsystem

This subsystem includes a client application, an API

proxy server, and multiple computing servers,

operating as follows:

 The developed client application is accessible

through the official website of the

EMULSION project [21], which is hosted on

an Odroid-HC2 web server with a 32-bit

ARM CPU: Samsung Exynos5422 ARM®

Cortex™-A15 Quad 2.0GHz/Cortex™-A7

Quad 1.4GHz. The web application sends

requests to the API proxy using the HTTP

GET commands and receives responses in a

JSON format.

 The API proxy is hosted on the same web

server and is implemented in PHP. When

requesting a route computation, it chooses a

random computation server, by taking into

account the server capabilities (more

powerful servers are called more often). The

received response is forwarded to the client

application. Necessary security measures are

taken by applying a unique secret key for

each computing server to prevent

unauthorized requests.

 The computing servers include:

 A PC with 2xCPU: Intel(R) Xeon(R)

Gold 6134 CPU @ 3.20GHz (cores: 16,

threads: 32). Only ¼ of its computing

capabilities are reserved for the route

computation task (threads: 8), whereas

the remaining capabilities are used for

performing other activities, as a

demonstration of the ability to use a

shared server.

 A PC with Intel(R) Xeon(R) CPU E3-

1220 v3 @ 3.10GHz (cores/threads: 4).

 15 single-board computers Odroid-N2+

with 64-bit ARM CPU: Amlogic

S922X, Quad Cortex-A73 2.4GHz and

Dual Cortex-A53 2GHz (cores/threads:

4+2).

When the compute script is called, it looks for

an available RAM cache. If such a cache is not

available, the data are loaded from the cached file

(Stage 1) and a new RAM cache is created.

During the trial experiments conducted with

single-board servers Odroid MC1 (with 4 servers

per unit) with 32-bit CPU (per server): Samsung

Exynos5422 ARM® Cortex™-A15 Quad

2.0GHz/Cortex™-A7 Quad 1.4GHz (cores/threads

per server: 8), the route computations often

exceeded two minutes. Therefore, this type of

single-board computer was excluded from the final

version of this subsystem. However, with another

software implementation of the computing process,

it would probably be possible to utilize such

inexpensive server hardware as well.

The used ARM computing servers (without the

two non-ARM servers) are depicted in Figure 3.

Fig. 3: The utilized Odroid single-board computers

(15x Odroid-N2+ and 5x Odroid-MC1)

6 Client Application

6.1 Setting up Maximum Values of Air

Quality Parameters

The user can specify a preferred maximum value for

each specific air pollutant (PM2.5, PM10, O3, CO,

SO2, NO2) and for the AQI, which must be not

exceeded in any case along the route.

6.2 Travel Modes

Different travel modes could be used by users, each

with its own exposure to polluted air and speed of

movement, which results in different inhaled doses.

In [22], [23], the rate of exposure to a particular air

pollutant on a route is calculated as follows:

       (3)

WSEAS TRANSACTIONS on INFORMATION SCIENCE and APPLICATIONS

DOI: 10.37394/23209.2024.21.52

Lazar Pendov, Zhanlin Ji, Ivan Ganchev

E-ISSN: 2224-3402

563

Volume 21, 2024

where  (L/min) denotes the minute ventilatory

rate (MVR),  (μg/m3) denotes the pollutant

concentration, and  (min) denotes the travel

time on the route.

Based on (3), we calculate the rate of exposure

to polluted air on a route, based on the AQI value,

as follows:

   󰇛 

󰇜 (4)

where denotes the AQI value of the th section

between two consecutive/neighboring

OpenStreetMap nodes on the route (calculated as

the average value of the interpolated AQI values of

the two nodes), denotes the travel time

through that section (calculated by dividing the

section length by the average travel speed on that

section), and  denotes the corresponding

coefficient of the travel mode, calculated by

comparing the MVRs for different travel modes,

using their average values reported by ChatGPT.

