Research On Intelligent Vehicle Tracking Control Based On Artificial Neural Network

With the continuous development of the automobile industry,a series of problems caused by environmental pollution and traffic accidents have been widely concerned by the society.Traffic safety problems occur frequently.In the closed-loop driving system consisting of road environment-car-driver,one of the main reasons for multiple occurrences is the driver has an accident due to differences in driving experience,physical fitness,physiological state,and various degrees of defects.Facing the increasingly severe situation,the development of new energy vehicles and intelligent driving systems for automobiles has become an urgent requirement of the current automotive industry,and smart cars are the research hotspots and difficulties in recent years.The key points of smart cars are the environment-aware system and the tracking control system.For the latter,the advantages and disadvantages of the vehicle’s path planning and following algorithms play a crucial role in the tracking control system.In this paper,the vertical and horizontal control contents of the tracking control of smart cars at home and abroad are studied,and the control methods of intelligent vehicle path following and obstacle avoidance lane change local path planning are proposed.Firstly,analyze the structure and working principle of vertical control systems and horizontal control systems.Based on the study of vehicle dynamics and vehicle tire model,the vehicle dynamics is decoupled.Based on the vehicle two-degree-of-freedom dynamics model and Newton’s law,the longitudinal control model of the vehicle is constructed based on the steady-state circular motion assumption and optimal premeasure.Secondly,based on the research and analysis of the dynamic model and the preview model,based on the intelligent control theory,combined with neural network and fuzzy control,a trajectory tracking controller is designed.Based on the advantages of selflearning ability of artificial neural network and the deep knowledge of the object,a prediction model of vehicle driving state based on BP(Back Propagation)neural network is designed.On the basis of this,this paper combines the control rules and strong robustness of fuzzy control to design a vehicle driving state control model based on Fuzzy Neural Network(FNN),and predicts the output of the model.As its input,the vertical and horizontal control of the smart car is studied as a whole.Then,the model is built in MATLAB/Simulink,and the simulation environment is set in CarSim.The reliability and adaptability of the designed controller are analyzed and verified by the joint simulation of MATLAB/Simulink and CarSim.Thirdly,based on ISO26262 and SOTIF,this paper studies and analyzes the safety status of smart cars in the tracking process under dynamic environment.Through the information obtained by the vehicle sensor,the positional relationship between the intelligent vehicle and the obstacle in the dynamic environment is analyzed,and the minimum safety index is calculated.After setting the priority order of tracking control targets,based on the traditional sinusoidal lane change curve and double arc curve,the local path based on big data and FNN fuzzy neural network is designed to fit and optimize the obstacle avoidance method.In addition,the validity of the parameters affecting the design of the local path is analyzed and verified.Finally,relying on the student team and the experimental group of the research group,the design and modification of the formula autonomous racing car and the SUV were carried out respectively.The debugging and experiment were carried out on the above,and the research content was verified by calibrating and collecting road information.