毕业生知识库

四旋翼飞行器因其适应场景广、结构紧凑、机动性强等特点，在各领域具有重要的应用价值与发展空间，其中无 GPS（Global Positioning System, GPS）环境下的实时定位与运动规划是四旋翼飞行器的关键技术之一。本文研究基于视觉惯性里程计的四旋翼飞行器定位及运动规划，主要研究内容概要如下:
1. 针对无 GPS 环境下四旋翼飞行器的实时定位问题，本文实现了一种基于非线性优化与 EKF（Extended Kalman Filter, EKF）的视觉惯性定位算法。算法采用非线性优化的方法融合视觉信息与外接 IMU（Inertial Measurement Unit, IMU）信息减小累计误差，利用 EKF 算法对非线性优化输出的位姿估计进行运动补偿，从而提高位姿估计帧率、优化位姿估计的平滑度，实现了对四旋翼飞行器位姿实时、准确的估计。
2. 实现了一种约束环境下的四旋翼飞行器运动规划算法。为求解约束环境下的四旋翼飞行器的目标轨迹，本文设计了四旋翼飞行器运动规划系统，该系统基于 A* 算法实现了前端的路径规划，采用 tie-breaker 技术与 Manhattan 启发函数优化 A* 算法搜索效率，利用骨架提取算法剔除 A* 算法输出路径中的冗余路径点，在后端基于 Minimum Snap 闭式解的方法求解出四旋翼飞行器运动的目标迹。
3. 构建了四旋翼飞行器定位导航实验平台并进行了实验研究。构建了四旋翼飞行器实时定位导航实验系统，利用飞行平台、视觉惯性单元、飞行控制模块、板载计算机、地面站集成了四旋翼飞行器实验平台，采用 ROS（Robot Operating System, ROS）架构实现了实时定位导航系统的集成与构建，开展了四旋翼飞行器实时定位导航实验，验证了定位算法的实时性与准确性、运动规划系统的安全性、高效性。
本研究实现了一种四旋翼飞行器自主定位导航系统，该系统使四旋翼飞行器能够在无 GPS 的约束环境下实时准确地定位并安全高效地抵达任务点，为四旋翼飞行器自主执行复杂任务提供了技术支撑。
 

The quadrotor has practical value and great potential in various fields due to its wide application, compactness, and flexibility. The real-time positioning and motion planning technology in GPS-denied environments is one of the keys for quadrotors. This thesis focus on the localization and motion planning for quadrotors with Visual-Inertial Odometry(VIO). The main contributions of this thesis are summarized as follows.
1. A visual inertial odometry algorithm using nonlinear optimization and EKF is proposed to solve the quadrotor real-time positioning problem. This thesis uses a nonlinear optimization method to fuse visual information and external IMU information to eliminate the accumulated errors. Subsequently, an EKF algorithm is used to perform motion compensation on the nonlinear optimization output to improve the frame rate and optimize the smoothness of the state estimation, facilitating the VIO to estimate thequadrotor state in real-time and accurately.
2. Quadrotor motion planning under confined environment is addressed. To calculate the target trajectory of quadrotors under confined environment, this research designs a quadrotor motion planning system. In the system front-end, waypoints are calculated by the A* algorithm using the tie-breaker technique mixed with Manhattan as a heuristic function and refined by a skeleton extraction algorithm. In the system back-end, the quadrotor target trajectory is calculated using the Minimum Snap closed-form solution.
3. The quadrotor positioning and navigation experiments are performed. This thesis builds a quadrotor real-time positioning and navigation system, which integrates a quadrotor, a visual-inertial unit, flight control module, an on-board computer, and ground station. This research then implements a real-time position and navigation system using ROS architecture, and performs the quadrotor positioning and navigation experiments. The experiment results demonstrate the real-time and accuracy of the proposed VIO algorithm, and validate the safety and efficiency of the motion planning system.
This thesis provides a feasible solution for the autonomous positioning and navigation of the quadrotor. Based on this solution, the quadrotor can accurately position itself in real-time and reach the mission point safely and efficiently in a GPS-denied and constrained environment, providing a technical basis for the quadrotor to execute various complex tasks autonomously.