A tight coupling mapping method to integrate the ESKF, g2o, and point cloud alignment

High-precision point cloud maps have drawn increasing attention due to their wide range of applications. In recent decades, point cloud maps are normally generated by simultaneous localization and mapping (SLAM) methods, which favor real-time performance over high precision. These methods generally focus on trajectory accuracy resulting in the unclearness of map accuracy. Therefore, to build a high-precision point cloud map and evaluate the mapping performance directly, this study proposes a tight coupling mapping method that integrates the error-state Kalman filter (ESKF), the general framework for graph optimization (g2o), and the point cloud alignment. An ESKF and a g2o are both used to improve the precision of the mapping process. Also, experiments based on a mobile mapping backpack prototype are conducted to verify the proposed method. Targets in the environment and a high-precision reference point cloud map are used to directly evaluate the map performance. The results indicate that the generated point cloud map is sufficiently precise and can reach the centimeter level.

EKF:

Extended Kalman filter

ESKF:

Error-state Kalman filter

ELS:

Enhanced line simplification

FG:

Factor graph

g2o:

General framework for graph optimization

GICP:

Generalized iterative closest point

ICP:

Iterative closest point

IMU:

Inertial measurement unit

LiDAR:

Light detection and ranging

NDT:

Normal distributions transform

SLAM:

Simultaneous localization and mapping

\(0_{3 \times 3}\) :

The zero matrix of size 3

\(a_{m}\) :

The output of the three-dimension accelerometer

\(a_{n}\) :

The white Gaussian noise of the accelerometer

\({\text{C}}_{{\text{L}}}^{{\text{I}}}\) :

The rotation matrix between LiDAR and IMU

\({\text{C}}_{{\text{i}}}^{{\text{X}}}\) :

The rotation matrix of Frame i in the coordinate system F_X

F_k :

The state transition matrix at time-step k

F b :

The body-fixed front-right-down coordinate system

F g :

The g2o optimization coordinate system

F n :

The local north-east-down coordinate system

H_{k + 1} :

The observation matrix at time-step k + 1

\({\text{I}}_{3 \times 3}\) :

The identity matrix of size 3

k :

Subscript k indicates the value is at time-step k

M :

Subscript M denotes the sensor measurement

\(n_{a}\) :

The Gaussian random walk noise of the accelerometer

\(n_{{\text{g}}}\) :

The Gaussian random walk noise of the gyroscope

\(p^{ + }\) :

The estimated position from the state equation

\(p_{{\text{L}}}\) :

The position from the vertexes in the g2o

\(q^{ + }\) :

The estimated quaternion from the state equation

\(q_{{\text{L}}}\) :

The position from the vertexes in the g2o

R:

The rotation matrix

T:

The superscript T represents the transpose operation

\({\text{t}}_{{\text{L}}}^{{\text{I}}}\) :

The translation vector between LiDAR and IMU

\({\text{t}}_{{\text{i}}}^{{\text{X}}}\) :

The position of Frame i in the coordinate system F_X

\({\text{v}}^{ + }\) :

The estimated velocity from the state equation

\({\text{v}}_{{\text{L}}}\) :

The velocity from the vertexes in the g2o

\(x_{k}\) :

The state vector is at time-step k

\(z_{k}\) :

The measurement vector is at time-step k

\(\delta p\) :

The position error

\(\delta q\) :

The quaternion error

\(\delta {\text{v}}\) :

The velocity error

\(\delta \beta_{a}\) :

The accelerometer bias error

\(\delta \beta_{{\text{g}}}\) :

The gyroscope bias error

\(\delta \theta\) :

The attitude angle error concerning the quaternion

\(\Delta t\) :

The discrete-time interval

\(\omega_{m}\) :

The output of the three-dimension gyroscope

\(\omega_{n}\) :

The white Gaussian noise of the gyroscope

\(\cdot\) :

The superscript \(\cdot\) indicates derivatives

\(\otimes\) :

The multiplication of the quaternion

\(*\) :

The superscript \(*\) represents the conjugate operation

This work was funded by The Hong Kong Polytechnic University (1-ZVN6, 4-BCF7), The State Bureau of Surveying and Mapping, P.R. China (1-ZVE8), and Hong Kong Research Grants Council (T22-505/19-N).

Authors and Affiliations

1. Department of Land Surveying and Geo-Informatics & Smart Cities Research Institute, The Hong Kong Polytechnic University, Hong Kong, China

   Bao Sheng, Shi Wenzhong, Fan Wenzheng, Chen Pengxin & Xiang Haodong

2. School of Remote Sensing and Information Engineering, Wuhan University, Wuhan, 430072, China

   Nie Mingyan

Corresponding author

Correspondence to Shi Wenzhong.

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Reprints and permissions