摘要：For pose estimation in visual servoing， by assuming the relative motion over onesample period to be constant, many existing works adopt a linear time-invariant (LTI) dynamicmodel. Since the standard feature point transformation is nonlinear, extended Kalman filtering.(EKF) has become popular due to its simplicity. Thus, the problem at hand becomes filtering ofan LTI system with a time-varying output matrix. To obtain satisfactory performance, accurateknowledge of the noise covariances is essential. Various methods have been proposed on howto adaptively update their values to improve performance. However,these techniques cannotguarantee the positive semidefiniteness (PSD) of the covariance estimates. In this paper，wepropose to apply the autocovariance least-squares (ALS) approach to covariance identificationin pose estimation. The ALS approach can provide reliable estimates of the covariance matriceswhile maintaining their PSD and imposing desired structural constraints.Our tests show thatusing the covariance estimates from the ALS method in EKF can reduce the average poseestimation error by more than 30% in simulation, and the average position estimation error byabout 30% using experimental data, respectively, compared to a hand-tuned EKF.
