Traffic Intersection Optimization Based on Random Forest and

SUMO Simulation in Xi’an

Xianqi Dai

Chang'an Dublin International College of Transportation, Chang'an Unversity, Shaanxi, China

Keywords: Traffic Optimization, AutoCAD, Random Forest, SUMO, Traffic Signal Control.

Abstract: With the rapid pace of urbanization, traffic congestion, especially at intersections, has become a significant

challenge in cities like Xi'an, China. This study conducted a comprehensive analysis of traffic congestion

issues, with a specific focus on the intersection of Shang Hong Road and Shang Ji Road, a critical bottleneck

area in Xi'an City. Utilizing AutoCAD for road layout optimization design, the paper simulated the optimized

solution using the SUMO simulation tool. Furthermore, a random forest model was employed to predict traffic

flow, leading to recommendations for optimizing signal light duration and lane configuration. The research

findings indicate that an enhanced traffic network design and signal light configuration can significantly

improve intersection throughput, reduce delay times, and enhance traffic safety. This study provides

scientifically grounded optimization suggestions for urban traffic management authorities, offering practical

measures to improve traffic efficiency, alleviate congestion, and contribute to safer and more sustainable

urban environments.

1 INTRODUCTION

With the acceleration of urbanization, the problem of

traffic pressure has gained increasing attention from

the public, especially in traffic intersections where

heavy congestions seriously hinder the smooth flow

of commuting vehicles (Afrin and Yodo, 2020). For

example, in many cities in China, due to heavy traffic

and unreasonable road design, intersections have been

held up for a long time, causing great inconvenience

to citizens' daily lives. These congestions influence

the travel safety and commute efficiency of citizens.

Moreover, extended periods of stagnation and low

speed of vehicles lead to more greenhouse gas

emissions, resulting in environmental pollution.

Therefore, it is of immense importance to improve the

overall performance of the existing urban

transportation system and the layout of the road

networks to alleviate the congestion at intersections

effectively.

However, the current traffic junction design is

inefficient for signal control and timing, and lane

settings have improper situations. For example, the

traffic signal timing scheme fails to fully consider the

https://orcid.org/0009-0007-9696-2130

dynamic change of traffic flow, resulting in the traffic

flow in peak hours not being effectively channelling.

The lane turning setting is not reasonable, the number

of lanes does not match the traffic flow, causing some

lanes in a saturated state for a long time. These

urgently need scientific analysis and optimization to

improve their operational effectiveness. In this

context, big data applications and intelligent

transportation systems are increasingly required to

provide data-driven solutions to optimize traffic

routes, traffic distribution, and general management.

(Brown et al., 2022). These integrations increase the

responsiveness and flexibility of urban transport

systems to reduce congestion and improve safety and

environmental quality (Haydari and Yilmaz, 2022).

The application of intelligent transportation

systems (ITS) and big data algorithms in urban traffic

management is getting more attention from academia

and industry. They can provide a scientific basis for

transportation management and optimization based

on real-time traffic flow monitoring and massive data

analysis (Cheng, Pang and Pavlou, 2020) According

to the research, scholars have proposed various

methods to improve the current traffic congestion in

intersections. For example, refining the timing of

Dai, X.

Trafﬁc Intersection Optimization Based on Random Forest and SUMO Simulation in Xi’an.

DOI: 10.5220/0013329600004558

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 1st International Conference on Modern Logistics and Supply Chain Management (MLSCM 2024), pages 301-309

