基于视觉反馈和行为树的无人机具身智能控制算法  

A Visual Feedback and Behavior Tree-Based Embodied Intelligence Control Algorithm for UAVs

作　　者：李增艺 LI Zengyi(China Railway 12th Bureau Group Railway Maintenance Engineering Co.Ltd.,Chengdu Sichuan 61000,China)

机构地区：[1]中铁十二局集团铁路养护工程有限公司,四川成都610000

出　　处：《铁道建筑技术》2025年第4期214-218,共5页Railway Construction Technology

基　　金：中铁十二局集团有限公司科技研发计划项目(集团2024研34号)。

摘　　要：提出一种基于视觉反馈和行为树控制的无人机具身智能运算方法,旨在提升无人机在地铁隧道等复杂环境中的自主性和智能性。该方法通过优化大规模语言模型将用户口令转化为机器指令,利用集成学习的视觉目标检测算法识别潜在风险目标,并设计行为树节点及体系架构以实现智能决策和行为规划。通过机器人操作系统(ROS)实现模块间信息交互,本套控制算法系统能够使无人机在不预设航点的情况下,有效理解用户指令,智能检测巡航过程的风险点并自主执行绕飞等无人机任务,有效应对动态环境和复杂任务,促进实现无人机具身智能应用。This study proposes a UAV embodied intelligence control algorithm based on visual feedback and behavior tree control.This algorithm aims to enhance the autonomy and intelligence of UAVs in complex flying environments like metro tunnels.By optimizing large-scale language models,this method converts human command to machine instruction.A visual target detection algorithm based on ensemble learning is utilized to identify potential risk targets.This approach also designs behavior tree nodes and system architecture in order to achieve intelligent decision-making and behavior planning.Through inter-modular information exchange in Robotic Operating System(ROS),this control algorithm allows UAVs to effectively interpret user commands,identify risk points intelligently during cruising and execute flying tasks like obstacle avoidance automatically,without preset waypoints.This method makes UAVs effectively adapt to dynamic flying environments and complex missions,thus advancing the application of UAV embodied intelligence.

关 键 词：地铁隧道 视觉反馈 行为树控制 具身智能 

分 类 号：V279.2[航空宇航科学与技术—飞行器设计]