Literature Review - CankayaUniversity/ceng-407-408-2021-2022-Smart-Traffic-System GitHub Wiki

  • Öz

Bu günlerde insanlar her yıl yaklaşık 40 saatini doğru kullanılabilecek bir reçel içinde harcıyorlar. Şehirlerde trafik yönetimi çoğunlukla trafik polisi ve yetersiz konfigürasyon nedeniyle araçların uzun süre beklemesine katkıda bulunan sinyal lambaları tarafından yönetilir. Ayrıca, birçok kentsel alanda, trafik sinyal ışıklarının çoğu, verimsiz konfigürasyonun bir nedeni olan sabit bir döngü protokolüne dayanmaktadır. Bu yüzden bu trafik sistemlerini otomatikleştirmeye ve iyileştirmeye ihtiyaç var. Trafik ışıklarının zamanlamaları, trafik sıkışıklığı vb. gibi sorunları çözebilecek ölçeklenebilir bir trafik sistemi geliştirmek için makine öğrenimi, yapay zeka (AI) ve bilgisayarla görmenin karma çalışması istenebilir. ve diğer arabalara göre daha uzun süre kırmızı ışıkta kalan miktarları. Makine öğrenimi, AI ve Vision tabanlı teknolojiler, tüm bu parametreleri kapsama potansiyeline sahiptir. Bu makale, bu teknolojilerin nasıl uygulanabileceği ve karşılaşacağımız zorluklar hakkında bir fikir vermektedir.

  • Abstract

These days, people waste most of their time waiting for traffic which this time could rather be used properly.Most cities all around the world, traffic is mostly just managed by traffic police and signal lights that contribute to the long wait we mostly have because of their inefficient configuration. Also, most of the time the traffic lights use fixed cycle protocol. This is an example for an inefficient configuration as well. That’s why there is a need to automate and improve these traffic systems. The mixed work of machine learning, artificial intelligence and computer vision can be used and integrated in the traffic system to get rid of long hours spent waiting in traffic. In addition, the system will identify the cars and their amount that are stuck under the red light for a longer time than the other cars.

1. Introduction

The main goal of the project that we are making is to change people’s lives for good purposes and make their lives easier. The population grows every single day. Therefore, smart traffic control can be a big problem. To find an answer to this problem, a lot of researchers' groups have done lots of studies. We also have the problem that some green lights occur for a very short amount of time and if there is a traffic jam, the same car might wait in the same red light for a long time. To prevent this problem, we can use computer vision technology to detect the cars that are stuck under the same light more than once and reduce the time of that red light the second time it occurs. By doing this, we give the chance to those cars which are stuck under the same light more than once. Thus, detecting all vehicles (cars, trucks, motorcycles etc.) is a very common problem. In this project, we want to detect the vehicles and also the count of them. By doing this, the system will decide the time of the traffic lights to occur and which road has the most traffic jams in that area. There will be some cameras which will be put in the street and from the top, the cameras will detect the vehicles and then give feedback to the system. Overall, this system will be a system that can identify all crowded or empty streets, vehicles, vehicles’ speeds.

2. Computer Vision and Machine Learning

Computer Vision is the field of computing with technologies that allow computers to identify and process preferences like humans. From dying to being processed, unintended, planning to achieve goals and aiming for them. The image in question may be the result of video, videos from multiple cameras of the same type, acquired with a 3D scanner, or enhanced devices such as an ultrasound device.Computer Vision studies first started in the 1950s. Commercially, it was first used in an application that perceived the difference between printing and handwriting in the 1970s. The real development in the field of Computer Vision has taken place in recent years. Both the rapid increase in the processing power of computers and the widespread use of the internet have resulted in a significant increase in the visual data we have.

Most of the machines we still use don't use image recognition and AI. These days we have self-driving cars that use computer vision which lets us not use the car manually. If they didnt use these technologies, they wouldn't know the differences between a person and a tree. And wouldn't know where they should drive and where they should avoid going. these technologies such as computer vision lets the system identify objects and people and relations of the environment. Computers learn the difference between these objects and people from image datasets, using them as training data and they design a model which helps the computer to predict the outcome. These datasets have huge amounts of data that can help the computer learn and train itself in time. about 2 billion photos are shared by users every day. And these images are used for prediction and training the models that we use to predict. The traffic management system has CCTV’s and sensors that helps the system see, identify and make momentary decisions. These sensors could be used to identify light weight/high weight vehicles and the number of the vehicles that are waiting on the road or just passing by. This knowledge can help us determine the times where there will be a high possibility of traffic.