ISBN: 978-989-758-738-2

301

signal lights, and upgrading lane Settings, could

alleviate traffic congestion, improve traffic efficiency,

and reduce environmental pollution. Eom and Kim

(2020) argue that a reasonable timing schedule for

signal lights can significantly improve the capacity

and safety of the roads. Zhao et al. (2013) suggest that

meticulously designed lane turn signs can ensure that

vehicles enter and exit smoothly without causing

severe traffic congestion. In addition, simulation

technologies such as Simulation of Urban Mobility

(SUMO) are becoming widely used in traffic

management and optimization. The results of these

simulations can work as vital support for renovating

the intersection lane layout (Shirazi, Morris and

Zhang, 2023). Through these analyses, the effect of

distinctive design plans can be evaluated to provide a

scientific basis for perfecting traffic intersections

To solve the traffic problem and the inefficiency

of the intersection, this paper carries out research

experiments to study the design problems in the

design of traffic intersections on Shang Hong Road

and Shang Ji Road in Xi'an City and proposes

improvements based on calculation and analysis. The

article will achieve the research goals based on the

following steps: Firstly, it will collect and analyze the

traffic flow during morning and evening and flat

peaks. Additionally, the paper will redesign the

junction using CAD software. Moreover, it will use

SUMO simulation software to simulate and analyze

the improved design and propose optimized signal

timing schemes and lane design suggestions based on

the simulation results. This study is expected to

provide scientific suggestions for urban traffic

management departments to improve the operation

efficiency and safety of traffic intersections.

2 RESEARCH AREA AND

METHODOLOGY

2.1 Overview of the Targeted Area

2.1.1 Introduction of the Intersection

With the progress of urbanization, the traffic pressure

of Xi 'an is increasing daily, and the traffic congestion

problem has become one of the bottlenecks restricting

the city's further development. Especially in some

important traffic nodes and economic development

areas, the traffic flow is dense, and the congestion is

serious. Therefore, it is of great practical significance

to solve the traffic problems in these areas. Shang

Hong Cross Road is a typical bottleneck area, located

at the intersection of Xi 'an Economic Avenue and

Shang Ji Road Science and Technology Development

Zone. The north-south main Road Shang Hong Road

and the National Highway 1310 Express Road

(known as Shang Ji Road) form this junction. The

intersection is located in a complex area of

infrastructure, the northern area is mostly residential

areas, and the south, east, and west are distributed

with several types of factories and commercial areas.

This diverse regional functional layout brings a lot of

commuting and coordination traffic needs. In addition,

there are many important traffic attractions near the

intersection, such as Xi 'an Economic Development

Zone No. 8 Primary School, Yurun Market, and

Master Kong Storage and Transportation Center,

which further aggravate the traffic flow and

congestion at the interchange. Therefore, it is of great

significance to study and optimize the traffic

organization and management of Shang Hong Cross

Road to improve the traffic situation in this area

(Figure 1).

Figure1: Actual map of the intersection (Photo/Picture

credit : Original).

2.1.2 Current Problems

The convergence of these major roads at the

intersection leads to an increasing flow of vehicles,

exacerbating the traffic congestion risk. Traffic

congestion is particularly severe during peak hours,

causing significant delays and increasing the risk of

accidents. The presence of important educational and

commercial establishments further increases the

traffic volume, with parents picking up students and

commercial vehicles moving in and out of storage

facilities, compounding the traffic problem. The

existing road infrastructure cannot accommodate such

a high volume of traffic, resulting in a bottleneck

phenomenon that affects not only the traffic in the

MLSCM 2024 - International Conference on Modern Logistics and Supply Chain Management

302

area but also the overall traffic flow in the technical

development zone. To address this issue,

comprehensive traffic management strategies and

infrastructure improvements are required to ease

congestion and ensure smoother traffic flow.

2.1.3 Data Collection

Through field measurement and mechanical counting

method, the study obtained the following traffic flow

of left-right turning and going straight through the

intersection within one hour of the morning and

evening peak. Through the vehicle conversion

coefficient table in Table 1,

Table 1: Vehicle conversion factor.

Vehicle

Type

Vehicle

Conversion

Facto

Description

Small

Vehicle

1.0

Passenger vehicles with

≤19 seats and cargo

vehicles with ≤2t load

capacity.

Medium

Vehicle

1.5

Passenger vehicles

with >19 seats and

cargo vehicles with >2t

to ≤7t load capacity.

