基于误差状态卡尔曼滤波估计的旋翼无人机输入饱和控制

Error State Kalman Filter Estimator Based Input Saturated Control for Rotorcraft Unmanned Aerial Vehicle

针对GPS(global positioning system)信号缺失环境下无人机自主飞行控制问题,设计了一种基于视觉与IMU(inertial measurement unit)融合的误差状态卡尔曼滤波(ESKF)框架,并在此基础上提出了一种新的输入饱和控制方法以进一步缓解视野约束以及运动模糊问题.不同于传统的扩展卡尔曼滤波(EKF)框架,本文设计的滤波框架是对误差状态进行更新与校正,而不是直接对系统状态进行估计.由于误差状态是小量,并且其线性程度较高,因此相对于系统状态局部线性化而言,误差状态的局部线性化的模型误差更小,进而可以提高状态估计的精度.基于ESKF框架得到的全状态估计,本文提出了一种新的线性与双曲正切混合的饱和函数,进而设计了输入饱和控制器并通过李亚普诺夫函数证明了闭环系统平衡点的渐近稳定性.最后,在旋翼无人机平台上的对比实验结果表明:本文ESKF方法得到的状态估计精度更高.另外,本文所提出的输入饱和控制方法有助于保证视觉特征在视野之内,并且比有界积分控制方法有更好的暂态以及稳态性能.

Regarding the autonomous flight control for unmanned aerial vehicle(UAV) in global positioning system denied environments, an error state Kalman filter(ESKF) framework is designed to fuse the vision and IMU(inertial measurement unit). On this basis, a novel input saturated control approach is proposed to further alleviate those issues due to motion blur and the field of view constraint. Different from the traditional extended Kalman filter(EKF) framework, the designed filter framework updates and corrects the error state rather than directly estimates the system state. Since the error state is a small variable with good linearity, the model error caused by its local linearization is smaller than that of the system state. Hence, the state estimation accuracy can be improved by using the error state. Based on the full state estimation in the ESKF framework, a novel hybrid linear and hyperbolic tangent saturation function is proposed to design the input saturated control approach. And it is shown that the equilibrium of the closed-loop system is asymptotically stable in the Lyapunov sense. Finally, comparative experimental results on the rotorcraft unmanned aerial vehicle(RUAV) demonstrate that the state estimation based on the proposed ESKF approach is more accurate. In addition, the proposed input saturated control method can help to keep the visual features in the field of view, and it has better transient and steady-state performances compared with the bounded integral control method.