Over the past several decades, the field of computing developed enormously. Earlier, everything was about logic and mathematical problems. This was an obstacle for solving real life intricacies. This was why we needed improved systems, to learn on its own and take decisions at the real time on its own. Machine learning varies based on the methods used. These methods are categorized as supervised, unsupervised, semi-supervised, reinforcement machine learning algorithms. Supervised machine learning algorithms can be used when we have learned in the past and we apply it to a new data using labeled examples to predict events of the future. Unsupervised machine learning ae mostly used when the information we have informations that are neither classified or labeled for training. Semi supervised machine learning method is mostly chosen when the labeled data needs skilled or relevant resources to train and learn. Reinforcement machine is a method that şnteracts with the environment by producing actions and discovers errors or rewords.

Fig.1 Workflow of Machine Learning

Machine learning is a branch of artificial intelligence that can create machines that can learn from their surroundings, people, and mistakes. Machine learning systems include recommendation systems, autopilot systems for planes, computer-aided translation of one language to another, and maps.

3. Previous Works Of Smart Traffic Systems

The problem of traffic road management has been an active problem for countless years. Many attempts to propose a solution for this problem of intelligent traffic control have been found in the past. For example, in [4], the design of Intelligent Traffic Light Controller Using Embedded System is developed. They utilize simple means to count the quantity of vehicles. In [5], It is proposed to create an intelligent traffic light system to prevent traffic accidents. In [6], an intelligent traffic light control method for crossroads based on extension theory is presented. An Artificial Intelligent (AI) Approach for Intelligent Traffic-Light Control is proposed in [7]. The most common method is to install a traffic signal at each road intersection to control access to the critical section (the junction). Several variants of priority access and critical sections are implemented. The round tape, for example, assigns a constant time for the open road (by turning on the green light), then a fixed time for the transition period (by turning on the yellow light), then closes the road (by turning on the red light), and repeats the process for the next street, and so on. Except in a few developed countries, this is the least efficient method of traffic management; however, it is the most widely used system in the world. In [8] an algorithm for managing the operation of a single traditional traffic light signal for an intersection with four-lane roads is presented, which proposes an adaptation of the traffic to the traffic conditions. Although it is stated that the proposed algorithm is adaptive, it takes into account fixed time periods similar to the study presented in [9]. Similarly, a calendar-based approach for progress reporting is proposed in [10]. [11] presents another study that uses multi-agent communication based on edge computing architecture and IoT for traffic light control. For global traffic sign management, the authors propose a multi-agent reinforcement learning system (MARL). Similar MARL-based studies are presented in [12] and [13]. Other agent-based approaches are given in [7,14]. In [15] to use Q Learning to maximize the number of vehicles crossing intersections. Similarly, [16] proposes deep reinforcement learning. In [17] the approach to the optimization of ant colonies is proposed and in [18] the approach to the optimization of artificial bee colonies is proposed. The social IoV proposed for traffic management [19] is shown in [20]. Several studies are based on the amount of traffic on the roads. In [21], for example, an adaptive algorithm is presented and evaluated. The goal of this study is to use V2V so that each vehicle can estimate the level of traffic congestion and redirect to the least congested route. [22] proposes yet another study based on V2V communication. [23] provides a traffic optimization framework based on vehicle redirection to reduce traffic congestion. Another study in [24] proposes a system based on V2I communications. Furthermore, this study takes into account the protection of incident detection as well as the spread of various types of attacks. Several experiments for the control of floating data-based traffic signs (FCD) are reported in [25]. FCD has also been used in [26] for vehicle tracking data management techniques. [27] proposes the Modular Timed Synchronized Petri Net model for traffic sign management in order to reduce environmental impact. [28] presents the maximum Pareto flow algorithm, and [29] proposes the cell genetics algorithm. Several research articles have been published on the use of cameras to count the number of vehicles for traffic management and optimization. [30] presents a recent study on the use of smart cameras on the Internet. The solution is based on WSN (Wireless Sensor Network) and VANET (Vehicular Ad-hoc Network) by connecting a large number of cameras in a dedicated infrastructure. The video feed from the cameras is routed to centralized servers for processing and extraction of useful traffic data that can be used to check traffic signs. [31, 32] provide additional studies based on the WSN. As shown in [33, 34], several IoT (Internet of Things)-based strategies are proposed. The authors of [35] propose using expert systems and artificial intelligence to process images extracted from camera systems for traffic management. [13] proposes a pheromone-based multi-agent system that is based on cameras and sensors. [36] presents an interesting study that selects the best charging station for electric vehicles based on traffic conditions in order to minimize travel time. [37] proposes a parallel algorithm for synchronizing intersections in large and dense areas, with the goal of improving Bus Rapid Transit based on average speed. [38] presents a similar study based on a hybrid heuristic approach. Finally, the authors of [39] propose that vehicle speed at the intersection be used as an optimization parameter for traffic light control.

