See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/379259894

Real Time Smart Road Traffic Flow Management System in a Developing

Country

Conference Paper · March 2024

CITATION READS

1 1,165

2 authors, including:

Shadrack Mambo
Walter Sisulu University
16 PUBLICATIONS 42 CITATIONS

SEE PROFILE

All content following this page was uploaded by Shadrack Mambo on 25 March 2024.

The user has requested enhancement of the downloaded file.

 Cape Town, South Africa, 5 – 7 March 2024

Real Time Smart Road Traffic Flow Management System in a

Developing Country
Joanne Mahugua*, Dr Shadrack Mambob
a,b
School of Engineering and Architecture, Kenyatta University, P.O. Box 43844-00100, Nairobi Kenya
* Corresponding Author – Email: [email protected]

Abstract

The Intelligent Transport System (ITS) project is part of the Nairobi Urban Transport Improvement Programme and is being
promoted by Kenya Urban Roads Authority (KURA) which manages, develops, rehabilitates and maintains all public roads in
cities and municipalities. The system is able to study traffic flow data in real-time, thereby allowing longer traffic flow on roads
with most traffic, thus minimising unnecessary waiting time.
The goal of this collaboration project between Kenyatta University (KU) and Kenya Urban Roads Authority (KURA) is to develop
a smart traffic control system that utilizes machine learning to analyze traffic flow. The project aims to improve the efficiency and
safety of traffic management in urban roads. The smart traffic control system was able to predict traffic patterns, optimize traffic
signals, and provide real-time traffic information to drivers, thus reducing congestion and improving overall traffic flow.
The research designed and developed an artificial intelligence model based on deep learning technology that automatically learns
traffic patterns, reads vehicle number plates, identify traffic violations using CCTV infrastructure thus directing traffic flow,
combating traffic offences and reducing accidents.

© 2024 Elsevier Inc. All rights reserved.

Keywords: Intelligent Transport System; Network Optimization; Traffic Flow Modelling

1. Introduction

Traffic situation in urban cities like Nairobi is getting worse due to unsystematic development and high inflow of
population. Nairobi, the biggest city in East Africa and capital city of Kenya, is the political and economic center of
the country with a population of 4.397 million (2019 KNBS Census). Perennial traffic jams in Nairobi has in turn
influences the city’s urban function, accessibility to social and economic services and people’s livelihood. Traffic
situation has not been solved by additional lanes, closure of some roundabouts and construction of missing links.
The Intelligent Transport System (ITS) applies artificial intelligence to manage urban traffic. It uses smart cameras,
variable timing traffic lights and a control centre to monitor and control traffic – easing the burden on traffic police
officers. The smart traffic control system also comes with a mobile electronic police system that allows flexible
monitoring of general traffic conditions.
The ITS project is part of the Nairobi Urban Transport Improvement Programme and is being promoted by Kenya
Urban Roads Authority (KURA) which manages, develops, rehabilitates and maintains all public roads in cities and

Click here to enter text.

2 Eng. Joanne Mahugu / African Transport Research Conference 2024, Cape Town, South Africa

municipalities. The system is able to study traffic flow data in real-time, thereby allowing longer traffic flow on roads
with most traffic, thus minimising unnecessary waiting time.
This research will investigate, develop and implement artificial intelligence (AI) enabled systems (smart systems)
to automatically learn traffic patterns using CCTV infrastructure installed on selected highways in Nairobi. It will also
aid in predicting and controlling traffic flow, implementing reversible tidal flow lanes, detecting, recording and
deterring traffic regulation offences, carrying out traffic analysis to improve road safety and efficient road use.
Smart traffic management systems enable integrated optimisation of road networks to match infrastructure demand
and supply in near real-time. It also manages traffic speeds, frequencies, and prioritisation of vehicles while abiding
by government regulations and safety requirements. Real-time data can be used to suggest alternate routes to drivers
when routes are congested and indicate to public transport operators and decision makers where user demand and
supply is located. Intelligence Traffic Systems are applied in traffic management and interfaces with other modes of
transportation to relieve traffic congestion, improve safety, and reduce emissions and reduce energy consumption (Ran
and Boyce, 2012).