Large

Vehicle

2.0

Cargo vehicles with >7t

to ≤14t load ca

acit

Trailer 3.0

Cargo vehicles with >7t

to ≤14t load capacity.

2.2 Methodology

AutoCAD software is produced by the United States

Autodesk Co., LTD. (Autodesk), which is an

automatic computer-aided design software, that can

be used to draw two-dimensional drawings and basic

three-dimensional designs, it does not need to know

programming, you can automatically draw, so it is

widely used in the world, can be used in civil

construction, decoration, industrial drawing,

engineering drawing, engineering drawing,

electronics industry, clothing processing, and other

fields. CAD tools can be used to design and optimize

road layouts, ensuring proper allocation of lanes for

vehicles, pedestrians, and cyclists to reduce traffic

congestion (Omura and Benton, 2019).

SUMO is an open-source traffic simulation

software that can simulate and analyze urban traffic

flow. Using SUMO, city planners can predict the

impact of different planning options and determine

the most efficient transportation solutions. Moreover,

with SUMO, it is possible to simulate urban traffic

conditions under different traffic flows and evaluate

the feasibility of different traffic plans. In addition,

traffic signal Settings can be optimized through

simulation to reduce vehicle waiting times and

improve overall traffic efficiency. This integrated

approach enables a comprehensive assessment and

improvement of urban traffic management strategies,

contributing to a more effective and efficient urban

traffic flow (Krajzewicz, 2010).

2.3 Traffic Flow Prediction Using

Random Forest Model

This paper will apply the Random Forest Model to

experiment with traffic flow prediction. Random

Forest is a supervised data mining algorithm. It is a

classifier model composed of multiple CART

decision trees. The resulting decision trees form a

random forest model (Liu and Wu, 2017).

The Random Forest model's prediction formula

can be succinctly expressed as:



∑





i=1

(1)

where (𝑌



) represents the estimated total traffic

volume, (𝑁) denotes the number of decision trees,

(𝑓



(

𝑋

)

) corresponds to the prediction outcome of

the (𝑖) −𝑡ℎ decision tree, and (𝑋) encapsulates the

input feature vector, including the time (in hours and

minutes). This ensemble method aggregates the

predictions from individual decision trees, effectively

capturing complex relationships within the data,

leading to more robust and accurate traffic volume

forecasts.

In this case, the formula of the Random Forest

Model can be used as:

𝑌









∑

𝑓



(

Hour



,Minute



)





(2)

Where(Y





) represents the estimated total traffic

volume at the time(t) , (N) denotes the number of

decision trees (set to 100 in this example),



(

Hour



,Minute



)

) corresponds to the prediction

outcome of the (i) − th decision tree based on the

hour and minute at time (t), and (X) encapsulates the

input feature vector, including the time (in hours and

minutes). This ensemble method aggregates the

predictions from individual decision trees, effectively

capturing complex relationships within the data,

leading to more robust and accurate traffic volume

forecasts.

Trafﬁc Intersection Optimization Based on Random Forest and SUMO Simulation in Xi’an

303

3 RESULT ANALYSIS

3.1 Using CAD to Upgrade the

Intersection Network

Figure 2: CAD drawing of the original intersection

(Photo/Picture credit : Original).

Figure 2 shows the existing road network structure,

the setting of turning lanes is not reasonable. Due to

the insufficient width of the lane or the unreasonable

setting of the signal light, vehicles interfere with each

other when turning, which is easy to cause a collision,

thus obstructing the smooth passage of traffic at the

intersection. Secondly, the setting of the safety islands

is not perfect, only the east and west safety islands are

set up. The north-south road surface is wider, but there

is no safety island, so it causes a safety risk for

pedestrians crossing the street. In addition, the

intersection does not effectively mark the specific

location of the non-motorized lane, resulting in the

phenomenon of grabbing the lane between non-motor

vehicles and motor vehicles, which not only causes

congestion but also causes security risks.