Coefficient  takes values between 0 and 1,

whereby a value of 1 means maximum exposure to

air pollution, and a value closer to 0 implies

minimum exposure. By comparing the MVRs of the

considered travel modes, shown in Table 4, it can be

seen that cycling leads to maximum exposure, so its

coefficient  is set to 1. The values of

other travel modes are calculated by dividing their

MVRs to the MVR of the cycling mode. More

accurate values of coefficient  can be obtained

with more precise empirical studies investigated and

taken into account in the future. If a user decides to

wear a protection mask when traveling, this could

be easily reflected in (4) by applying a

corresponding protection coefficient  .

In each travel mode, for each road type, there is

a default value for the average speed, which can be

adjusted in the application itself depending on the

legal regulations for road traffic (e.g., according to

maximum speed limits set), and can be further

adjusted by the user depending on her/his abilities

and habits related to the specific travel mode used.

Some travel modes may be prohibited on certain

road types (e.g., a car cannot be used within a

pedestrian zone).

6.3 Routes’ Start and End Points

In the current implementation, the selection of the

start and end points of a route is done by the user by

clicking on the map with the left (for the start point)

and right (for the end point) mouse buttons. The

specified coordinates will hardly coincide exactly

with the coordinates of any OpenStreetMap node in

the dataset. Therefore, the closest node existing on

the map is used for the route generation.

Table 4. Different travel modes and their

corresponding minute ventilatory rates (MVRs)

and coefficients

Travel

mode

MVR

(L/min)



Driving a car

(with closed windows)

0.200

Driving a car

(with open windows)

0.229

Motorcycling

0.286

Cycling

1.000

Walking

0.429

6.4 Route Types

In addition to the ‘least air pollution exposure’

route, the developed web-based application can

generate and recommend also the fastest route and

shortest route for traveling. All route computations

are performed by means of the Dijkstra algorithm,

using, respectively, the ‘exposure rate’, defined in

(4), as a cost for the ‘least air pollution exposure’

route, the ‘time’ for the fastest route, and the

‘length’ for the shortest route.

6.5 Results

The final result for each generated route type

includes a route visualization, presented as a set of

ordered GPS coordinates used for route

visualization on the map, as shown in Figure 4(a)

(each route type is drawn in a different color), and

information about air quality parameter values on

each route type, as presented in Figure 4(b). More

specifically, the developed application generates

three different types of routes for traveling between

any two points in the study area. In the example

presented in Figure 4, these are (i) the healthiest

route with a minimum exposure rate of 1120.45

AQI.minutes, (ii) the fastest route with a minimum

travel duration of 1:29 h, and (iii) the shortest route

with a minimum travel distance of 133.069 km.

Additional data include the GPS coordinates of the

start and end points (respectively,

(37.53357480/114.51482110) and

(36.99950980/115.51223270) in this example), the

exposure rate, and the name of the server used for

the computation of each route type.

WSEAS TRANSACTIONS on INFORMATION SCIENCE and APPLICATIONS

DOI: 10.37394/23209.2024.21.52

Lazar Pendov, Zhanlin Ji, Ivan Ganchev

E-ISSN: 2224-3402

564

Volume 21, 2024

(a)

(b)

Fig. 4: Sample route generation: (a) different route

types (i.e., healthiest, fastest, and shortest route)

between two points in the study area; (b) the air

quality parameter values on the healthiest route,

generated in (a)

7 Conclusion

This paper has presented an elaborated distributed

computing architecture for the generation of the

healthiest routes (more precisely, the ‘least air

pollution exposure’ routes), performed in near real-

time by means of the dynamic Dijkstra algorithm,

based on the air quality index (AQI). In addition, the

fastest and shortest routes are generated as well. The

generated routes are visualized on a map and

recommended to users through a specially

developed web-based application, as part of the

client tier of the supporting IoT platform

EMULSION. The generated healthiest routes, in

particular, allow the users to avoid air-polluted areas

posing particular health risks to them.

Future work will be focused on the development

of suitable models to predict the future AQI, based

on historical data, current meteorological data, and

weather forecasts, for the purposes of smart

proactive “healthy routes” planning for outdoor

activities of users. The routes will be initially

preplanned, with the possibility to be dynamically

changed later, if needed, depending on the current

environmental conditions. The incorporation of such

health-related criteria into existing navigation

systems and applications for route generation and

recommendation is envisaged as an important

functionality extension of the latter, [5].

