Supplementary Materials:
Night-Rider: Nocturnal Vision-aided Localization in Streetlight Maps Using Invariant Extended Kalman Filtering
Tianxiao Gao1, Mingle Zhao1, Chengzhong Xu1, and Hui Kong1
1Tianxiao Gao, Mingle Zhao, Chengzhong Xu and Hui Kong are with the State Key Laboratory of Internet of Things for Smart City (SKL-IOTSC), Faculty of Science and Technology, University of Macau, Macau, China. ({ga0.tianxiao, zhao.mingle}@connect.umac.mo, {czxu, huikong}@um.edu.mo)
This document provides supplementary materials for the ICRA 2024 submission “Night-Rider: Nocturnal Vision-aided Localization in Streetlight Maps Using Invariant Extended Kalman Filtering”.
-A Derivation of Error Propagation Model
From the Eq. (8), we need to derive the linearized model of right invariant error and bias error .
For the states fitting into Lie group, the derivative of right invariant error associates with the log of invariant error based on .
|
|
|
(27) |
With Eq. (3), the derivative of is deduced as below.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
(28) |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
(29) |
|
|
|
|
|
|
|
|
|
|
|
|
(30) |
On the other hand, the differential of bias errors is calculated as
|
|
|
(31) |
Therefore, the linearized model can be written as in Eq. (9).
-B Derivation of and
The covariance matrix of reprojection error in Eq. (18) for the streetlight observation and cluster is derived as below:
|
|
|
(32) |
|
|
|
|
|
|
|
|
|
(33) |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
(34) |
where and is a preset parameter.
Similar to , the variance of angle error in Eq. (19) also originates from the covariance of detected streetlight observation and current pose.
|
|
|
|
|
|
(35) |
|
|
|
|
|
|
|
|
|
(36) |
|
|
|
|
|
|
|
|
|
|
|
|
(37) |
|
|
|
|
|
|
|
|
|
|
|
|
(38) |
where denotes the homogeneous coordinates. .
-C Derivation of the Jacobian Matrix of the Camera-based Observation Model
For the derivation of the Jacobian matrix in Eq. (24), we transform the observation function in Eq. (23) as:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
(39) |
The Jacobian matrix is formulated as:
|
|
|
|
|
|
|
|
|
|
|
|
(40) |
where denotes the third row of and is the third element of . By substituting Eq. (-C) into Eq. (-C), the Jacobian matrix of the camera-based observation model can be derived.
|
|
|
|
|
|
|
(41) |
|
|
|
(42) |
|
|
|
(43) |
|
|
|
|
|
|
|
|
|
|
|
|