Fig.2 RECENT ACADEMIC STUDIES IN VEHICLE DETECTION AND TRAFFIC LIGHT SYSTEM

4. Proposed System and Challenges

To make a smart traffic system, we are building a new system using artificial intelligence and CCTV(Camera-Real time video) image processing together. Cameras mounted on every place that have traffic lights record thousands of hours of video daily, which contain very useful information and can be used to reduce the overcrowding on the streets. With the help of the video footage data, data set of traffic videos and machine learning algorithms, it is possible to manage traffic light duration. Pattern recognition can be used to identify the vehicle type, count number of vehicles, and categorize lightweight and heavy vehicles. This would improve the system to control the traffic lights to decrease the traffic. The machine learning algorithm can perform better as the training data sets contain enough data to predict. These data sets are growing every day which can help our system to give suggestions about improving roads like “Build a bridge here” etc. Sometimes during the overcrowding situations, managing traffic light duration is not enough. So we are building a suggestion AI for improving roads . Applying machines to do this job can be very effective and reliable. The proposed system will be a synchronised intelligent traffic light which will improve overcrowding. The proposed model will be composed of two components: a monitoring system and a control system. Both components are able to integrate together and use data of the same directional backward signals simultaneously to make intelligent decisions efficiently. Figure shows the proposed model architecture.

Fig. 2 Proposed Model

5. Conclusion

Simply put, it is possible to create a system capable of effectively solving one of our daily urban problems by combining the main developing IT fields. The model proposed here should perform better over time with help of the AI, because as the amount of data collected increases, it will be possible to make better predictions. Therefore, a large and reliable data set will contribute to efficient estimates. Moreover,if it is not enough to collect data, you must be able to validate, connect and access data dynamically to extract meaningful insights and implement it in real time, which will help forecast traffic trends and effectively regulate traffic. .Thus, the most useful and powerful technology nowadays, machine learning, AI and computer vision will help us to automate our traffic systems. 6. References