References:

[1] Ribeiro, P. and J.F.G. Mendes, Healthy routes

for active modes in school journeys.

International Journal of Sustainable

Development and Planning, 2013. 8(4): pp.

591-602. DOI: 10.2495/SDP-V8-N4-591-602.

[2] Gao, L.-N., F. Tao, P.-L. Ma, C.-Y. Wang, W.

Kong, W.-K. Chen, and T. Zhou, A short-

distance healthy route planning approach.

Journal of Transport & Health, 2022. 24: pp.

1-14. DOI: 10.1016/j.jth.2021.101314.

[3] Zou, Z., T. Cai, and K. Cao, An urban big

data-based air quality index prediction: A case

study of routes planning for outdoor activities

in Beijing. Environment and Planning B:

Urban Analytics and City Science, 2019.

47(6): pp. 948-963. DOI:

10.1177/2399808319862292.

[4] Air Quality Index (AQI) Basics, [Online].

https://www.airnow.gov/aqi/aqi-basics/

(Accessed Date: September 20, 2024).

[5] Ganchev, I., Z. Ji, and M. O'Droma,

Horizontal IoT Platform EMULSION.

Electronics, 2023. 12(8): pp. 1-21. DOI:

10.3390/electronics12081864.

[6] Shafer, J., S. Rixner, and A.L. Cox. The

Hadoop distributed filesystem: Balancing

portability and performance. Proc. of 2010

IEEE International Symposium on

Performance Analysis of Systems & Software

(ISPASS), pp. 122-133. White Plains, NY,

USA, 2010. DOI:

10.1109/ISPASS.2010.5452045.

[7] Goodchild, M.F., Geographic information

system, in: Liu, L., Özsu, M.T. (eds)

Encyclopedia of Database Systems. Springer,

Boston, MA, 2009. pp. 1231-1236. DOI:

10.1007/978-0-387-39940-9_178.

[8] Dijkstra, E.W., A note on two problems in

connexion with graphs. Numerische

Mathematik, 1959. 1(1): pp. 269-271. DOI:

10.1007/BF01386390.

WSEAS TRANSACTIONS on INFORMATION SCIENCE and APPLICATIONS

DOI: 10.37394/23209.2024.21.52

Lazar Pendov, Zhanlin Ji, Ivan Ganchev

E-ISSN: 2224-3402

565

Volume 21, 2024

[9] Hall, R.W., The Fastest Path through a

Network with Random Time-Dependent

Travel-Times. Transportation Science, 1986.

20(3): pp. 182-188.

[10] Pellazar, M.B. Vehicle route planning with

constraints using genetic algorithms. Proc. of

National Aerospace and Electronics

Conference (NAECON'94), vol. 1, pp. 111-

118. Dayton, OH, USA, 1994. DOI:

10.1109/naecon.1994.333010.

[11] Pattnaik, S.B., S. Mohan, and V.M. Tom,

Urban bus transit route network design using

genetic algorithm. Journal of Transportation

Engineering, 1998. 124(4): pp. 368-375. DOI:

10.1061/(ASCE)0733-947X(1998)124:4(368).

[12] Chien, S., Z.W. Yan, and E. Hou, Genetic

algorithm approach for transit route planning

and design. Journal of Transportation

Engineering, 2001. 127(3): pp. 200-207. DOI:

10.1061/(ASCE)0733-947X(2001)127:3(200).

[13] Boyan, J. and M. Mitzenmacher, Improved

results for route planning in stochastic

transportation. Proc. of the 12th Annual ACM-

SIAM Symposium on Discrete Algorithms, pp.

895-902. Washington, D.C., USA, 2001.

DOI: 10.1145/365411.365803.

[14] Nikolova, E., M. Brand, and D.R. Karger,

Optimal route planning under uncertainty.

Proc. of the 16th International Conference on

Automated Planning and Scheduling, pp. 131-

140. Cumbria, UK, 2006.

[15] Hatzopoulou, M., S. Weichenthal, G. Barreau,

M. Goldberg, W. Farrell, D. Crouse, and N.

Ross, A web-based route planning tool to

reduce cyclists' exposures to traffic pollution:

a case study in Montreal, Canada.