1.1. Objectives

The main objective of this research is to design, implement and evaluate the effectiveness of a real-time AI powered
smart road system prototype integrating a deep learning model and IoT technology for improving road safety in Nairobi
through monitoring and detecting traffic conditions and traffic regulation offences. The project aims to:
1. Investigate the performance of predictive models for predicting and controlling traffic flow;
2. To design a prediction framework based on IoT and deep learning model;
3. Develop an intelligent road safety system prototype;
4. Investigate the performance/applicability of the developed system prototype in improving road safety.

2. Literature Review

With the current rapid growth in population, cities have experienced a massive expansion in the number of business
activities and daily traffic flow both in and out of the city. This has led to severe traffic congestion. Some research
solutions to this congestion include traffic flow control and management, robust parking systems, city surveillance for
security, among other solutions.
In Nairobi and other urban cities, the major traffic management systems adopted include manual controls using
traffic policemen or the use of a predefined time-based signaling system. The drawback of this signal system is that
the traffic management system functions irrespective of traffic flow. Smart traffic control systems act according to the
real-time traffic conditions. They use a central control system, smart signal lights and intelligent cameras and queue
systems.
Globally, there are 1.35 million road traffic related deaths annually, with road traffic cases being the 8th leading
causes of death for people of all ages (WHO 2018). In Kenya there were 12,839 road related death and injury cases in
2018 and 15,696 cases in 2019 making them one of the alarming causes of death in Kenya (Muguro, Sasaki,
Matsushita, & Njeri, 2020). These high cases of death and injuries have motivated researchers to focus on the creation
of intelligent traffic management systems. These systems provide solutions such as traffic prediction, traffic light
optimization, traffic parameter awareness, accident monitoring and control, and road network analysis amongst others.
The implementation of appropriate intelligent systems minimizes the occurrence of traffic violations.
Traffic violations are likely to cause accidents, and thus if addressed in time will greatly reduce the deaths and
injuries. Traffic accidents cause disturbance in traffic operations and severe urban problems which lead to injuries,
fatalities and loss in economic growth. Most hospitals in Kenya are overwhelmed by an increased number of patients
who report daily for medical attention for injuries from these accidents (Zhang, He, Gao, & Ni, 2018).
With the use of modern technology these traffic offences can be detected and addressed in real time or even before
they occur. The emerging IoT technology allows connectivity via the internet of systems, services, and devices such
as sensors and actuators that allow data to be gathered from the environment and easily shared for timely decisions to
be made (Khan & Khan, 2018). The integration of IoT and deep learning technologies can be used to optimize traffic
lights to reduce traffic congestion and to accurately detect and predict traffic flow. Such IoT and AI algorithms may
 Author name / Computer Vision and Image Understanding 000 (2017) 000–000 3

also be used to detect and predict the occurrence of moving violation traffic offences and cause the appropriate actions
to be taken to minimize the side effects of their occurrence, conduct accident monitoring control and provide
comprehensive road network analysis.
Several studies have shown the need for a traffic flow control and management system in Nairobi (Mbae, 2015).
Such a system will provide reliable information on matters of road and transport safety and inform critical decisions
in implementing related policies. The system will provide adequate data for real time traffic flow analytics, prediction
of moving violations and other hazards that may have negative effects on traffic flow. This would consequently
minimize traffic congestions and improve traffic efficiency and safety.
A traffic flow control system will also promote a reduction in road traffic accidents and a decrease in air pollution,
optimize citizens' transportation and safety experience and help demarcate and warn of black-spots (accidents prone
areas) within the traffic network. Intelligent timing of traffic signals will go a long way in minimizing congestion at
busy intersections (Rusyaidi and Ibrahim, 2020).

