Authors: S.Likhitha Priya, Assistant Professor B.Narsimha
Abstract: The transportation networks, economy, and quality of life in cities throughout the world are all negatively impacted by the pervasive problem of traffic congestion. Congestion relief is of critical importance in India since the country's fast urbanization and population increase are making the situation worse. By integrating user and expert perspectives and using a data-driven approach to prepare a knowledge base for traffic congestion, this thesis presents a holistic strategy to address road traffic congestion. The system is tailored to the lndian scenario in the context of inter-urban highways. One module of this thesis is based on user survey methodology, while the other is centered around applying deep learning techniques to analyze the dynamics of congestion. Several facets of congestion prediction are covered in this thesis's objectives. Through observation and feedback, it seeks to understand the causes that contribute to congestion and identify its characteristics from the perspective of travelers. To achieve this goal, we analyzed data from thirteen separate sites along the undivided interurban highway that were selected based on their mixed land use pattern and where congestion is an issue. The survey questionnaire yielded 206 responses, which constitute the data set. The Mann Whitney U test was used to see if there is any heterogeneity between the two traveler groups, drivers and passengers. This goal assesses the features of congestion using the Relative Impoftancc Index (RII) and also scores their preference for measures to alleviate it, in addition to identifying the attributes of congestion. Additionally, it suggests legislative recommendations based on the opinions of travelers in an effort to enhance their traveling experience on interurban highways. The second aim of the thesis uses exploratory factor analysis on the characteristics derived from the first objective and empirical knowledge to investigate the interrelationships among the congestion factors. This objectivc has sampled data from 282 replies and 5 experts. The professionals at Thcsc come from a variety of backgrounds in the transportation industry. After conducting an exploratory factual analysis on the traffic congestion attributes identified in objective one using interpretive knowledge based on the experiments, the fiuzy based total interpretive structural modeling technique (Fwzy TISM) was used to model the interrelationships between the factors. This technique provides both direct and indirect links of cause and effect between the factors. Micmac analysis has been used to determine the independent, dependent, linkage, and autonomous components in the connection diagram that was derived via Fuzzy TISM. Eight important factors contributing to traffic jams on interurban highways were identified through exploratory factors analysis. These factors include inadequate road geometry (Cl), environmental factors (C2), external events (C3), inefficient public transportation (C5), driving behavior (C6), special events (C7), and regional economic dynamics (C8). Using the r- uzzy TISM technique, we may partition the variables into different levels of the hierarchical connection diagram. At the Level I factor level, we have C3, C4, and C6. The second Lcvel component is C5, the third is Cl and CT, and the fourth is C2 and C8. Based on the results of the fuzzy Micmac analysis, we can deduce that factors such as Regional Economic Dynamics (C8), Special Events (C7), and Environmental Factors (C2) are independent variables that impact most of the factors, despite having less driving power. The occurrence of these variables also affects the linkage factor. In this study, archaic traffic management (C4) is a linkage factor that has a high driving power and a high dependence power owing to its unstable proper. It impacts and is affected by other factors such as road geometry dynamics (C3), logistical events (C4), inefficient public transportation facilities (C5), and driving behavior (C6). This thesis's first module is complete with the first two objectives. Objectives three and four are located in the second root of the thcsis plant. The third objective of the project is to automate the process of traffic volume determination utilizing modern dccp lcarning dctcctors and trackers. This would enable the collecting of real-time data for congestion prediction. This objective has been tested on a bespoke dataset consisting of 10,000 virtual reality images from India. The photos in this dataset depict a variety of highway situations with several automobiles. A model for vehicle categorization and traffic volume counting has been developed. It has been trained using YOI-O variants, a single stage object identification methodology, and Byte trackcr, a counting method. The results of the traffic volume counter are examined on the intersection scenes and mid blocks of the intcr urban highways on videos ranging from one minute to five minutes. The short videos achieve an accuracy of 98 degrees of freedom, while the long videos reach an accuracy of up to 87 degrees of freedom. In order to build a comprehensive knowledge base for predicting congestion states, the fourth objective of the study suggests a framework for a knowledge base prediction system regarding road congestion. This framework integrates the results of objectives one and three by combining the impedance factors identified in objective one with traffic flow variables obtained from objective three. The 84-hour data sample includes both weekdays and weekends, and this target has been executed on that basis. The ANFIS (Adaptive Network based fuzzy inferencc system) technique, which is based on several membership functions, has been used to analyze and generate the if-then rules for decision makers. It demonstrates that the existence of rnultiple impedance factors is responsible for the differences in congestion levels on weekdays and weekends. Finally, to improve the efficacy and precision of congestion prediction, the proposed knowledge-based basc road traffic system makes use of both uscr-drivcn insights and cxpert knowlcdgc. By combining user feedback with expert analysis and automated data collection techniques, the system provides a comprehensive view of traffic congestion and how to predict it. In the Indian context, where varied traffic situations and intricate urban landscapes call for precise prediction systems, this comprehensive methodology is of utmost importance. The system is able to produce strong and trustworthy predictions because it uses a combination of fuzzy logic and deep learning techniques, which allow it to deal with the inherent uncertainty and complexity of real-world traffic scenarios. Additionally, the system may enhance its predictive capabilities through the incorporation of user feedback and cxpcrt insights, guaranteeing its relevance and efficacy in ever-changing traffic scenarios. Finally, as a solution to the widespread issue of traffic congestion in India, the proposed knowledge-based road traffic congestion prediction system brings hope. The system offers a comprehensive framework for understanding, analysing, and forecasting congestion lcvcls by combining USCr perspectives, cxpcrt knowledge, and sophisticated technology. Improving traffic management strategies, reducing congestion, and making urban transportation systems in India more efficient might all be possible with the use of this system.
