Research On Intelligent Vehicle Forward Guidance Area Planning Based On Prediction Model

The end-to-end intelligent vehicle is a kind of intelligent car directly controlled by artificial intelligence algorithm,which can directly output the control command of the vehicle according to the data input of the sensor.When designing the driving model of an end-to-end intelligent vehicle,engineers no longer need to explicitly define the various scenes that appear during the driving process.They only need to use the human driving data set to train the artificial intelligence algorithm model and the driving model be capable of driving a car.Different from the previous intelligent vehicle path planning,the research of this paper is aimed at the end-to-end intelligent vehicle,planning the forward guidance area for the end-toend intelligent car and guiding the driving direction of the vehicle.According to the characteristics of the end-to-end driving model,this paper firstly designs the overall control framework for end-to-end intelligent vehicle applications with the end-to-end driving module and the assisting driving module as the core.The framework has the characteristics of portability and scalability.The forward guidance area planning sub-module which is a part of the auxiliary driving module is responsible for converting the information of the upper navigation system into a control language acceptable to the end-to-end driving model,and is a bridge for communication between the upper navigation system and the end-to-end driving model.In this paper,the nonlinear intelligent vehicle model is directly discretized,and the Euler method,the second-order RK method and the fourth-order RK method are compared in terms of numerical stability,precision and efficiency.And the second-order RK method is used to construct the discrete intelligent vehicle prediction model.After obtaining the discrete intelligent vehicle prediction model,the constraints are added to the model and the planning target route consisting of the transition route and the target route is constructed.Based on this,the objective function with the adaptive function of the working condition is constructed.The NLP problem will be solved by SQP algorithm.After the discrete solution results are obtained,the interpolation with a derivative method is used to make the planning results can be expressed explicitly by mathematical functions,and the result will be extended to both ends perpendicular to the driving direction of the intelligent vehicle and the forward guidance area is created.In order to verify the effectiveness of the designed algorithm,this paper builds a hardwarein-the-loop platform based on PreScan and ROS,designs the software architecture,and codes the algorithm with C++.Then,we tested three typical scenarios,the good results show that the algorithm described in this paper is feasible.