UC Riverside

Simultaneous Localization and Mapping (SLAM) is a fundamental task in robotics and computer vision, allowing a robot to build a map of its environment while estimating its own pose. In real-time SLAM scenarios, the processing time plays a critical role in achieving accurate and timely results. However, varying processing times due to computational load or system constraints can directly impact the performance of SLAM algorithms.

In this thesis, a new resource-efficient 3D SLAM framework is proposed using adaptive interval rates under consistent uncertainty. The proposed approach incorporates an adaptive mechanism that dynamically adjusts SLAM parameters based on the available processing time, allowing the system to adapt to changing computational loads and achieve real-time performance while maintaining accuracy. Specifically, upgrading the popular method, named Tightly-coupled Lidar Inertial Odometry via Smoothing and Mapping (LIO-SAM)[17] is utilized, which combines LiDAR and inertial sensor data for 3D SLAM.

The proposed approach utilizes processing time measurement and control of a key SLAM parameter such as key-frame frequency. By dynamically adapting this parameter, the system can optimize its performance based on the available processing time, maximizing the efficiency of system resources. We conduct extensive experiments and evaluations on real-world datasets to validate the effectiveness of the proposed approach and compare it with existing SLAM approaches.

The results demonstrate that the adaptive-SLAM approach enhances the adaptability and efficiency of LIO-SAM in real-time visual odometry scenarios. It achieves accurate and timely pose estimation while effectively utilizing the available processing time, making it suitable for applications that require real-time and adaptive SLAM capabilities, such as autonomous robotics, augmented reality, and virtual reality. The findings of this thesis contribute to the field of SLAM by providing a novel approach for maximizing the utilization of processing time in SLAM algorithms, opening up new possibilities for robust and efficient localization and mapping in dynamic environments.