Figure 3: CAD drawing of the improved intersection

(Photo/Picture credit : Original).

The upgrade plan for the network is shown in

Figure 3. In the figure, this research changed the

steering of the north-to-south lanes and concentrated

the right-turn lanes on the right side of the road. This

way, there will be no congestion and traffic standstill

caused by straight-going vehicles and right-turning

vehicles competing under the same green light.

Secondly, considering the width of the intersection

and the possibility of pedestrians crossing the street

twice, this paper added two safety islands at the east-

west zebra crossing. According to the standard, the

width of the safety island is at least 1.5 meters, and if

it needs to accommodate wheelchairs, etc., the width

needs to be set to 2 meters. At the same time, the

safety island design used in the paper will have

sections at both ends that are raised above the ground

to enhance the safety of pedestrians and prevent

vehicle strikes. At the same time, the middle part of

the safety island is level with the ground, which is

convenient for the elderly and disabled people to use.

In addition, the middle part of the safety island is

coated with anti-slip lines to prevent slipping in rain

and snow and improve visibility, and its two ends are

also coated with yellow reflective material, which is

easier to see at night (Li, Yang and Yin, 2010). Finally,

this research set up a non-motorized lane, as shown in

the brown area in the figure, which separates the non-

motorized traffic flow from the motorized traffic flow,

improving the efficiency and safety of the intersection

3.2 Prediction Result Analysis

3.2.1 Morning Peak Analysis

As shown in Figure 4, during the morning peak, the

traffic flow prediction indicates that in the coming

hour, there will be an increasing number of vehicles

passing through the intersection, especially between

8:45 and 9:20. Traffic started at 8:45 am with 169

vehicles, then steadily increased to 194 vehicles at

8:55 am and reached its peak at 9:20 am with 209

vehicles. Subsequently, traffic decreased slightly to

164 vehicles at 9:30 a.m. and 152 vehicles at 9:40 a.m.

This trend indicates that traffic congestion will

gradually increase during peak hours in the morning,

especially around 9:20 am, when traffic may reach

saturation. Therefore, to reduce traffic congestion

around 9:20, it is suggested that the design of traffic

lights be adjusted, traffic flow should be restricted

during rush hour, and management of the surrounding

area should be optimized.

MLSCM 2024 - International Conference on Modern Logistics and Supply Chain Management

304

Figure 4: Traffic volume forecast in the morning peak (Photo/Picture credit: Original).

Figure 5: Traffic volume forecast in the evening peak (Photo/Picture credit: Original).

For traffic from north to south, as forecast,

morning peak traffic is steady but tends to decrease.

From 8:45 to 12:22, it gradually decreases until 9:40

to 64h. The traffic volume in this direction is small

and will decrease gradually. However, attention must

be paid to traffic management to prevent unforeseen

events from causing disruptions. For example, early

morning traffic management should be improved,

sufficient labor and equipment should be provided,

and accidents or technical failures should be recorded

promptly to reduce the risk of traffic accidents.

3.2.2 Evening Peak Analysis

During evening peak periods, north-south traffic

shows significant variations, especially from north to

south. Concrete data in Figure 5 show that north-south

traffic increased from 142 at 6:35 to 193 at 6:40, with

a peak of 200 at 7:10. Traffic subsequently fluctuated

between 175 and 200 vehicles and continued until

7:30. Traffic from the south to the north, on the other

hand, is relatively stable and is expected to fluctuate

between 113 and 137 vehicles between 6:35 and 7:30.

This forecast indicates that the north-to-south

tracks could face significant congestion, especially

during the evening rush hour at approximately 7:10.

In addition, large gaps between north and south can

lead to bottlenecks at intersections, further

exacerbating congestion. It is therefore recommended

that the operating time of the traffic lights be

dynamically adjusted during the evening rush hour to

give priority to north-south traffic. At the same time,

consideration could be given to creating additional

temporary lanes in the north and south directions to

optimize the use of the lanes and allow rush-hour

vehicles to move quickly in the priority sections.