3. Materials and Methods

The proposed research will investigate, develop and implement artificial intelligence algorithms to automatically
learn traffic patterns using CCTV infrastructure in Nairobi along major highways to aid in directing traffic flow,
detecting traffic regulation offences and reducing accidents. This research was subdivided into: a critical literature
review on the application of AI algorithms into traffic management; evaluating the current use of CCTV in KURA
traffic control systems; development of a machine learning system prototype and an AI-enabled traffic control pilot
study. A mixed methods approach was adopted, with qualitative, quantitative research design, system design and
empirical proof of concept study.
Data annotation can be done manually or automatically depending on the uniqueness of the attribute to be
identified. The process includes drawing bounding images to specify a feature in the images such as number plates.
Moreover, specified images can be captured to indicate an offense such as crossing the junction while the lights are
red (Sumalee and Ho, 2018).

3.1. Qualitative Research Design

The qualitative component of the research was carried out to assess the current challenges in traffic management.
This was done through analysis of data currently held by KURA. This phase involved focus group discussions with
stakeholders in KURA in order to evaluate existing challenges and gather requirements for the proposed system. The
output of this phase was a factual report on perceived and actual shortcomings of existing traffic control systems
leading to clear requirements for an intelligent traffic flow management system.

3.2. Quantitative Research Design

The quantitative component of the research involved collection of streaming data from CCTV cameras, and
converting this data into discrete data. The output of this phase is data that can be used to train a machine learning
algorithm for traffic control. Depending on big data, which is the voluminous data received from CCTV systems, the
system will analyse the flows of traffic data and come up with predictions that will enable it to improve traffic
management at selected junctions. Using IBM Big Data Analytics, the selected algorithm was trained, tested and
improved using lab-based tests.
The output of this phase was an AI-enabled system that is adaptive to traffic flows at a specified junction. This
system will have components to predict and detect moving violation traffic violations and produce road network
analysis.
4 Eng. Joanne Mahugu / African Transport Research Conference 2024, Cape Town, South Africa

Theory Building
Desktop Research
Critical Literature
Review

Pre-
Experimentation
development System
Optimization of
-Analyze Development
algorithm
current KURA Prototype
Lab-based simulation
traffic data Development
Refinement of algorithm
-FGD with
traffic experts

Pilot Study
Proof of concept

Fig. 1. Research Design

4. Results and Discussion

An AI-enabled system that is adaptive to traffic flows at a specified junction was developed. This system has
components to predict and detect moving violation traffic violations and produce road network analysis. These
include:
• Closed-circuit television (CCTV), also known as video surveillance. This is the use of video cameras to transmit a
signal to a specific place, on a limited set of monitors. The CCTVs cameras were used to collect video stream
data on a particular road junction:
• The Network Video Recorder (NVR) which is a computer system that records video footage and stores it on a hard
disk, a mass storage device, or cloud storage. NVRs are paired with digital internet protocol (IP) cameras to
create a video surveillance system. They are used to temporarily store the video footages and images before
transmitting them to the Data Centre. They also allows the Edge AI Processor to access the video streams from
the CCTV camera. Additionally they powers the CCTV cameras.
• Edge Router which is a device that connects an internal network to the internet. The edge router allows other
components to access the video stream data from the Network video recorder. It offers a robust security network
when accessing data in the CCTV network system. It allows data to be transmitted to the Kenya Urban Roads
Authority Data Centres via the fibre network system. It also allows the Edge AI processor to access the video
stream from the Network video recorder.
• Onsite Data Center which is a facility that centralizes an organization's shared IT operations and equipment for
the purposes of storing, processing, and disseminating data and applications. Data centers contain physical or
virtual servers that are connected internally and externally through networking and communication equipment to
store, transfer and access digital information. The Kenya Urban Roads Authority Data Centre was used for long
term storage of video stream collected by different CCTV camera from various intersections. The data was
transmitted to the centre via optical fibre cables. The data centre video streams can be accessed, replayed,
reviewed and extracted by the traffic controller administrators.
 Author name / Computer Vision and Image Understanding 000 (2017) 000–000 5

