# Lidar SLAM for Autonomous Driving Vehicles

**Authors:** Farhad Aghili  
**arXiv:** [2208.11855](https://arxiv.org/abs/2208.11855)  
**Date:** 2022-08-25  
**Summary page:** https://www.summarizepaper.com/en/paper/2208.11855v1/

## Key points

- Comprehensive study on Lidar-based Simultaneous Localization and Mapping (SLAM) for autonomous driving vehicles
- Fusion of data from landmark sensors and strap-down Inertial Measurement Unit (IMU) using adaptive Kalman filter (KF)
- Proposal of self-tuning filter to estimate IMU calibration parameters and measurement noise covariance
- Derivation of closed-form expressions for key components of SLAM system for real-time implementation
- Analysis of landmark alignment to examine system observability
- 3D SLAM system remains observable under certain conditions, ensuring reliable localization and mapping in autonomous driving scenarios
- Advancement of Lidar-based SLAM techniques through investigation of data fusion with IMU using adaptive Kalman filter
- Enhanced accuracy through estimation of IMU calibration parameters and measurement noise covariance
- Real-time implementation enabled by derived closed-form solutions
- Examination of system observability provides insights into robust performance in challenging environments.

## Summary

This paper presents a comprehensive study on Lidar-based Simultaneous Localization and Mapping (SLAM) for autonomous driving vehicles. The authors investigate the fusion of data from landmark sensors and a strap-down Inertial Measurement Unit (IMU) using an adaptive Kalman filter (KF). They also explore the observability of the system. In addition to estimating the vehicle's states and landmark positions, the authors propose a self-tuning filter that estimates the IMU calibration parameters and the covariance of the measurement noise. This approach allows for more accurate localization and mapping in real-time scenarios. To facilitate real-time implementation, the authors derive closed-form expressions for key components of the SLAM system, including the discrete-time covariance matrix of the process noise, state transition matrix, and observation sensitivity matrix. These closed-form solutions enable efficient computation and make them suitable for real-time applications. Furthermore, an analysis of landmark alignment is conducted to examine system observability. It is concluded that under certain conditions such as proper alignment between landmarks, the 3D SLAM system remains observable. This finding provides valuable insights into ensuring reliable localization and mapping in autonomous driving scenarios. Overall, this research contributes to advancing Lidar-based SLAM techniques for autonomous vehicles by investigating data fusion with an IMU using an adaptive Kalman filter. The proposed self-tuning filter enhances accuracy by estimating IMU calibration parameters and measurement noise covariance while derived closed-form solutions enable real-time implementation. Examining system observability sheds light on ensuring robust performance in challenging environments.