When the morning and evening peak traffic

forecasts are considered, traffic varies between

periods and directions. During the morning peak,

traffic reaches its highest at 9:20 (209 vehicles) before

tapering off from north to south to the lowest (64

vehicles). During the evening peak, the traffic from

north to south fluctuates widely, up to 200 vehicles,

while the flow from south to north is more stable.

Trafﬁc Intersection Optimization Based on Random Forest and SUMO Simulation in Xi’an

305

These data show that some important sections present

a clear risk of congestion during peak hours and

require appropriate management measures.

To effectively address these challenges, improve

real-time traffic monitoring and use intelligent

transport systems (its) to dynamically adjust signaling

periods and conduct traffic strikes by adding

temporary lanes and optimizing lane use. Tempering

returns should be considered. In addition, continuous

data analysis and forecasting will provide an

important reference for future traffic management and

will help to develop more scientific traffic channeling

scenarios and improve the overall efficiency of traffic

exploitation.

3.3 SUMO Simulation Analysis

3.3.1 Network Structure and Signal

Optimization

The diagram below shows the typical design of a four-

lane crossing in SUMO. To manage the volume of

traffic in the north-south direction, the emphasis is on

improving the tracks in the north-south direction. The

design uses a crossing plan to handle north-south

traffic. The track configuration in all directions

includes a left, right, and right turn. The tracks

intersect in the central section of the crossing,

allowing safe and smooth movement of vehicles in all

directions. At the same time, detectors are installed at

each entrance. These detectors can monitor traffic in

real-time in all directions and provide data to

dynamically adjust signal synchronization. Crossings

are designed to maximize vehicle traffic efficiency

and reduce potential collisions and congestion.

Figure 6: Road structure image after improving the lane

network using SUMO (Photo/Picture credit : Original).

The diagram below shows the optimal signal time

at this crossing. Based on traffic data provided by the

SUMO simulation tool, the green light time is

dynamically adjusted according to traffic to reduce

vehicle delays and improve the effectiveness of bans.

Especially in directions where there is significant

traffic in the north-south direction, the duration of the

green light was extended, allowing vehicles from that

direction to cross the crossing quickly, reducing

waiting times and potential congestion. As shown in

the chart, from north to south, the green light is 20

seconds long, the red light is 63 seconds long, and the

yellow light is fixed for 3 seconds in a signal period.

At the same time, the green light is 40 seconds long

and the red light is 25 seconds long in a signal period

from south to north.

With the improved traffic light configuration, it is

possible to effectively avoid the north-south and east-

west traffic lights at the same time and reduce traffic

congestion. This time allocation, which allows

vehicles to travel in one direction at all times, reduces

the risk of accidents and improves crossing safety. A

reasonable allocation of time can improve crossing

Figure 7: Signal light timing improvement diagram (Photo/Picture credit: Original).

MLSCM 2024 - International Conference on Modern Logistics and Supply Chain Management

306

capacity. By analyzing traffic in different directions,

this program defines a reasonable green light time

based on actual traffic, which ensures the efficient

passage of traffic. Also, switch traffic lights regularly

every week to make traffic flow smoother. The

optimized signal configuration scheme ensures the

stability of the road sign system in the actual

operation by clearly setting the green, yellow, and red

times. This stable signal switching reduces driver

uncertainty and improves road traffic reliability.

3.3.2 Simulation Results Comparison

This study analyses in detail traffic performance

indicators before and after upgrading the crossing

network and traffic light optimization using SUMO

simulation data. The indicators analyzed are the

Average Travel Time, the Average Speed, the Average

Time Lost, the Average Time Lost Within, and the

Average Duration Within. The following is a

summary of the results of the analysis.

