Decision-making Modeling Of Lane Changing Behavior For Autonomous Vehicles In Urban Road Environment

Recent years, autonomous vehicles gradually become a hot topic, there are many colleges, traditional car companies and internet companies begin to study on it, and they all have developed to a certain level, but there are still many problems need to solve to make the autonomous driving vehicles drive in the real urban roads. This thesis researches the field of the decision-making for autonomous vehicles to make autonomous vehicles drive in the urban roads freely. Contraposing the behavior of lane changing in the urban roads environment, this thesis put forward a lane changing decision-making model based on driver experience.In this thesis, a systematical intuitive decision-making for autonomous vehicle has been put forward to imitate the driver’s decision-making process. The method combines similarity matching, online learning mechanism and prediction together. Similarity matching can make a decision based on previous learned knowledge, while online learning can enrich the knowledge database, and prediction can make the system have reasoning common sense to produce decisions in unfamiliar and incomplete traffic scenarios. However, due to the limited time, this thesis only researches the offline learning part of intuitive decision-making model. For lane changing behavior, this thesis put forward a lane changing decision model for autonomous driving based on the human driving experience. Using rough set and neural network fusion algorithm to extract the human drivers’ lane changing rules, and artificial neural network algorithms to ensure the consistency of rules’ extraction result during the rule extraction process with rough set. After the extraction of rules, a layered library of lane changing rules had been established by using hierarchical state machine method. Through the rules library, human drivers’ experience can be applied to autonomous driving decisionmaking model. Then the co-simulation with Prescan and Simulink/Stateflow were built to simulate the urban road environment and verify the algorithm, the simulation results showed that with this method the autonomous vehicles can change lane safety in the urban traffic and the rules are effective.Meanwhile, in order to verify the feasibility of lane changing decision-making model of autonomous vehicles in real urban road environment, firstly the co-simulation with V-rep and Visual Studio were built to verify the safety of lane changing decision-making model, and then testing the algorithm based on the autonomous vehicle BYD in the third ring road of Beijing. The Experimental results showed that autonomous vehicle can change lane safety in the urban road environment with the decision-making model which is established in this thesis. Finally, the similarity of autonomous vehicle lane changing decision-making model with human driver had been analyzed, and the analysis results showed that the lane changing decision-making model of autonomous vehicle established in this thesis was similar to human driver’s decision, and the offline learning method was effective.