• AI Edge Processor is a combination of an Edge Computing and Artificial Intelligence. Edge computing allows
analysis and processing of data at the point of data collection. Artificial Intelligence mimics human reasoning,
hence this allows high level analytics combined with automation. It was used to analyse the video stream, hence
produce parameters such as queue length, number of vehicles, type of vehicles, traffic violations, traffic
conditions i.e., accidents on the junction. The AI Edge processor sends the information to the signal processor
systems that is used as input for traffic lights control. The AI Edge processor also sends information on the traffic
violation to the e-police system as response to curb road intersection and junctions traffic rules violation which
mainly results to accidents and traffic snarl ups. Additionally, the AI Edge processor can extract additional data
required by traffic engineers in development of road planning and design. Its benefits includes:
1. Zero latency in processing the video stream data
2. Real time analytics, hence able to intelligently detect the traffic conditions onsite.
3. Scalability: while many CCTV camera are installed with onboard AI analytics systems, they are not custom
made to suit the needs of a particular country infrastructure and traffic control needs.

Fig. 2. System Design

5. Conclusion

An efficient ITS traffic flow management system can be used to predict future traffic flow situation. Traffic signal
optimisation and coordination can greatly enhance the movement of vehicles along a traffic network. Delay on the
network and queue lengths on approaches to intersection can be reduced, while increasing overall network operating
speed and improving emissions levels. It is proposed that similar traffic signal coordination systems can be used in
Africa to reduce congestion in cities without the need for expensive updating of traffic signal timings through
scheduled traffic counts and manual traffic signal optimisation.
Real-time optimisation of the intersections along a road network can respond to changing traffic patterns over time,
caused by traffic incidents or special events. With this information, traffic management agencies can advance
strategies to realize the best scenario and prevent the congestion from getting worse, to create connected cities with
seamlessly integrated and efficient transport networks. Intelligence Traffic Systems are applied in traffic management
and interfaces with other modes of transportation to relieve traffic congestion, improve safety, and reduce emissions
and reduce energy consumption (Ran and Boyce, 2012). Additionally, government transport agencies can generate
revenue through automation of traffic violations.
The key deliverables that were realized at the end of the research were:
1. A smart (AI-enabled) prototype to predict and monitor traffic flow;
2. An IoT enabled prototype to detect, report and prevent moving violation traffic offenses;
3. A smart real-time traffic flow management system that uses AI to intelligently direct traffic.
6 Eng. Joanne Mahugu / African Transport Research Conference 2024, Cape Town, South Africa

References

1. Kenya National Bureau of Statistics (KNBS) (2019). National Population Census Report.
2. Khan, M. A., & Khan, S. F. (2018).IoT based framework for Vehicle Over-speed detection. 1st International
Conference on Computer Applications and Information Security, ICCAIS 2018, 2018–2021.
https://doi.org/10.1109/CAIS.2018.8441951
3. Mbae, 2015.Nairobi to have 1.35 million vehicles in 2030.The Standard.
https://www.standardmedia.co.ke/entertainment/city-news/2000154645/nairobi-to-have-135-million-vehicles-in-
2030
4. Muguro, J. K., Sasaki, M., Matsushita, K., &Njeri, W. (2020). Trend analysis and fatality causes in Kenyan
roads : A review of road traffic accident data between 2015 and 2020 Trend analysis and fatality causes in
Kenyan roads. Cogent Engineering, 7(1). https://doi.org/10.1080/23311916.2020.1797981.
5. Ran, B., & Boyce, D. (2012). Modeling Dynamic Transportation Networks: An Intelligent Transportation System
Oriented Approach. Springer Science & Business Media.
6. Rusyaidi, M., & Ibrahim, Z. (2020). A Review: An Evaluation of Current Artificial Intelligent Methods in Traffic
Flow Prediction. In IOP Conference Series: Materials Science and Engineering (Vol. 917, No. 1, p. 012063). IOP
Publishing.
7. Sumalee, A., & Ho, H. W. (2018). Smarter and more connected: Future intelligent transportation system.
International Association of Traffic and Safety Sciences, 42(2), 67-71.
https://doi.org/10.1016/j.iatssr.2018.05.005
8. WHO (2018). Global Status Report On Road Safety 2018.
9. Zhang, Z., He, Q., Gao, J., & Ni, M. (2018). A deep learning approach for detecting traffic accidents from social
media data. Transportation Research Part C: Emerging Technologies, 86(November 2017), 580–596.
https://doi.org/10.1016/j.trc.2017.11.027

View publication stats