Robust LM visual inertia odometer and UWB hybrid high-precision indoor positioning method

Abstract

The invention provides a robust LM visual inertia odometer and UWB hybrid high-precision indoor positioning method. The method comprises the following steps of deploying a certain number of UWB base stations in an indoor open area, installing and combining UWB and camera equipment, positioning a to-be-detected area, and conducting interpolation through a cubic spline curve in the positioning process, and adopting an improved LM algorithm to optimize a model used for visual VIO and UWB fusion, thereby effectively improving the precision of indoor positioning.

Description

technical field [0001] The invention relates to the technical field of indoor positioning, in particular to a hybrid high-precision indoor positioning method based on robust LM visual inertial odometer and UWB. Background technique [0002] Visual navigation and positioning is to extract different images through the visual system during camera movement, by detecting the changes of these different images, extracting and matching the same feature points, and judging the movement changes of the feature points to estimate the motion of the camera. Because visual odometry (VO) cannot track well in the face of simple rotation, it is usually combined with low-cost IMU and other inertial sensors in practical applications. After the combination, visual inertial odometry (VIO) can be very robust. Great app. Although VIO has high positioning accuracy in the case of good lighting conditions and image quality, it is essentially a local positioning algorithm due to the lack of global pos...

AI Technical Summary

Problems solved by technology

The steepest descent method cannot determine the convergence step size, a small step size will lead to slower convergence, and a larger step size will cause the algorithm to diverge

The Gauss-Newton method is approximated by the second-order Taylor expansion. Only near the expansion point will there be a better approximation effect. If the Hessian matrix is ​​a non-singular matrix, it may cause the algorithm to diverge.

LM can overcome the above shortcomings very well and has been widely used in many nonlinear optimization examples, but LM also has its own defects.

Although LM can overcome the influence of the ill-conditioned solution of the Hessian matrix, because the LM algorithm uses the least squares estimation as the convergence condition, it cannot resist the interference of the gross error of the sensor data center. Ensure the stability and reliability of UWB positioning

If the LM algorithm is directly used for optimization, the resulting fusion solution will deviate from the true value

Images