- Artificial intelligence (AI)-based autonomous traffic regulation refers to the management and control of traffic flow. In order to collect real-time data on traffic conditions, sensors, cameras, and communication networks are used. This data is then evaluated and processed by AI algorithms to produce insights and make judgement. AIpowered autonomous traffic regulation aims to increase system efficiency by reducing congestion, increasing safety, and all of the above. The advantage of using autonomous traffic regulation utilizing AI is the ability to process and collect large real time data and conclusions are drawn. This enables the system to adjust the traffic flow fast in response to shifting traffic circumstances. Algorithms based on AI can also be used learn from previous traffic patterns and situations to create future forecasts and conclusions that are more accurate. For autonomous traffic regulation, a variety of AI algorithms, which includes reinforcement learning machine learning, deep learning, can be applied. Algorithms based on Deep learning can be used to interpret photos, video data from cameras, spotting patterns and trends in traffic data can be achieved through machine learning algorithms. Algorithms for reinforcement learning can be used to learn from the past and make choices based on reward signals. To guarantee their dependability and safety, it is crucial to make sure that these systems are designed and deployed with the proper protections. This AI-powered system can also adjust in real-time to shifting traffic patterns and road conditions, making the traffic regulating process more responsive and dynamic. As a result, there may be an improvement in traffic-related emissions reductions and fuel efficiency. Overall, the AI is used for the development of intelligent transportation systems which has advanced significant, which has the potential to revolutionize traffic management and assure a more effective, safe, and sustainable transportation system.
The automation of traffic management and control is accomplished here by development of an autonomous traffic regulator. It enhances the safety and effectiveness of roadways by using a variety of technologies such as cameras, signal controllers and artificial intelligence algorithms to detect and adapt to traffic patterns in realtime. Reduced traffic congestion, lower accident risk, and improved vehicle flow are the objectives of an autonomous traffic regulator. Image detection, image processing, density calculation, communication networks, an efficient signal switching algorithm and a centralized control system are essential parts of an autonomous traffic regulation system.
Autonomous Traffic Regulators use a combination of such technologies and algorithms to collect and analyze data about the flow of traffic. This information is then used to control the traffic lights, which in turn help us in regulating the flow of traffic, time spent by each vehicle on the road and lesser time delays. This further helps in reducing congestion and hence reducing the carbon emissions on the road.
The ATR system's ability to reduce travel time and fuel consumption is one of its main advantages. The ATR represents a significant advance in the future of traffic management given the rising demand for smart cities and the creation of intelligent transportation systems. It has the ability to fundamentally alter how we control traffic and guarantee a more effective, secure, and sustainable transportation system. Increased road safety is a benefit of the ATR system as well. The system can make decisions that reduce the danger of crashes and other traffic-related occurrences by assessing real-time data on traffic patterns and road conditions. This can aid in lowering the amount of collisions and fatalities on the roads, hence enhancing the safety of the roadways for all users. But there are enormous potential advantages, and technology is developing quickly.
The ATR system offers a viable answer to one of the most critical issues facing modern cities as they continue to grow and traffic congestion worsens.
The necessity to overcome the difficulties traditional traffic management systems confront is driving the development of autonomous traffic regulators. The increased needs of modern transportation and the growing complexity of urban traffic networks have shown the current traffic management methods to be insufficient. 0000-00002-3181-6104 (Dr. Sreelatha); 0000-0002-9144-7021 (BS Mahalaxmi); © 2023 Copyright for this paper by its authors.
Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
CEUR Workshop Proceedings (CEUR-WS.org)
One of the main problems that autonomous traffic regulators seek to address is traffic congestion. In addition to wasting time and fuel, it also causes air pollution and traffic collisions. Traditional traffic management systems rely on time-consuming, ineffective manual interventions to control traffic. Road safety is another problem that autonomous traffic regulators seek to solve. As per the World Health Organization records, road accidents are the ninth main common cause of mortality worldwide and the main reason for the death among young people. Real-time detection and accident prevention are limitations of conventional traffic management systems. To improve traffic safety, autonomous traffic regulators make use of cutting-edge technologies including object identification, weighted assignments to objects, and real-time traffic monitoring. Furthermore, conventional traffic management systems are frequently created for a particular site and are not adaptable enough to accommodate shifting traffic patterns. This may result in an ineffective utilization of the road system, especially during rush hour. By minimizing travel time, fuel use, and emissions, autonomous traffic regulators can also increase the overall effectiveness of the transportation network. Autonomous traffic regulators can save travel times and use less fuel by enhancing traffic flow and minimizing congestion. Additionally, autonomous traffic regulators can lower car emissions by lowering the amount of traffic accidents.
