An ANFIS Design for Prediction of Future State of a Vehicle in Lane Change Behavior Ali Ghaffari Mechanical Engineering Department Islamic Azad University, South Tehran Branch Tehran, Iran ghaffari@kntu.ac.ir Saeed Arvin Mechanical Engineering Department Islamic Azad University, South Tehran Branch Tehran, Iran me.saeid_arvin@yahoo.com Alireza Khodayari Mechanical Engineering Department K. N. Toosi University of Technology Tehran, Iran khodayari@ieee.org Fatemeh Alimardani Mechatronics Engineering Department K. N. Toosi University of Technology Tehran, Iran f.alimardani@ee.kntu.ac.ir Abstract— The lane change maneuver is one of the most common driving behaviors in highways or rural roads. Each driver usually conducts at least one lane change maneuver during his trip. Therefore, it is chosen as the object behavior in this study and novel adaptive neuro-fuzzy inference models are proposed for this behavior. These models are able to simulate and predict the future behavior of a Driver-Vehicle- Unit in the lane change maneuver for various time delays. Using the field data, the outputs of the models are validated and compared with the real traffic data. The simulation results show that these models have a very close compatibility with the field data and reflect the situation of the traffic flow in a more realistic way. Keywords- adaptive neuro-fuzzy inference system, lane change, Intelligent Transportation System, Driver-Vehicle-Unit I. INTRODUCTION Nowadays, intelligent transportation systems (ITS) play an important role in transportation industry. ITS is a multidisciplinary area with its focus on incorporating up-to- date information technologies of all kinds in the transportation field [1]. These systems help to decrease the traffic flow and increase the safety of the passengers, using computers, telecommunications, and advanced control systems [2]. One category of the ITS sub-systems are microscopic models of traffic flow which lane change behavior models are among the most important ones. The goal of the lane change models is to obtain the desired behavior of a Driver-Vehicle-Unit (DVU) in the lane change process. Fig. 1 shows a typical situation of a lane change maneuver. When the necessity to change the current lane arises, the distances between the main vehicle and other vehicles should be checked before any decision making. If the distances were safe enough to prevent accidents, the lane change maneuver can get started. To perform the maneuver, the vehicle initiates to move to the adjacent lane. By starting to move to the left lane, the heading angle of the vehicle begins to increase until the vehicle gets to the middle of the left lane. At this point, the maneuver is completed and the vehicle can arrange to move in the straight path again. As a result, the heading angle begins to decrease [3]. Fig. 1. Lane change behavior [3]. To develop microscopic traffic simulation of high fidelity, researchers are often interested in imitating human's real driving behavior at a tactical level. Fig. 2 shows the model structure of a DVU in which the detailed driver actions become internal [2] and [4]. Many models for the lane change maneuver have been presented yet [5-6]. Toledo-Moreo and Zamora-Izquierdo presented a lane change prediction model, for collision avoidance in highways, using interactive multiple models (IMM) method. In this model, the GPS and INS systems are used as the sensor unit. This model is able to predict the position and the heading angle of the vehicle during a lane change maneuver [7]. Dogan et. al. applied a back-propagation neural network to develop a lane change model. To train his model, he used the data of real experiments. This model offered a representation of driver’s lane change behavior in order to predict the driver’s intentions as a first step towards a realistic driver model [8]. Seimenis and Fotiades presented a mathematical lane change model by using Clothoidal Theory and Bezier Points. In this study, the lane change trajectory points were approximated using a polynomial that is called s-series. This model can change curvature radius during lane change path regularly using continuous monitoring of centrifugal acceleration through velocity control [9]. Ahle and Soffker presented a lane change model based on the relationships governing the parameters and situations of the operator. In this study, first, various situations of the vehicle and actions of operators (mean braking, driving and lane changing) are defined. Then, a lane change algorithm base on situation-operator mode1 (SOM) was presented [10]. Hsu 2011 IEEE International Conference on Control System, Computing and Engineering 978-1-4577-1642-3/11/$26.00 ©2011 IEEE 156