As shown in Tables 6 and 7, the average left-hand

car travel time increased from 47.97 seconds to 43.32

seconds, an improvement of approximately 9.7%. The

average driving time of direct line vehicles decreased

by 21.7% from 21.84 seconds to 17.15 seconds. The

average driving time of vehicles turning to the right

decreased from 17.91 seconds to 14.05 seconds, an

improvement of 21.5%. The bar chart in Figure 8

demonstrates directly that these reductions show that

optimization measures have significantly improved

driving efficiency in all directions.

As shown in Figure 8, the average speed in all

directions was increased by optimization. Tables 6

and 7 indicates that the average speed of the left cars

increased by 42.0% from 2.76km /h to 3.92km /h. The

average speed of straight-line vehicles increased by

14.8% from 6.16 km/h to 7.07 km/h. The average

speed of right turns increased from 4.23 km/h to 4.42

km/h, an increase of 4.5%. The increase in average

speed means smoother traffic and less congestion.

Table 2 and Table 3 show that the average lost time

is an important measure of the total number of

vehicles delayed at crossings. Optimized traffic

decreases in all directions. The average time lost by a

left-hand vehicle decreased from 44.00 seconds to

37.23 seconds, a reduction of 15.4%. The average

time lost for direct line vehicles decreased by 11.8%

from 17.76 seconds to 15.67 seconds. The average

time lost to a right turn decreased from 15.61 seconds

to 10.01 seconds, a reduction of 35.7%. These results

show that optimization significantly reduces delay

time in all directions.

Figure 8: Comparison of different indicators before and after optimization (Photo/Picture credit: Original).

Trafﬁc Intersection Optimization Based on Random Forest and SUMO Simulation in Xi’an

307

Table 2: Simulation results of different indexes before optimization.

Original Average Travel

Time

Average

Spee

Average Time

Los

Average Time Lost

Within

Average Duration

Within

Lef

47.97 2.76 44.00 30.47 32.10

Strai

21.84 6.16 17.76 34.93 36.95

17.91 4.23 15.61 39.43 41.56

Table 3: Simulation results of different indexes after optimization.

Improved Average

Travel Time

Average

Spee

Average Time

Los

Average Time

Lost Within

Average

Duration Within

Lef

43.32 3.92 37.23 24.34 30.26

Strai

17.15 7.07 15.67 30.45 33.18

14.05 4.42 10.01 30.92 35.62

The average time lost within the crossing has also

been improved as shown in Figure 8. The average

internal time lost by left-hand cars decreased from

30.47 seconds to 24.34 seconds, a reduction of 20.2%.

The average loss of internal time for cars in direct

service was reduced from 34.93 seconds to 30.45

seconds, a reduction of 12.7%. The average loss of

internal time for cars that made a right turn was

reduced from 39.43 seconds to 30.92 seconds, a

decrease of 21.6% as shown in Tables 6 and 7. It

reduces the stay time of vehicles inside the

intersection and improves the efficiency of the

passage.

The average crossing time has also been improved

through optimization. The average length of a left turn

decreased by 5.7% from 32.10 seconds to 30.26

seconds. The average duration of direct line vehicles

decreased by 10.3% from 36.95 seconds to 33.18

seconds. Average right-hand vehicle travel time

decreased from 41.56 seconds to 35.62 seconds, a

reduction of 14.2% as shown in Tables 6 and 7. These

improvements have significantly reduced the average

stopping time of vehicles at crossings.

Overall, improvements to the crossing system and

optimization of signal lights have significantly

improved traffic efficiency. It reduces the time and

waste of time on the road and also increases the

average speed of vehicles, as demonstrated in Figure

9, after optimization, the number of vehicles queuing

in the lane is significantly reduced. These

improvements will improve traffic flow and reduce

congestion, making an important contribution to

urban traffic management.

Figure 9: Comparison of the original lane queue (Left) with

the improved lane queue (Right) (Photo/Picture credit :

Original).

