基于AI芯片和机器视觉的协同控制实现

Implementation of Collaborative Control Based on AI Chips and Machine Vision

目前,对于如何实现无人集群在无网络、无GPS环境中的自主导航研究较少。因此,针对这种特殊的应用场景,基于机器视觉相关算法,本文创新性地提出了一套采用集中式编队方式的地空协同控制系统。该系统由空中侦察机、空中跟随机、地面侦察车和地面跟随车组成。空中侦察机采用VINS-MONO技术进行导航,提供位置基准;同时利用CVM-Net技术将空中侦察机与地面侦察车的拍摄图像匹配,得到二者相对位置,用于引导地面跟随车和空中跟随机。该系统采用YOLO-LITE检测算法,可实现实时目标检测;采用华为"麒麟970"智能芯片作为核心硬件,实现信息的实时无回传处理。利用该系统,地面与空中无人集群可在无网络、无GPS的环境进行自主导航定位与协同控制,在紧急搜救和军事侦察等领域具有广阔的应用前景。

Currently,there are few studies on how to realize autonomous navigation of unmanned cluster system in a networkless and GPS-free environment.For this special application scenario,this paper innovatively proposes a ground-air cooperative control system using a centralized formation approach based on machine vision.The system consists of air reconnaissance aircrafts,air-following aircrafts,ground reconnaissance vehicles and ground-following vehicles.The air reconnaissance aircrafts use VINS-MONO to navigate and provide a position reference.At the same time,the CVM-Net is used to match captured images of air reconnaissance aircrafts and ground reconnaissance vehicles to obtain the relative position of them,which is used to guide the ground-following vehicles and air reconnaissance vehicles,and YOLO-LITE detection algorithm is used to realize target real-time detection.The system uses Huawei Kirin 970 as the core hardware,which can realize real-time and no-return processing of information.Through this system,the ground and air unmanned cluster can perform autonomous navigation and coordinated control in the networkless and GPS-free environment,which has broad application prospects in the fields of emergency search and rescue and military investigation.