伴随着传感器技术以及嵌入式计算技术的发展，无人机在各领域的应用正处于快速发展期。由于无人机保有量和使用频率的不断增加，对无人机飞行实施有效的监管已成为现实的迫切需求。针对无人机飞行安全监管的现状，以及当前对于一整套基于云平台的监管系统以及传感融合定位模块的完整解决方案研究的不足，本文充分利用传感融合定位算法的优势，设计并实现了满足“小型化、集成化、低成本”的无人机实时融合定位模块，并针对实现过程中出现的算法自适应能力不足和计算复杂度过大的问题，对自适应算法和低计算复杂度算法进行了深入研究。首先设计并实现了使用加速度计、陀螺仪、磁力计、气压计以及GNSS的低成本无人机融合定位模块硬件。为了更好地提高其定位效果，利用了六位置确定性误差校准和基于Allan方差的随机误差建模方法作为使用前的标定及数据处理。对比了EKF、UKF、USQUE三种算法在飞行轨迹跟踪实现中的效果，验证了定位模块硬件的可用性并为后续的算法研究提供了硬件基础和算法参考。然后针对融合定位模块在实现过程中存在的传感器随机特性不稳定或未知的问题，本文利用变分贝叶斯推断的思想，提出了针对融合定位场景的VBEKF-P和VBUKF-P算法，以满足在传感器随机特性不稳定情况下的自适应状态估计。利用飞行轨迹跟踪实验，验证了算法在低估过程噪声协方差的情况下的高精度和自适应能力。鉴于自适应算法本身存在的计算消耗过大的问题，本文利用了融合算法过程中中间变量稀疏和对称的特性，并采用了级联模型的思路，提出了一种降低计算复杂度的算法。该算法将一体式的模型解耦为姿态环和速度/位移环，并分别对上述两个环节使用EKF和VBEKF-P算法进行状态估计。经过实验验证，所提出的低计算复杂度算法能够满足设计条件下的计算复杂度要求。为了实现最终的无人机监管功能，本文设计了相应的低空安全飞行监管系统，包括流程设计、云系统架构设计与实现以及相应的功能效果验证。并完成了融合定位模块的运行逻辑及上位机软件的开发，最终实现一整套无人机低空安全监管解决方案。

With the development of sensor technology and embedded computing technology, the application of UAV is in a period of rapid development. Due to the increasing amount and the frequently usage of UAV, effective supervision of UAVs has become a very realistic demand. In view of the current demand for UAV flight safety supervision and the current lack of research on a complete set of supervision system based on cloud platform and the complete solution of sensor fusion positioning module, this paper makes full use of the advantages of sensor fusion positioning algorithm to design and implement a "miniaturized, integrated and low-cost" UAV real-time fusion positioning module. According to the problems of insufficient adaptive ability and excessive computational complexity in the implementation process, the adaptive algorithm and low computational complexity algorithm are studied.Firstly, the hardware of low-cost UAV fusion positioning module using accelerometer, gyroscope, magnetometer, barometer and GNSS is designed and implemented. In order to better improve its positioning effect, six-position deterministic error calibration and random error modeling method based on Allan variance are used as calibration and data pre-processing. The performance of EKF, UKF and USQUE algorithms in flight trajectory tracking are compared, the availability of positioning module hardware is verified, and a reference is provided for subsequent algorithm research.Then, aiming at the problem that the random characteristics of the sensor are unstable or unknown in the implementation of the fusion positioning module, using the idea of variational Bayesian inference, VBEKF-P and VBUKF-P algorithms for the fusion positioning scene are proposed to meet the demand of adaptive state estimation. The flight trajectory tracking experiment is used to verify the high accuracy and adaptive ability of the algorithm under the condition of underestimating the process noise covariance.In view of the excessive computational consumption of the adaptive algorithm, this paper utilizes the sparse and symmetrical characteristics of intermediate variables in the process of fusion algorithm, and adopts the idea of cascade model to propose an algorithm to reduce the computational complexity. The algorithm decouples the integrated model into attitude loop and velocity / displacement loop, and uses EKF and VBEKF-P algorithms to estimate the state of the above two loops respectively. Experimental results show that the proposed low computational complexity algorithm can meet the computational complexity requirements under design conditions.In order to realize the final UAV supervision function, the corresponding flight safety supervision system is also designed, including process design, cloud system architecture design and implementation, and corresponding function performance verification. The operation logic development of the fusion positioning module and the upper computer software development are also finished, and finally realize a complete set of UAV low altitude safety supervision solution.