Environmental Research, 2013. 123: pp. 58-

61. DOI: 10.1016/j.envres.2013.03.004.

[16] Zou, B., S. Li, Z. Zheng, B.F. Zhan, Z. Yang,

and N. Wan, Healthier routes planning: A new

method and online implementation for

minimizing air pollution exposure risk.

Computers, Environment and Urban Systems,

2020. 80: pp. 1-11.

DOI: 10.1016/j.compenvurbsys.2019.101456.

[17] Hertel, O., M. Hvidberg, M. Ketzel, L. Storm,

and L. Stausgaard, A proper choice of route

significantly reduces air pollution exposure - a

study on bicycle and bus trips in urban streets.

Science of The Total Environment, 2008.

389(1): pp. 58-70.

DOI: 10.1016/j.scitotenv.2007.08.058.

[18] Zahmatkesh, H., M. Saber, and M.

Malekpour, A New Method for Urban Travel

Rout Planning Based on Air Pollution Sensor

Data. Current World Environment, 2015.

10(Special-Issue1): pp. 699-704.

DOI: 10.12944/CWE.10.Special-Issue1.83.

[19] Mahajan, S., Y.-S. Tang, D.-Y. Wu, T.-C.

Tsai and L.-J. Chen, CAR: The Clean Air

Routing Algorithm for Path Navigation With

Minimal PM2.5 Exposure on the Move. IEEE

Access, 2019. 7: pp. 147373-147382.

DOI: 10.1109/ACCESS.2019.2946419.

[20] Arsanjani, J.J., A. Zipf, P. Mooney, and M.

Helbich, OpenStreetMap in GIScience:

Experiences, Research, and Applications, in

Lecture Notes in Geoinformation and

Cartography. 2015, Springer International

Publishing: Imprint: Springer, Cham.

[21] EMULSION project, [Online].

http://emulsion.science (Accessed Date:

September 20, 2024).

[22] Li, H.-C., P.-T. Chiueh, S.-P, Liu, and Y.-Y.

Huang, Assessment of different route choice

on commuters' exposure to air pollution in

Taipei, Taiwan. Environment Science

Pollution Research, 2017. 24(3): pp. 3163-

3171. DOI: 10.1007/s11356-016-8000-7.

[23] Ramos, C.A., H.T. Wolterbeek, and S.M.

Almeida, Air pollutant exposure and inhaled

dose during urban commuting: a comparison

between cycling and motorized modes. Air

Quality Atmosphere and Health, 2016. 9(8):

pp. 867-879. DOI: 10.1007/s11869-015-0389-

WSEAS TRANSACTIONS on INFORMATION SCIENCE and APPLICATIONS

DOI: 10.37394/23209.2024.21.52

Lazar Pendov, Zhanlin Ji, Ivan Ganchev

E-ISSN: 2224-3402

566

Volume 21, 2024

Contribution of Individual Authors to the

Creation of a Scientific Article (Ghostwriting

Policy)

The authors equally contributed to the present

research, at all stages from the formulation of the

problem to the final findings and solution.

Sources of Funding for Research Presented in a

Scientific Article or Scientific Article Itself

This publication has emanated from joint research

conducted with the financial support of the

Bulgarian National Science Fund (BNSF) under the

Grant No. KP-06-IP-CHINA/1 (КП-06-ИП-

КИТАЙ/1) and the National Key Research and

Development Program of China under Grant No.

2017YFE0135700.

Conflict of Interest

The authors have no conflicts of interest to declare.

Creative Commons Attribution License 4.0

(Attribution 4.0 International, CC BY 4.0)

This article is published under the terms of the

Creative Commons Attribution License 4.0

https://creativecommons.org/licenses/by/4.0/deed.en

_US

WSEAS TRANSACTIONS on INFORMATION SCIENCE and APPLICATIONS

DOI: 10.37394/23209.2024.21.52

Lazar Pendov, Zhanlin Ji, Ivan Ganchev

E-ISSN: 2224-3402

567

Volume 21, 2024