SAITraC - Smart Artificial Intelligence based Traffic Controller

Pakistan is a developing nation and requires progress in certain areas. In urban areas, problem of traffic congestion has arisen mainly due to increase in number of personal vehicles. Secondly, due to insufficient capacity of vehicles on roads. This causes huge amount of traffic in certain areas dur

Project Title

SAITraC - Smart Artificial Intelligence based Traffic Controller

Project Area of Specialization

Artificial Intelligence

Project Summary

Pakistan is a developing nation and requires progress in certain areas. In urban areas, problem of traffic congestion has arisen mainly due to increase in number of personal vehicles. Secondly, due to insufficient capacity of vehicles on roads. This causes huge amount of traffic in certain areas during certain times of the day. Also, the infrastructure (i.e. bridges or roundabouts) can not be made due to less space. We are proposing to build an intelligent traffic management system, which will have the capability to alter the traffic signal green light time according to number of vehicles present at an intersection.

The aim of this project is to replace the conventional traffic management system with this intelligent one. Initially, we will take a video and apply object detection using SSD MobileNet V2 Model. Then using the Machine Learning algorithms we will calculate the timer for Green-Light. The final prototype i.e., real-time implementation will be done after successful results. 

Project Objectives

• Design an intelligent traffic management system to replace the conventional system with SAITraC
• Identifying and tracking cars and their density in a lane using TensorFlow Object Detection API's
• Implement this structure using Machine Learning Neural Networks to control traffic flow smoothly
• Real-time traffic light switching according to current traffic density
• Efficient management of time and resources

Project Implementation Method

The implementation of this project will require a LCD screen with a video dataset, Jetson Nano 4GB and on-screen Traffic Lights. The screen is splitted in four parts having videos of four lanes of the intersection. On the top of each lane, there is an on-screen traffic signal. All this scenario is being controlled by a Jetson Nano which takes these videos as its input. It processes the inputs and controls the timer for green light. Based on those timers, we then control the working of the traffic signal. This is a practical demonstration of a real time intelligently working traffic signal. This process of learning the situation of traffic continuously and making decisions based on that will be repeated after every interval of 10 minutes. 

• In this process, we started meetings with Safe City Capital Command and Control Center Islamabad. Asked them for real time traffic data at Faisal Chowk Islamabad. 
• We trained our neural network based on Real-time Traffic Dataset recorded on Faisal Chowk Intersection Islamabad. And successfully got new intelligent timers for the Green-Light.
• A system which actively detects the vehicles from a webcam, counts the density, and based on these traffic conditions sets the Green-Light timer of a traffic signal

Benefits of the Project

Proposed Benefits

• Less expensive than sensors
• Virtually no new hardware installation
• Save Fuel Consumption which will reduce Transportation Costs, etc.
• Implementation of traffic rules could be made easier

Impact of Project on Society

• Save human time wasted in traffic daily & Reduce stress of drivers
• It will be eco-friendly, i.e. less air pollution in consistent area, as it will reduce emission of CO2
• No need of man-power as the system will be autonomous

Impact of Project on Economy

• The reduction in fuel consumption will lead to decrease in fuel subsidies, this directly effects our GDP.

Technical Details of Final Deliverable

The first stage is to detect vehicles present in all lanes of traffic intersection and make a datset. For that, we will check from the video whenever all signals are red, it extracts a frame. We apply object detection on frame and recognize all vehicles in it. Then we count the no. of detected vehicles.

We need to analyze the dataset, and extract data of "Density of Vehicles passing through the intersection at different Time Intervals from respective Lanes." All information is recorded in an Microsoft Excel file. The dataset contains density of vehicles passing through the intersection at different times in different lanes. We then measure the actual timers set on signal and also the intelligent timers needed to make the system Artificially Intelligent. 

This dataset is fed to a deep learning neural network algorithm (80%train data and 20%test data) which predicts the new intelligent timers required to make signal intelligent. In this way, Density calculation is done. Now that all traffic on four lanes of intersection is calculated, we will set priority (as 1, 2, 3, 4) of lanes on basis of which signal will get Green Light.

This process continues for next 10 minutes and in mean time all the above procedure will be repeated again.

Final Deliverable of the Project

HW/SW integrated system

Core Industry

Transportation

Other Industries

Education , Telecommunication

Core Technology

Artificial Intelligence(AI)

Other Technologies

Internet of Things (IoT)

Sustainable Development Goals

Good Health and Well-Being for People, Affordable and Clean Energy, Decent Work and Economic Growth, Industry, Innovation and Infrastructure, Sustainable Cities and Communities, Life on Land

Required Resources

If you need this project, please contact me on contact@adikhanofficial.com