Calibration of the internal and external parameters of wheeled robot mobile chasses and inertial measurement units based on nonlinear optimization

Gang Peng, Zezao Lu, Zejie Tan, Dingxin He, Xinde LiarXiv 2020 · arXiv

Introduction

For autonomous mobile robots, precise localization and navigation depend heavily on the accurate calibration of onboard sensors. This includes both intrinsic parameters (e.g., wheel diameters, sensor biases) and extrinsic parameters (the spatial relationship between sensors like IMUs and the robot's chassis). While internal IMU calibration is well-established, determining the precise rotation and translation between an IMU and a wheel-based chassis often requires specialized external equipment. Small misalignments—such as a few degrees of tilt or centimeters of offset—can lead to significant drift in pose estimation. This paper presents a self-contained, two-step calibration method that utilizes only the robot's existing sensors to align the IMU with the chassis and tune wheel odometer parameters.

Methodology: A Decoupled Approach

The proposed method simplifies the complex 6-degree-of-freedom (DoF) calibration problem by decoupling it into two distinct stages: tilt estimation and planar optimization. This approach assumes that the camera-IMU system has already undergone internal calibration.

The two-step calibration pipeline: PCA for tilt and nonlinear optimization for heading and displacement.

1. Tilt Calibration via PCA

When a robot operates on a flat horizontal plane, its primary rotation occurs around the vertical (Z) axis. By collecting gyroscope data during planar motion and applying Principal Component Analysis (PCA), the system identifies the dominant axis of rotation. This axis corresponds to the chassis's Z-axis, allowing for the direct calculation of the IMU's pitch and roll relative to the ground.

2. Heading and Displacement Optimization

Once the tilt is "locked in," the remaining extrinsic parameters—the heading (yaw) angle and the XY displacement—are resolved alongside the wheel odometer's internal scale factors. This is achieved by comparing the trajectory generated by Visual-Inertial Odometry (VIO) with the trajectory from the wheel encoders.

Coordinate frames involved in the calibration: Camera (C), Body (B), Fake-body (F), and Odometry (O).

The authors formulate a nonlinear least squares problem using the Google Ceres library. The optimization minimizes the "residual" error between the relative poses calculated by the VIO and those derived from the wheel odometry. This step effectively calibrates the IMU's position on the chassis while simultaneously correcting for manufacturing variances in wheel diameter and friction (scale factors).

Experimental Setup and Results

The method was evaluated using a custom Mecanum wheel omnidirectional platform equipped with an Intel RealSense ZR300 depth camera and a Bosch BMI055 IMU. The VINS-Mono algorithm provided the VIO ground truth for the optimization process.

Validation through repeated experiments demonstrated high precision and repeatability:

  • Spatial Consistency: The standard deviation for the IMU's X-axis and Y-axis positions were 1.21 cm and 0.4 cm, respectively.
  • Heading Accuracy: The heading angle showed a standard deviation of only 0.128°, indicating extremely high repeatability.
  • Wheel Calibration: The system successfully identified wheel scale factors (e.g., 99.7% and 103.7%), accounting for real-world physical discrepancies.

Conclusion and Limitations

By exploiting planar motion constraints and decoupling the optimization problem, this method provides a "field-ready" solution for robot calibration without requiring external markers or motion capture systems. However, certain limitations remain. Because the robot moves on a 2D plane, the vertical displacement (Z-axis height) of the IMU relative to the chassis is not mathematically observable and must be measured manually. Additionally, the method relies on a flat calibration surface and a feature-rich environment to ensure the reliability of the VIO system. Future work may address these constraints to enable calibration in more diverse environments.

Summary updated

Abstract

Mobile robot positioning, mapping, and navigation systems generally employ an inertial measurement unit (IMU) to obtain the acceleration and angular velocity of the robot. However, errors in the internal and external parameters of an IMU arising from defective calibration directly affect the accuracy of robot positioning and pose estimation. While this issue has been addressed by the mature internal reference calibration methods available for IMUs, external reference calibration methods between the IMU and the chassis of a mobile robot are lacking. This study addresses this issue by proposing a novel chassis-IMU internal and external parameter calibration algorithm based on nonlinear optimization, which is designed for robots equipped with cameras, IMUs, and wheel speed odometers, and functions under the premise of accurate calibrations for the internal parameters of the IMU and the internal and external parameters of the camera. All of the internal and external reference calibrations are conducted using the robot's existing equipment without the need for additional calibration aids. The feasibility of the method is verified by its application to a Mecanum wheel omnidirectional mobile platform as an example, as well as suitable for other type chassis of mobile robots. The proposed calibration method is thereby demonstrated to guarantee the accuracy of robot pose estimation.
BibTeX
@misc{peng2020calibration,
  author = {Gang Peng and Zezao Lu and Zejie Tan and Dingxin He and Xinde Li},
  title = {{Calibration of the internal and external parameters of wheeled robot mobile chasses and inertial measurement units based on nonlinear optimization}},
  year = {2020},
  eprint = {2005.08284},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2005.08284},
}
Figure