The prediction of the signal timing of an isolated intersection using a combination of anfis and population-based optimization algorithm

Due to the rapidly increasing number of vehicles and pedestrian population, traffic congestion has become one of the issues to be addressed specifically in metropolitan areas. An increase in the number of accidents, air pollution, and loss of energy and time due to the queue of traffics are among the most important problems caused by traffic congestion. Urban traffic is dynamic in nature. The purpose of traffic signals is to provide safe travel and decrease the time of traffic delays by increasing the road capacity at crowded intersections. Nowadays, methods to optimize traffic signal timing are based on mathematical models that don't adequately capture traffic dynamics at an intersection. Therefore, it has become inevitable to use adaptive traffic control models as a replacement for low-performance mathematical models in order to overcome the traffic congestion problem at the intersections during peak hours. Istanbul is the largest city in Turkey. In international, commercial, political, academic, social, artistic, and cultural terms, a city hosts numerous tourists annually. Istanbul covers an area of approximately 76,5 million square meters on the European side. It has a 200 million seating capacity together with more than 350 flight points, which are considered to be great advantages on the world lists. Therefore, all the above-mentioned activities are sufficient to show the level of traffic the city is exposed to. The data of the current thesis study was collected from three-phase and four-way intersections in Cumhuriyet road, Yarimburgaz 34303, Kuchukchekmece, Istanbul and converted into a data format that can be used in excel environment. After the volume count related to an intersection was completed, the peak hour standard obtained. High vehicle volume (super-saturated situations) at certain times of the day (morning, noon, and evening) indicates the times of peak hour occurrence. In the current study, the VISSIM simulation model was utilized to obtain high quality, and error-free inputs suitable for the ANFIS architecture. In the VISSIM traffic simulation environment, traffic volume data and other related parameters were entered to model the intersection, and optimum time of the green signal, length of queue of traffic, and traffic delay values were obtained by utilizing the design feature of the program. After the design was completed, we observed that there were improvements in the values obtained, and the saturation degree at the intersection remained below the critical values. The ANFIS-Type1 architecture we designed incorporates three inputs and one output. ANFIS is based on the first order SUGENO-FIS which learns and changes the rules of the system, adaptively. ANFIS training means the calculation of the precursor and resultant parameters using an optimization algorithm. Achieving successful results with ANFIS depends directly on the optimization algorithms that are trained for. In the current study, single-objective (GA, PSO, HS) and multi-objective (MOPSO, NSGAΙΙ) algorithms were used to regulate the parameters and minimize errors during the training period. By analyzing simulations completed by the single-objective algorithms, we discovered that ANFIS-GA resulted in better estimations than ANFIS-PSO and ANFIS-HS. In addition, By analyzing simulations completed by the multi-objective algorithms, it was found out that ANFIS-MOPSO resulted in better estimations than ANFIS-NSGAΙΙ.