A Research And Implementation Of Cloud-controlled Autonomous Driving System Used In The Constrained Scene

In recent years,with the rise of artificial intelligence,autonomous driving technology is developing rapidly.The current mainstream of vehicle autonomous driving systems is expected to achieve autonomous driving in general road scene through vehicle-mounted equipment and computing platform,which are not only costly but also difficult to put into practical use.While driving conditions can be clearly defined under constrained scene with less uncertainty,it is suitable for bringing autonomous driving technology into practical.Besides,with the further development of Internet of Things(IoT)technology,the autonomous driving technology gradually develops towards network connected.Taking advantage of cloud computing,the network connected autonomous driving system can reduce vehicle-mounted computing cost,and lays the fast-iterative system foundation for the big-data processing in machine learning methods.Based on the constrained scene with less uncertainty,this thesis study on a cloud-controlled autonomous driving system,which realizes the practical use of autonomous driving with low cost.Drawn from the IoT system architecture,the proposed system includes:network connected vehicle system,vehicle cloud-controlled system,application systems,and blockchain-based service and data sharing platform.Among them,the main contents of this thesis are focus on the network connected vehicle system and the vehicle cloud-controlled system.As a data source and execution terminal,the network connected vehicle system is responsible for real-time control,perception and data acquisition tasks,while the vehicle cloud-controlled system is responsible for non-real-time computing,vehicle monitoring of data storage tasks.The two systems provide autonomous driving interface for the application system to extend the application area of autonomous driving,and manages the resource and executing tasks of the network connected vehicles.The innovation points of this thesis are as follows:(1)cloud-controlled autonomous driving is realized and provided for the application system as a a basic service.Receiving the requests from the application system,the vehicle cloud-controlled system makes trajectory planning and continuous monitoring for target vehicle.Based on the received trajectory high-precision localization,the target vehicle executes trajectory tracking to realize autonomous driving.(2)Achieved efficient autonomous driving data collection and storage for future algorithm simulation and rapid iteration,including logs,videos and various sensor dataIn order to verify the validity of the systems and methods,the experiments of cloud-controlled autonomous driving have been carried out in the simulation environment and the constrained scene.The results show that the systems described in this thesis have practical value.