深度相机与微机电惯性测量单元松组合导航算法

针对室内非结构环境下，全球定位系统、无线电定位等定位手段使用困难，轮式里程计在楼梯等场合易出现打滑或空转而误差较大，单一视觉传感器或微机电系统（MEMS）惯性测量单元（IMU）很难实现高精度自主定位，以及传统的视觉与MEMS IMU组合导航算法复杂、计算量大、导航精度低等问题，提出一种适用于室内的深度相机与MEMS IMU松组合导航算法。MEMS IMU预积分结果作为改进迭代最近点（ICP）算法的迭代初值，大幅减少了迭代次数；通过深度相机和MEMS IMU分别计算载体的位置并作差运算，将位置差值作为量测信息，使用扩展卡尔曼滤波估计MEMS IMU的导航误差，修正航位推算的结果；利用Kinect v1深度相机和MTI 100-IMU搭建的平台进行实验验证。结果表明，基于MEMS IMU辅助的改进ICP算法能够减少迭代次数约50%，基于位置差值的深度相机与MEMS IMU松组合算法导航定位误差小于总里程的10%.

Integrated Navigation Algorithm of Depth Camera/MEMS IMU

GPS, radio positioning and other positioning methods are difficult to use in the indoor environment. Wheel odometer is easy to slip or idle in stairs and other occasions. Single vision sensor or micro-electromechanical system (MEMS) inertial measurement unit (IMU) is difficult to achieve high-precision autonomous positioning. The traditional vision/MEMS IMU integrated navigation algorithm is complex, computationally expensive and of low accuracy. In order to solve these problems, a loose coupled navigation algorithm based on depth camera/MEMS IMU is proposed for indoor application. The pre-integration results of MEMS IMU are used as the initial value of the iterative closest point (ICP) algorithm, which greatly reduces the number of iterations. The position of a carrier is calculated by depth camera and MEMS IMU, respectively. The position difference is obtained by subtracting the two positions. The errors of MEMS IMU is estimated and the result of dead reckoning is corrected by using the position difference as the measuring information of EKF. The results of experiments with Kinect v1 and MTI 100-IMU show that the improved ICP algorithm based on MEMS IMU can effectively reduce the number of iterations by about 50%. The depth camera/MEMS IMU loose coupled algorithm based on position difference can make the position errors less than 10% of the total mileage.