4 CONCLUSION

This research focused on a comprehensive analysis

and design optimization to address traffic congestion

at the intersection of Shang Hong Road and Shang Ji

Road in Xi'an City. The project involved enhancing

the road network layout through AutoCAD software

and simulating the proposed improvements using the

SUMO simulation tool. Additionally, a random forest

model was utilized to forecast traffic flow, leading to

recommendations for optimizing signal light

durations and lane arrangements.

The findings indicate that the revised traffic

network design and signal configurations can enhance

the intersection's capacity, minimize vehicle delay

times, and boost overall traffic safety. Specifically,

The average travel time of left-turning, right-turning,

and straight-moving vehicles experienced a

significant reduction, resulting in improvements of

9.7%, 21.7%, and 21.5% respectively. Furthermore,

there was a notable increase in average travel speeds:

left-turning vehicles improved from 2.76 km/h to 3.92

km/h; straight-moving vehicles increased from 6.16

km/h to 7.07 km/h; and right-turning vehicles rose

MLSCM 2024 - International Conference on Modern Logistics and Supply Chain Management

308

slightly from 4.23 km/h to 4.42 km/h—all indicating

that optimized designs effectively reduce delays while

enhancing vehicle throughput efficiency.

Looking into the future, it is recommended that

real-time monitoring systems be further integrated

with adjustments made to signal lights and lane

configurations in response to evolving traffic patterns

over time. In the future, consider introducing deep

learning models, such as long short-term memory

networks (LSTMs), to capture time-dependent

relationships in traffic flow. Reinforcement learning

algorithms can also be used to optimize signal timing

dynamically and adjust based on real-time data. They

could provide dynamic solutions for urban

transportation systems while promoting smoother

flows and environmental sustainability within city

traffic networks.

REFERENCES

Afrin, T. and Yodo, N. 2020. A Survey of Road Traffic

Congestion Measures towards a Sustainable and

Resilient Transportation System. Sustainability, [online]

12(11), p.4660.

Brown, C., Hardman, M., Davies, N. and Armitage, R. 2022.

Mobility as a Service: Defining a Transport Utopia.

Future Transportation, 2(1), pp.300–309.

Cheng, Z. (Aaron), Pang, M.-S. and Pavlou, P.A. 2020)

Mitigating Traffic Congestion: The Role of Intelligent

Transportation Systems. Information Systems Research,

31(3).

Eom, M. and Kim, B.-I. 2020. The traffic signal control

problem for intersections: a review. European Transport

Research Review, 12(1).

Haydari, A. and Yilmaz, Y. 2022. Deep Reinforcement

Learning for Intelligent Transportation Systems: A

Survey. IEEE Transactions on Intelligent

Transportation Systems, 23(1), pp.11–32.

Krajzewicz, D. 2010. Traffic Simulation with SUMO –

Simulation of Urban Mobility. Fundamentals of Traffic

Simulation, 145, pp.269–293.

Li, L., Yang, X. and Yin, L. 2010. Exploration of Pedestrian

Refuge Effect on Safety Crossing at Signalized

Intersection. Transportation Research Record: Journal

of the Transportation Research Board, 2193(1), pp.44–

50.

Liu, Y. and Wu, H. 2017. Prediction of Road Traffic

Congestion Based on Random Forest. [online] IEEE

Xplore.

Omura, G. and Benton, B.C. 2019. Mastering AutoCAD

2019 and AutoCAD LT 2019. Hoboken, Nj: Sybex, A

Wiley Brand.

Shirazi, M.S., Morris, B.T. and Zhang, S. 2023. Intersection

Analysis Using Computer Vision Techniques with

SUMO. Intelligent Transportation Infrastructure,

[online] 2.

Zhao, J., Ma, W., Zhang, H.M. and Yang, X. 2013.

Increasing the Capacity of Signalized Intersections with

Dynamic Use of Exit Lanes for Left-Turn Traffic.

Transportation Research Record: Journal of the

Transportation Research Board, 2355(1), pp.49–59.

Trafﬁc Intersection Optimization Based on Random Forest and SUMO Simulation in Xi’an

309