  1. Emad Issa Abdul Kaream and Aman Jantan, “Intelligent Traffic Light Control Using Neural Network with Multi-Connect Architecture”, National Conference on Information Retrieval and Knowledge Management (CAMP08) in Kuala Lumpur,Malaysia, 2008.
  2. Emad Issa Abdul Kaream, k. N. M., Hussein A. Moussa, “Gray Image Recognition Using Hopfield Neural Network With Multi- Bitplane and Multi-Connect Architecture”, Proceedings of the international Conference on Computer Graphics, Imaging and Visualisation (CGIV'06) IEEE, 2006.
  3. S. S. Chavan, R. S. Deshpande, and J. G. Rana, “Design of intelligent traffic light controller using embedded system,” in Proceedings of the 2009 Second International Conference on Emerging Trends in Engineering & Technology, pp. 1086–1091, Maharashtra, India, December 2009.
  4. K. Matsuzaki, M. Nitta, and K. Kato, “Development of an intelligent traffic light for reducing traffic accidents,” in Proceedings of the 2008 International Conference on Control, Automation and Systems, pp. 443–447, Hanoi, Vietnam, December 2008.
  5. K.-H. Chao, R.-H. Lee, and K.-L. Yen, “An intelligent traffic light control method based on extension theory for crossroads,” in Proceedings of the 2008 International Conference on Machine Learning and Cybernetics, pp. 1882–1887, Kunming, China, July 2008.
  6. V. Hirankitti and J. Krohkaew, “An agent approach for intelligent traffic-light control,” in Proceedings of the First Asia International Conference on Modelling & Simulation (AMS’07), pp. 496–501, Phyket, Thailand, March 2007.
  7. M.-D. Pop, “Traffic lights management using optimization tool,” Procedia-Social and Behavioral Sciences, vol. 238, pp. 323–330, 2018.
  8. H. Joo, S. H. Ahmed, and Y. Lim, “Traffic signal control for smart cities using reinforcement learning,” Computer Communications, vol. 154, no. 3, pp. 324–330, 2020.
  9. M. Radivojević, M. Tanasković, and Z. Stević, “The adaptive algorithm of a four way intersection regulated by traffic lights with four phases within a cycle,” Expert Systems With Applications, vol. 166, p. 2021, 2019.
  10. M. Drapalyuk, S. Dorokhin, and A. Artemov, “Estimation of efficiency of different traffic management methods in isolated area,” Transportation Research Procedia, vol. 50, pp. 106–112, 2020.
  11. A. Yousef, A. Shatnawi, and M. Latayfeh, “Intelligent traffic light scheduling technique using calendar-based history information,” Future Generation Computer Systems, vol. 91, pp. 124–135, 2019.
  12. Q. Wu, J. Wu, J. Shen, B. Yong, and Q. Zhou, “An edge based multi-agent auto communication method for traffic light control,” Sensors, vol. 20, no. 15, pp. 4291–4316, 2020.
  13. T. Wu, P. Zhou, K. Liu et al., “Multi-agent deep reinforcement learning for urban traffic light control in vehicular networks,” IEEE Transactions on Vehicular Technology, vol. 69, no. 8, pp. 8243–8256, 2020.
  14. K. L. Soon, J. M.-Y. Lim, and R. Parthiban, “Coordinated traffic light control in cooperative green vehicle routing for pheromone-based multi-agent systems,” Applied Soft Computing, vol. 81, Article ID 105486, 2019.
  15. N. Kumar and S. S. Rahman, “Deep reinforcement learning with vehicle heterogeneity based traffic light control for intelligent transportation system,” in Proceedings of the 2019 IEEE International Conference on Industrial Internet (ICII), pp. 28–33, Orlando, FL, USA, November 2019.
  16. P. M. Kumar, U. Devi Gandhi, G. Manogaran, R. Sundarasekar, N. Chilamkurti, and R. Varatharajan, “Ant colony optimization algorithm with internet of vehicles for intelligent traffic control system,” Computer Networks, vol. 144, pp. 154–162, 2018.
  17. K. Gao, Y. Yicheng Zhang, A. Sadollah, and R. Rong Su, “Improved artificial bee colony algorithm for solving urban traffic light scheduling problem,” in Proceedings of the 2017 IEEE Congress on Evolutionary Computation (CEC), pp. 395–402, San Sebastián, Spain, June 2017.
  18. K. M. Alam, M. Saini, and A. E. Saddik, “Toward social internet of vehicles: concept, architecture, and applications,” IEEE Access, vol. 3, pp. 343–357, 2015.
  19. M. S. Roopa, S. Ayesha Siddiq, R. Buyya, K. R. Venugopal, S. S. Iyengar, and L. M. Patnaik, “Dynamic management of traffic signals through social IoT,” Procedia Computer Science, vol. 171, pp. 1908–1916, 2019.
  20. T. S. Gomides, R. E. De Grande, A. M. D. Souza, F. S. H. Souza, L. A. Villas, and D. L. Guidoni, “An adaptive and distributed traffic management system using vehicular ad-hoc networks,” Computer Communications, vol. 159, pp. 317–330, 2020.
  21. A. Bazzi, A. Zanella, and B. M. Masini, “A distributed virtual traffic light algorithm exploiting short range V2V communications,” Ad Hoc Networks, vol. 49, pp. 42–57, 2016.
  22. Z. Cao, S. Jiang, J. Zhang, and H. Guo, “A unified framework for vehicle rerouting and traffic light control to reduce traffic congestion,” IEEE Transactions on Intelligent Transportation Systems, vol. 18, no. 7, pp. 1958–1973, 2017.