Zaatouri et al. [1] introduces a traffic light control system which accepts the YOLO (You Only Look Once) algorithm for detecting the objects. The system dynamically adjusts traffic light timings by analyzing vehicle and pedestrian presence, aiming to enhance traffic flow and alleviate congestion. By utilizing YOLO's efficiency and accuracy, the proposed system helps to self-adaptive approach for optimizing traffic signal operations.
Liu et al. [2] presents an approach that combines the YOLO network with the anchor box mechanism to improve object detection accuracy and efficiency. Experimental results demonstrate the effectiveness of the proposed method in detecting objects in real-time scenarios. The research contributes to the advancement of object detection techniques by leveraging the capabilities of the YOLO network and introducing the anchor box mechanism.
The authors Pratama B et al. [
Bhave N et al. [
By applying reinforcement learning algorithms, the system learns optimal traffic signal policies for different traffic scenarios. Experimental results from Palladam, India, demonstrate the effectiveness of the proposed approach in reducing traffic congestion and improving overall traffic management by adapting to changing traffic patterns.
Garg et al. [
Kwon J et al. [
Lee et al. [
The paper discusses the application of machine learning algorithms to predict traffic conditions and optimize traffic signal timings for improved traffic flow. The proposed approach aims to enhance the performance and efficiency of autonomous vehicle systems by leveraging machine learning capabilities in traffic control.
Tiwari et al. [
Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
CEUR Workshop Proceedings (CEUR-WS.org)
Lorencik D et al. [
Asha C S et al. [
De Oliveira L F P et al. [
Peiyuan Jiang et al.[
Maqbool S et al.[
The research contributes to the field of intelligent transportation systems by providing an effective solution for vehicle detection and tracking.
Kumari R et al.[
This research suggests a real-time method of traffic signal control which is based on traffic movement. They have the features of the opposing traffic flows at the signalized road crossing thanks to computer vision and machine learning.
You Only Look Once, a cutting-edge real-time item detection system, does this. It is built on deep conventional neural networks (YOLO). Then, traffic signal phases are optimized based on data that has been gathered, namely line length and waiting time per vehicle, to allow the greatest number of vehicles to pass safely with the shortest amount of waiting time. YOLO's accuracy and real-time efficiency made it possible to substitute the policeman in traffic control optimization.
Deep learning is used to create a novel adaptive traffic light control algorithm that complies with safety standards. Realtime detection and vehicle monitoring with duration before exiting the intersection are made feasible by YOLO v2. In fact, the controller uses the YOLO model to determine how many vehicles are in each lane and how long they will wait when the light turns yellow. The duration of the following phase is determined to reduce waiting time based on the maximum and average waiting times for each lane and the length of the queue.
Without disrupting the cycle order, our approach gives preference to those who have waited the longest. See Fig. 1 for a discussion of this algorithm. 0000-00002-3181-6104 (Dr. Sreelatha); 0000-0002-9144-7021 (BS Mahalaxmi); © 2023 Copyright for this paper by its authors.
Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
CEUR Workshop Proceedings (CEUR-WS.org)
We explored and simulated several visual degenerative processes. The excellent results have been produced by Deep learning-based object detection. As there are many numerous problems with issues like photographs when shooting happens in the real world, which includes issues such as noise, blurring, rotational jitter, etc. The advantages of these issues on object detection is important.. In the beginning, they developed the models for degraded photos mostly by applying mathematical models to produce degraded images based on common data sets. They then trained the network to adapt to the challenging real-world environment using these models. Based on the YOLO network developed image degradation model and incorporated conventional image processing techniques to emulate the issues present in real-world shooting, using traffic signs as an example. We examined the impacts of various degradation models on object detection after developing the various degradation models. In order to increase the average precision (AP) of traffic sign detection in real scenarios, we trained a strong model using the YOLO network. In addition to improving the accuracy of object detection, our work has also shown that it is possible to train models that are robust to a variety of visual degradations. This is important for applications such as selfdriving cars, where the ability to detect objects in degraded conditions is essential for safety.
Finally, we enhanced the model's capacity for generalizing complex images. We used the YOLO neural network to assess the traffic signs as our study object. A new picture degradation model was created as a result, using various deteriorated photographs as test sets. After that, they altered the source network and used several degradation techniques to the training set. Then, they used more intricate degradation processes to the training sets to produce an improved and broadly applicable detection network. In conclusion, the model's capacity for generalization had been strengthened, and object detection had become more precise.
