The project belongs to the science and technology field: Computer science and technology/Computer application technology/Computer decision support system Main technical innovation content: Road traffic safety issues have always been the focus of social attention, and research on advanced safety technologies related to them has received increasing attention. Based on the conceptual system of human-vehicle-road-environment collaboration, this project has established a logical system with information flow as the main line, and built a vehicle-road collaborative satellite positioning information system consisting of a mobile access layer, a public transmission layer and a control center. Including modules such as cloud and cloud hybrid fusion technology, vehicle-road collaboration and multiple information interaction, BP neural network-based safe speed warning, and RNN-based pedestrian identification. The system can dynamically collect vehicle status information and driving environment information, and realize proactive safety control and danger warning of vehicles through deep learning and dynamic decision-making to ensure the driving safety of public transportation vehicles and avoid traffic accidents. 1) Cloud and cloud hybrid fusion technology combines fog computing and vehicle networking technology to design an information collection architecture with low latency and high reliability in a fog computing environment. This architecture mainly takes the urban road environment as an application scenario and consists of vehicles, roadside units (RSU), fog equipment and cloud computing data centers. The vehicle is wirelessly connected to the RSUs deployed on both sides of the road. The RSUs transmit the collected vehicle information to nearby fog equipment through a wired connection. The fog equipment temporarily stores and calculates the received information, and transmits the information to the Cloud Virtual Machine. 2) Vehicle-road collaboration multi-information interaction Starting from the aspects of channel access protocols and multiple communication mode fusion, combined with the characteristics of the transportation system, the method to optimize the information interaction performance of the vehicle-road collaboration system is studied. By analyzing different MAC layer handshake mechanisms and typical backoff algorithms, and combining the traffic characteristics of vehicle-road collaborative systems with the characteristics of IEEE802.11p, an adaptive channel access protocol is proposed. Integrate vehicle density and data retransmission times into the optimized handshake mechanism and adaptive adjustment backoff algorithm to reduce the possibility of transmission channel collisions between nodes and improve network performance. 3) Safe speed warning based on BP neural network By analyzing various factors affecting the safe speed of public transportation and combining deep learning algorithms, a safe speed warning model of public transportation based on BP neural network is established. Comprehensively considering the main factors affecting the safe speed of public transportation vehicles, safe speeds under different weather conditions, different lights, different time periods, different driving experiences, and different road congestion levels can be obtained, which can provide data support for the judgment and early warning of dangerous conditions such as lane deviations and vehicles receiving vehicles too close. 4) RNN-based pedestrian recognition uses vehicle-mounted cameras and vehicle-mounted ultrasonic/lidar technology to detect pedestrian motion trajectory information, uses D-S evidence reasoning algorithm to preprocess the information, and then uses information acquisition delay, satellite positioning error and road surface adhesion coefficient As input factors, the position of pedestrian targets is accurately and quickly estimated through multi-source mixed error compensation in the vehicle-road collaborative environment of RNN neural network. Intellectual property rights: A total of 5 patents have been produced so far, including 2 authorized invention patents: the control method of a GPS positioning vehicle-mounted recorder and its system control unit, and a vehicle-road collaborative satellite positioning information system. 5 computer software copyrights are registered: Yaowei Car Butler Software V1.1, Yaowei Satellite Positioning Monitoring Software V1.0, Yaowei Smart Bus Operation and Scheduling Software V1.0, Yaowei Official Vehicle Dispatch and Management Software V1.0, Yaowei Official Vehicle Safety Management Information Platform Software V1.0. Application and promotion status: As of March 2019, this project has been directly applied to customers such as Shanghai Jiading Public Transportation Company, Feng County Public Transportation Company, Jiangsu Province, Shanghai Pudong New District Diplomatic Relations and Relations Commission, Shanghai City Government Affairs Administration Bureau and subordinate district-level Government Technical Affairs Administration Bureaus, achieving a total sales of 300 million yuan. In order to integrate into the construction of smart cities, the vehicle-road collaborative satellite positioning informatization system of this project can seamlessly connect with existing informatization systems and gradually improve the informatization, dataization and intelligence level of the entire public transportation system.