  23. V.-T. Ta and A. Dvir, “A secure road traffic congestion detection and notification concept based on V2I communications,” Vehicular Communications, vol. 25, Article ID 100283, 2020.
  24. V. Astarita, V. P. Giofrè, G. Guido, and A. Vitale, “A review of traffic signal control methods and experiments based on floating car data (FCD),” Procedia Computer Science, vol. 175, pp. 745–751, 2020.
  25. S. Brakatsoulas, D. Pfoser, and N. Tryfona, “Practical data management techniques for vehicle tracking data,” in Proceedings of the 21st International Conference on Data Engineering (ICDE’05), pp. 324-325, Tokyo, Japan, April 2005.
  26. H. Lamghari Elidrissi, A. Nait-Sidi-Moh, and A. Tajer, “Knapsack problem-based control approach for traffic signal management at urban intersections: increasing smooth traffic flows and reducing environmental impact,” Ecological Complexity, vol. 44, Article ID 100878, 2020.
  27. S. Kamishetty, S. Vadlamannati, and P. Paruchuri, “Towards a better management of urban traffic pollution using a pareto max flow approach,” Transportation Research Part D: Transport and Environment, vol. 79, Article ID 102194, 2020.
  28. A. Villagra, E. Alba, and G. Luque, “A better understanding on traffic light scheduling: new cellular GAs and new in-depth analysis of solutions,” Journal of Computational Science, vol. 41, Article ID 101085, 2020.
  29. W. C. Tchuitcheu, C. Bobda, and M. J. H. Pantho, “Internet of smart-cameras for traffic lights optimization in smart cities,” Internet of Things, vol. 11, Article ID 100207, 2020.
  30. G. Padmavathi, D. Shanmugapriya, and M. Kalaivani, “A study on vehicle detection and tracking using wireless sensor networks,” Wireless Sensor Network, vol. 2, no. 2, pp. 173–185, 2010.
  31. B. Zhou, J. Cao, and H. Wu, “Adaptive traffic light control of multiple intersections in WSN-based ITS,” in Proceedings of the 2011 IEEE 73rd Vehicular Technology Conference (VTC Spring), Budapest, Hungary, May 2011.
  32. L. F. P. Oliveira, L. T. Manera, and P. D. G. Luz, “Smart traffic light controller system,” in Proceedings of the 2019 Sixth International Conference on Internet of Things: Systems, Management and Security (IOTSMS), pp. 155–160, Granada, Spain, October 2019.
  33. I. M. Albatish and S. S. Abu-Naser, “Modeling and controlling smart traffic light system using a rule based system,” in Proceedings of the 2019 International Conference on Promising Electronic Technologies (ICPET), pp. 55–60, Gaza, Palestinian Territories, October 2019.
  34. P. Barbecho Bautista, L. Lemus Cárdenas, L. Urquiza Aguiar, and M. Aguilar Igartua, “A traffic-aware electric vehicle charging management system for smart cities,” Vehicular Communications, vol. 20, Article ID 100188, 2019.
  35. S. Nesmachnow, R. Massobrio, E. Arreche et al., “Traffic lights synchronization for Bus Rapid Transit using a parallel evolutionary algorithm,” International Journal of Transportation Science and Technology, vol. 8, no. 1, pp. 53–67, 2019.
  36. X. Cabezas, S. García, and S. D. Salas, “A hybrid heuristic approach for traffic light synchronization based on the MAXBAND,” Soft Computing Letters, vol. 1, Article ID 100001, 2019.
  37. J. Li, Y. Zhang, and Y. Chen, “A self-adaptive traffic light control system based on speed of vehicles,” in Proceedings of the 2016 IEEE International Conference on Software Quality, Reliability and Security Companion (QRS-C), pp. 382–388, Vienna, Austria, August 2016.
  38. Lin Zhang, Junhao Huang, Xiyuan Li, Lu Xiong,“Vision-based Parking-slot Detection: A DCNN-based Approach and A Large-scale Benchmark Dataset”, 2017 IEEE Transactions on Image Processing, vol. 27, no 11, pp. 5350-5364, Nov. 2018
  39. Nilakorn Seenouvong, Ukrit Watchareeruetai, Chaiwat Nuthong, Khamphong Khongsomboon, “A computer vision based vehicle detection and counting system”, 2016 8th International Conference on Knowledge and Smart Technology (KST), IEEE, 2016.
  40. N.V. Chawla, K.W. Bowyer, L.O. Hall, and W.P. Kegelmeyer, “SMOTE: Synthetic minority over-sampling technique,” J. Artificial Intell. Research, vol. 16, no. 1, pp. 321–357, Jan. 2002.
  41. T. Kong, A. Yao, Y. Chen, and F. Sun, “Hypernet: Towards accurate region proposal generation and joint object detection,” in Proc. IEEE Int. Conf. Comput. Vis. Pattern Recognit., Jun. 2016, pp. 845–853.
  42. C. Harris and M. Stephens, “A combined corner and edge detector,” in Proc. 4th Alvey Vis. Conf., Sep. 1988, pp. 147–151.
  43. N. Dalal and B. Triggs, “Histogram of oriented gradients for human detection,” in Proc. IEEE Int. Conf. Comput. Vis. Pattern Recognit., Jun. 2005, pp. 886–893.
  44. J. Redmon, S. Divvala, R. Girshick, and A. Farhadi, “You only look once: Unified, real-time object detection,” in Proc. IEEE Int. Conf. Comput. Vis. Pattern Recognit., Jun. 2016, pp. 779–788.
  45. W. Liu, D. Anguelov, D. Erhan, C. Szegedy, S. Reed, C. Yu, and A.C. Berg, “SSD: Single shot multibox detector,” in Proc. Euro. Conf. Comput. Vis., Oct. 2016, pp. 21–37.
  46. J. Redmon and A. Farhadi, “Yolo9000: Better, faster, stronger,” in Proc. IEEE Int. Conf. Comput. Vis. Pattern Recognit., Jul. 2017.