Through estimations of traffic density on road layouts, we suggest adaptive traffic signals to regulate their timing. Several different road designs are subjected to image processing to determine the traffic density. Later, the traffic density is used to determine when the traffic signal will turn on. To assess the performance of their suggested method and compare it to a fixed-time traffic light system, the authors created a simulation model. The simulation's findings demonstrated that, in comparison to the fixed-time system, the adaptive traffic signal system was able to decrease the average vehicle waiting time and increase traffic flow. A server that collects data and manages traffic light operations at crossroads is also present. Real-world road conditions are used to validate the whole set of algorithms, including those for calculating traffic density and timing of traffic lights. The results obtained demonstrate the accuracy with which the traffic density sensing system can accurately determine the time of a traffic light. By ensuring that the green light is on for a longer period of time when there is a high volume of traffic on the road and a shorter period of time when there is less traffic, the system was able to lessen congestion.
Canny edge detection algorithm is one of the important used techniques in image processing. 0000-00002-3181-6104 (Dr. Sreelatha); 0000-0002-9144-7021 (BS Mahalaxmi); © 2023 Copyright for this paper by its authors.
Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
CEUR Workshop Proceedings (CEUR-WS.org)
using the Contrast Limited Adaptive Histogram Equalization (CLAHE) equation The image is processed before the edge of the image is determined. By adaptively adjusting the contrast difference, this technique can be utilized to lessen the noise that results from using a camera with low performance or from taking photos at night.
A method of image screening known as bilateral filtering provides a smoothing operation while preserving the image's edge structure. In other words, the image is edge-preserving smoothing via bilateral filtering .The two processes that make up bilateral filtering are selection and filtering. Here, Bilateral filtering is employed in this study to eliminate noise on coloring that was created in the first step. The goal of the selection procedure is to take the surrounding pixels into account. A delimiter function based on the difference in pixel values is the criteria function that is utilized. The filtering procedure itself then applies linear (using kernel box or Gaussian) or nonlinear (median filter) filtering techniques. The range of pixels included in the selection process and the maximum distance that passes the selection process are two parameters for the bilateral filtering algorithm that must be manually defined.
The last method to identify traffic congestion is the binary threshold. This procedure's major goal is to separate the automobiles from the background (road). In order to clearly identify the region that includes the object and backdrop of the image, the binary threshold converts the image to a binary or black-and-white image. The Region of Interest (ROI) of the path, which serves as the observation's focal point, is where the segmentation process is restricted.
By counting the black and white pixels in the ROI, one can determine the traffic density. The following formula is used to determine the traffic density formula:
The use of Reinforcement Learning (RL) and Object Detection to improve traffic flow and reduce congestion is explored. The system uses object detection algorithms to detect and count vehicles at intersections and RL algorithms to determine the optimal signal timings. The signal timings are then adjusted in real-time based on the traffic conditions to reduce wait times and improve traffic flow.Our suggested system is a fully functional model that includes hardware, software, algorithms for object identification and reinforcement learning. Following is a description of how each component like Actions and State-Action Pair work.
The agent's Actions are determined by how the agent perceives the environment. The potential green phase timings of the traffic signal are the activities of our RL agent. These show how many seconds have passed since the green phase began whereas the State-Action pair is a mapping which is associated with Q-values, known as the state space representation. The State space is represented by a Matrix in our implementation. Each cell displays a Qvalue for a possible State and action pair.The authors 0000-00002-3181-6104 (Dr. Sreelatha); 0000-0002-9144-7021 (BS Mahalaxmi); © 2023 Copyright for this paper by its authors.
Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
CEUR Workshop Proceedings (CEUR-WS.org) evaluated the performance of the proposed system using simulations and compared it to a traditional fixed-time signal control system.
The results showed that the proposed system was able to significantly reduce average wait times and improve traffic flow compared to the fixed-time system. In conclusion, the study demonstrates the potential benefits of using RL and object detection in traffic signal control systems to improve traffic flow and reduce congestion. The authors suggest that their proposed system could be implemented in real-world scenarios and further research is needed to validate the results and refine the algorithms. real-time traffic optimization over several road crossings using just the raw pixel input from CCTV cameras. By enhancing traffic flow and decreasing the average amount of time a vehicle spends at an intersection, it is demonstrated that this set of traffic control agents significantly outperforms independently running adaptive signal control systems. In a scenario where each agent only has access to the partial state of the traffic environment, they have shown that a centralized controller is capable of fostering a principled learning strategy between the signal control agents, leading to the positive emergence of cooperative behavior among them.The performance of the proposed system was evaluated using simulations and compared to a traditional fixed-time signal control system.
The results showed that the proposed system was able to significantly reduce average wait times and improve traffic flow compared to the fixed-time system.
