Design And Implementation Of Aircraft Routing System Based On Heterogeneous Multicore Embedded Platform

Due to the various drawbacks of traditional embedded devices and obstacle avoidance algorithms in the field of obstacle avoidance of drones,it is not suitable for various intelligent path-finding methods.Then computer vision played an irreplaceable role.Due to the large amount of computer vision computation,it depends on the processing performance of the embedded computing unit.In this regard,FPGA,GPU,and DSP computing platforms have their advantages and disadvantages.This article weighs the pros and cons of various processing platforms,selects the ARM + GPU platform as a real-time computing platform.The main tasks of this paper are as follows.:1.Binocular stereo vision algorithm is transplanted to the ARM + GPU heterogeneous processing platform.The NEON and CUDA technology are used to optimize the algorithm so that it can process binocular images collected by binocular cameras in real time and generate 3D point cloud images.2.Designing an obstacle extraction algorithm to efficiently extract obstacle information from the 3D point cloud image generated in the first step and convert it into a data format that the path planning algorithm can handle.3.Paper designs and optimizes the dynamic window approach(DWA).DWA use the obstacle information generated in the second step,calculate the flight path of the aircraft,and generate control commands to control the flight of the aircraft.4.Paper using the SDK provided by the aircraft to control the flight of the aircraft in real time for the test environment.It receives the flight control command generated in the third step.5.Design a reasonable system framework,integrate the algorithms in steps 1-4 into the framework,and make the overall framework highly modular.Each step can be independently updated.6.The entire routing system to be built is tested in real environments.Ensure the correctness and robustness of the system.7.The binocular image captured during the navigation of the aircraft is transmitted through the network to the PC,and an open-source SLAM algorithm is used on the PC to perform real-time 3D reconstruction of the flight environment.The system implements the above functions and algorithms in the embedded devices on the aircraft,so that the aircraft can reconstruct the three-dimensional map in the flight environment in real time(20 fps),and can avoid the obstacles in real time in an unknown environment.At the same time,it is also possible to communicate with the ground station during the flight process,send the image to the ground station,and use the high-speed computing equipment of the ground station to do other things,like SLAM.Although this article only implements the aircraft obstacle avoidance function in the end,the three-dimensional map generated during the flight can be used in more scenes.