紧耦合INS/视觉相对位姿测量方法

Relative pose measurement based on tightly-coupled INS/Vision

研究了一种紧耦合INS/视觉相对位姿测量方法在无人机自主空中加油中的应用。该方法直接将特征点的图像坐标作为滤波器输入,避免了求解复杂的非线性位姿方程,尤其是特征点提取不全时,该方法具有较强的鲁棒性。引入相对惯导误差建立了增广状态模型,根据杆臂效应详细推导了紧、松两种耦合模式的量测方程。采用扩展卡尔曼滤波算法估计误差状态,并校正惯导输出获取精确的相对位姿信息。仿真结果表明,与松耦合模式相比,紧耦合在提高系统实时性的同时可获得更高的测量精度,位置误差小于0.1 m,姿态角误差小于3'。

The relative pose measurement based on tightly-coupled INS/vision is investigated and applied to autonomous aerial refueling for unmanned aerial vehicles（UAVs）.By inputting feature image coordinates to the filter directly,the measurement does not need to solve complicated nonlinear line of sight pose equations and exhibits robustness especially in feature extraction scarcity.By introducing relative inertial errors,the extended state model is established,and the tightly and loosely coupled measurement models are derived using level-arm effect.The state errors estimated by extended kalman filter algorithm are applied to correct INS outputs,which can obtain precise relative pose consequently.Simulation results show that the tightly-coupled INS/Vision could achieve more accurate estimation and higher real time performance than loosely-coupled.The position error is less than 0.1 m,and the attitude error is less than 3＇.