A system of numerous, coordinated traffic signal control systems is suggested to be used. It presents a study on a traffic optimization system that uses multi-agent deep reinforcement learning (RL) to control the traffic lights at multiple road intersections. The system uses live camera feeds to detect and count vehicles at each intersection and adjust the signal timings in real-time to optimize traffic flow. The authors used a multi-agent deep RL algorithm to train the system, where each intersection was treated as an independent agent. In this study, multi-agent deep reinforcement learning (DRL) is applied for the first time to 0000-00002-3181-6104 (Dr. Sreelatha); 0000-0002-9144-7021 (BS Mahalaxmi); © 2023 Copyright for this paper by its authors.
Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
CEUR Workshop Proceedings (CEUR-WS.org)
In order to comprehend the flow of traffic and make judgments about traffic control, vehicle counting is a crucial component of traffic management. The current techniques for counting vehicles take a long time, require a lot of work, and are inaccurate. This method locates and recognises automobiles in real-time video footage by using the object detection algorithm YOLO, which is based on deep learning. Then, to precisely count the number of vehicles on the road, the Correlation Filter method is applied. It can be concluded that the YOLO and Correlation Filter algorithms can be used to automate vehicle counting in traffic control systems. The suggested approach works well for precisely counting automobiles and following their movements. Future research should concentrate on enhancing the algorithms' accuracy and integrating the approach with other traffic control systems.
b. Traffic density is calculated by using processing time of the algorithm similarly image need to be acquired which is decided by the green light duration.
the green light starts i.e. the average time required to cross the signal by each class of vehicle. g. The minimum and maximum time limit for the green light duration -to prevent starvation.
When the algorithm is initially run, it sets the default time for the first signal of the first cycle and all following cycles' signals as well as the times for all other signals of the first cycle. The main thread manages the timer of the current signal, and a second thread is initiated to handle vehicle detection for each direction. The detecting threads take a snapshot of the next direction when the current signal's green light timer (or the subsequent green signal's red light timer) reaches zero seconds. The next green signal's timer is set when the result has been parsed. While the main thread is reducing the time remaining on the current green signal's timer, all of this is occurring in the background. As a result, there won't be any latency during the timer's assignment.
The next signal turns green for the duration specified by the algorithm when the current signal's green timer reaches zero.
To improve traffic management, it is possible to specify the average amount of time it takes for each class of vehicle to 0000-00002-3181-6104 (Dr. Sreelatha); 0000-0002-9144-7021 (BS Mahalaxmi); © 2023 Copyright for this paper by its authors.
Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
CEUR Workshop Proceedings (CEUR-WS.org) cross an intersection based on the location, i.e., the region, the city, the locality, or even the intersection itself. For this, information from the relevant transportation authorities can be analyzed. The picture is taken when there are exactly zero seconds till the signal that will turn green next. This allows the system to process the image, count the number of vehicles in each class present in the image, and determine the green signal duration in a total of 5 seconds (equivalent to the value of the yellow signal timer). and set the red signal time for the following signal as well as the times for this signal appropriately. The average speeds of vehicles at startup and their acceleration times were utilized to determine the best green signal time based on the number of vehicles of each class at a signal, and from there, an estimate of the average time each class of vehicle takes to cross an intersection was found. The following formula is then used to get the green signal time. where: ●
vehicles of each class of vehicle at the signal as detected by the vehicle detection module, Green Signal Time(GST) averageTimeOfClass is the average time the vehicles of that class take to cross an intersection, noOfLanes is the number of lanes at the intersection ● ● ●
The vehicle detection module's traffic density data is used by the Signal Switching Algorithm to set the green signal timer and update other lights' red signal timers. Additionally, it cycles through the signals in accordance with the timers. The detection module's information on the vehicles that were picked up by the algorithm, as described in the preceding section, serves as its input. This data is presented in JSON format, with the confidence and coordinates serving as the values and the label of the object being detected as the key. To determine the total number of vehicles in each class, this data is analyzed next. Following this, the signal's green signal time is determined and assigned, and the red signal times of other signals are calculated. To accommodate any number of signals at an intersection, the algorithm can be scaled up or down.
To model actual traffic, Pygame was used to create a simulation from scratch. It helps with system visualization and comparison with the current static system. There are 4 traffic lights at a 4-way intersection there. Each signal has a timer on top that displays the amount of time until it changes from green to yellow, yellow to red, or red to green. The quantity of vehicles that have passed through the intersection is also shown next to each light. There are cars, bikes, buses, trucks, rickshaws, and other vehicles coming from all directions. Some of the vehicles in the rightmost lane turn to cross the intersection to increase the realism of the simulation. When a vehicle is generated, random numbers are also used to determine whether or not it will turn. It also has a timer that shows how much time has passed since the simulation began.
