Research And Implementation Of Cloud Platform Of Autonomous Driving Based On B/s Architecture

Autonomous driving has become a research hotspot in the automotive field.Due to the limitation of national laws and other reasons,autonomous driving simulation technology is more and more favored by researchers and enterprises because it can save a lot of human and material costs.Recently,with the development of cloud computing technology,the cloud platform for autonomous driving is becoming popular.Uber and other foreign driverless research teams have begun to develop a web browser based cloud platform for autonomous driving.Domestic universities and enterprises engaged in autonomous driving are still entering this fields.However,most of these cloud simulation techniques are in the initial stage.Therefore,it is necessary to study open source and controllable cloud platform based on browsers.The main work of this paper is as follows:(1)We developed a cloud platform for autonomous driving simulation based on Browser/server(B/S)architecture.First,the front-end human-computer interaction interface based on browser is developed,the back-end server supporting high concurrency is built,and several AI modules for automatic driving are integrated.Relying on the system,users can initially complete the training or remotely execute trained models of AI algorithms for autonomous driving.(2)The browser interface with human-computer interaction and web server based on HTML5 and Web GL technologies were built.The interface allows users to upload a variety of3 D simulation models and set corresponding parameters,including cars,roads,obstacles,etc.,and build common scenes of cars in the process of driving,including: single car overtaking,avoiding the car behind,waiting for pedestrians,etc.Based on the above functions,the interface allows users to upload AI codes and models for remote cloud training and remote execution.The interface model supports the access to the background server database,and develops the user registration and login functions.The background cloud server adopts thread pool,non blocking socket and other technologies,which can support high concurrent multi-user requests..(3)The sensing module of laser point cloud and path planning module of the autopilot cloud platform are integrated to verify the effectiveness and reliability of the simulation platform.Based on Python and Pytorch technology,the laser point cloud sensing module builds a point cloud sensing server,and uses the classic "second" target detection algorithm to test the feasibility of the point cloud sensing module,and completes the target detection of cars,pedestrians and bicycles.The latter is based on node.js technology to build a server for path planning,and uses the classical "graph neural search" algorithm to test the feasibility of the path planning module,and finally realizes the automatic driving of the car in the simulation cloud platform in the browser.