Design And Implementation Of Intelligent Car Based On FPGA Cloud Acceleration And Its Application In Specific Scenes

With the development of intelligent technology,smart cars are widely used in teaching,automatic driving simulation research,and intelligent vehicle research.At the same time,in the specific scenario such as path planning or obstacle avoidance,smart cars start using computationally intensive algorithms such as neural networks with the development of deep learning.However,the computing resources of the general smart car computing core cannot meet the computing ability requirements of the computationally intensive algorithm.Therefore,in order to solve the problem of insufficient computing resources of smart cars,this thesis proposes a solution to accelerate the computationally intensive problem in smart cars by using FPGA cloud acceleration technology.This thesis also designs and implements a smart car based on FPGA cloud acceleration technology.Based on the related research,this thesis analyzes the functional requirements and non-functional requirements,and proposes a solution to solve the intelligent and computational intensive algorithm using FPGA cloud acceleration technology.And the application scenarios and overall design of smart cars based on FPGA cloud acceleration are proposed.According to the overall design,the hardware and software of smart car based on PYNQ framework and the hardware and application program of FPGA cloud accelerator based on YOLO-v2 deep learning model are designed.And a smart car system based on FPGA cloud acceleration is finally implemented.At the end,the system is tested in an obstacle avoidance scenario based on the COCO data set.The results show that the smart car system based on FPGA cloud acceleration designed and implemented in this thesis can greatly accelerate the calculation of the computationally intensive algorithm of YOLO v2 deep learning model,and meet the requirements.