In this study, we examined that Autonomous Traffic Regulator reduced travel times by up to 28%.
A study by the Indian Institute of Technology, Bombay examined that the traffic congestion in India cost the country an estimated $100 billion per year in lost productivity and fuel cost. So if our model is implemented we can save the fuel cost by an estimated figure of $26 billion.
The below graphs help us in understanding the efficiency and effectiveness of our proposed system vs. the traditional automatic traffic light control system that is already in use, by comparing the number of vehicles crossing the signal per second unit of time: 0000-00002-3181-6104 (Dr. Sreelatha); 0000-0002-9144-7021 (BS Mahalaxmi); © 2023 Copyright for this paper by its authors.
Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
CEUR Workshop Proceedings (CEUR-WS.org) 0000-00002-3181-6104 (Dr. Sreelatha); 0000-0002-9144-7021 (BS Mahalaxmi); © 2023 Copyright for this paper by its authors.
Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
CEUR Workshop Proceedings (CEUR-WS.org)
main uses for autonomous traffic regulators. Autonomous traffic regulators monitor and manage traffic flow using real-time data from sensors and algorithms, which helps to ease congestion and improve traffic flow. As a result, the transportation system becomes more effective and less time and fuel are lost due to congestion.
on the road in real time, autonomous traffic regulators also improve road safety. For the purpose of preventing accidents, sensors and algorithms can identify risky driving practices, poor road conditions, and seasonal patterns. Environmental Sustainability: By lowering transportation-related pollutants, autonomous traffic controllers can help support environmental sustainability. Autonomous traffic regulators can optimize traffic flow and ease congestion while cutting down on fuel consumption, which lowers emissions.
Emergency Response: Autonomous traffic controllers can help with emergency response initiatives. Autonomous traffic regulators can modify traffic lights and infrastructure in the case of a natural disaster, auto accident, or other emergency circumstance to enable a smooth flow of emergency vehicles and help the evacuation of impacted areas.
Autonomous traffic regulation using AI has the potential to greatly increase road safety and traffic flow. In order to improve traffic flow, AI algorithms can study traffic patterns, forecast congestion, and dynamically change signal timings. This may lead to shorter travel distances and less fuel use, as well as fewer pollution and accidents. Additionally, To improve road safety by detecting and responding to potential hazards, such as vehicles driving erratically or pedestrians crossing the street illegally AIPowered Traffic management system can be implemented.. However, the implementation of AI-based traffic regulating systems also brings up significant ethical and privacy issues, in addition to technical difficulties such as ensuring the resilience, dependability, and explain ability of AI algorithms. In addition, thorough examination of a number of legal and regulatory issues, such as culpability in the event of accidents or system failures, is necessary for the implementation of autonomous traffic regulation using AI. A lot of money must be invested, and many parties, including communities, businesses, and governments, must work together to integrate AI-based systems with current infrastructure. Despite these difficulties, autonomous traffic management using AI offers a lot of potential for the future of mobility and transportation. Artificial intelligence (AI)based technologies can offer real-time, data-driven solutions to enhance traffic flow and safety on our roads by utilizing the power of machine learning and computer vision. But it's crucial to make sure that the deployment of these technologies is carried out in a morally righteous, accountable, and open way, taking into account the potential risks and advantages for all stakeholders.
It is essential to make sure that autonomous traffic regulators are deployed in a way that is visible, accountable, and beneficial to society as a whole. Additionally, it's crucial to approach the deployment of autonomous traffic regulators holistically, taking into account not only the technological elements but also the social, economic, and political ramifications.
[1] Zaatouri, Khaled & Ezzedine, Tahar. (2018). A SelfAdaptive Traffic Light Control System Based on YOLO. 16-19. 10.1109/IINTEC.2018.8695293. [2] C. Liu, Y. Tao, J. Liang, K. Li and Y. Chen, "Object Detection Based on YOLO Network," 2018 IEEE 4th Information Technology and Mechatronics Engineering Conference (ITOEC), Chongqing, China, 2018, pp. 799803, doi: 10.1109/ITOEC.2018.8740604. 0000-00002-3181-6104 (Dr. Sreelatha); 0000-0002-9144-7021 (BS Mahalaxmi); © 2023 Copyright for this paper by its authors.
Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
CEUR Workshop Proceedings (CEUR-WS.org) .[15] Y. Guo, Q. Gao and F. Pan, "Trained Model Reuse of Autonomous-Driving in Pygame with Deep Reinforcement Learning," 2020 39th Chinese Control Conference (CCC),