UA-MPC: U ncertainty- A ware M odel P redictive C ontrol for Motorized LiDAR Odometry
Jianping Li, Xinhang Xu, Jinxin Liu, Kun Cao · 2024
Read UA-MPC: U ncertainty- A ware M odel P redictive C ontrol for Motorized LiDAR Odometry by Jianping Li, Xinhang Xu, Jinxin Liu, Kun Cao (2024) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2412-13873.txt.
Introduction
Jianping Li and co-authors give a motorized laser scanner a smarter way to swing its beam. Their 2024 paper presents UA-MPC, a control strategy that weighs how uncertain each scan direction is. A single rotating scanner widens the field of view cheaply, but constant-speed spinning wastes effort in complex scenes.
The controller predicts how observable each direction is by ray tracing, then models that pattern to set motor speeds scene by scene. A realistic open-source simulator let the team test control ideas safely. In both simulated and real tests, position error fell by over sixty percent. Scanning efficiency dipped by less than two percent, a trade the authors call well worth it.
Catalog record
| CREATOR | Jianping Li, Xinhang Xu, Jinxin Liu, Kun Cao |
|---|---|
| YEAR | 2024 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 33 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2412-13873
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
UA-MPC: U ncertainty- A ware M odel P redictive C ontrol for Motorized LiDAR Odometry Jianping Li, Xinhang Xu, Jinxin Liu, Kun Cao Abstract Accurate and comprehensive 3D sensing using LiDAR systems is crucial for various applications in photogrammetry and robotics, including facility inspection, Building Information Modeling (BIM), and robot navigation. Motorized LiDAR systems can expand the Field of View (FoV) without adding multiple scanners, but existing motorized LiDAR systems often rely on constant-speed motor control, leading to suboptimal performance in complex environments. To address this, we propose UA-MPC, an uncertainty-aware motor control strategy that balances scanning accuracy and efficiency. By predicting discrete observabilities of LiDAR Odometry (LO) through ray tracing and modeling their distribution with a surrogate function, UA-MPC efficiently optimizes motor speed control according to different scenes. Additionally, we develop a ROS-based realistic simulation environment for motorized LiDAR systems, enabling the evaluation of control strategies across diverse scenarios. Extensive experiments, conducted on both simulated and real-world scenarios, demonstrate that our method significantly improves odometry accuracy while preserving the scanning efficiency of motorized LiDAR systems. Specifically, it achieves over a 60% reduction in positioning error with less than a 2% decrease in efficiency compared to constant-speed control, offering a smarter and more effective solution for active 3D sensing tasks. The simulation environment for control motorized LiDAR is open-sourced at: https://github.com/kafeiyin00/UA-MPC.git . † † footnotetext: J. Li, X. Xu, J. Liu, S. Yuan, and L. Xie are with School of Electrical and Electronic Engineering, Nanyang Technological University, 50 Nanyang Avenue, Singapore. K. Cao is with the Department of Control Science and Engineering, College of Electronics and Information Engineering, and Shanghai Research Institute for Intelligent Autonomous Systems, Tongji University, Shanghai, China. (E-mail: [email protected], [email protected], [email protected], [email protected], [email protected], [email protected]) I Introduction Accurate and comprehensive 3D sensing of the environment using a LiDAR system is a cornerstone of advancements in photogrammetry [ 1 , 2 , 3 ] and robotics [ 4 ] . The high-resolution 3D point clouds generated by LiDAR systems enable a wide range of critical applications, including facility inspection [ 5 ] , where detailed structural analysis is required; BIM [ 6 ] , for precise 3D digitalization; and autonomous robot navigation [ 7 , 8 ] , where accurate environmental awareness is essential for path planning and obstacle avoidance [ 9 ] . However, a single LiDAR typically suffers from a limited Field of View (FoV), which can restrict its usability in tasks requiring full environmental coverage. To overcome this limitation, rotating the LiDAR with a motor to create a motorized LiDAR system significantly enhances the FoV without the need for additional LiDAR [ 10 ] . This approach not only improves coverage but also offers advantages in terms of reduced weight, size, and system complexity, making it particularly well-suited for applications on lightweight and resource-constrained platforms, such as unmanned aerial vehicles (UAVs) [ 11 , 12 ] and mobile robots [ 13 , 14 ] . Figure 1: Coordinates and mechanical design of the proposed motorized LiDAR system. Thanks to the compact design and expansive FoV provided by rotating LiDAR systems equipped with motors, LO using motorized LiDAR systems has emerged as a key area of research over the past decade [ 15 , 13 , 14 ] . Early pioneering efforts [ 16 , 17 , 18 ] demonstrate the integration of 2D LiDAR with motors, enabling 3D sensing and the realization of 3D Simultaneous Localization and Mapping (SLAM) for robotic platforms. These foundational works highlight the potential of motorized LiDAR systems for dynamic and high-resolution environmental mapping. With advancements in LiDAR manufacturing technologies, the integration of 3D multi-beam LiDAR into motorized systems has further expanded their capabilities, as validated by recent studies [ 19 ] . To enhance SLAM accuracy on the motorized LiDAR system, several approaches have employed elastic mapping techniques [ 20 , 21 ] , which adaptively refine the map structure to achieve better accuracy. Additionally, innovative motor control strategies have been explored, such as the Lissajous rotating scheme [ 22 ] , which utilizes two motors to generate optimized scanning patterns, significantly improving data quality for downstream SLAM processes. These existing studies collectively demonstrate the significant advancements made possible by motorized LiDAR systems. However, most existing studies on motorized LiDAR-based LO [ 14 , 15 , 16 ] have employed constant-speed motor settings, determined largely by empirical experience rather than scene-specific optimization. Active LiDAR SLAM is gaining increasing attention from researchers due to its ability to dynamically adapt sensor behavior or vehicle paths to optimize performance in complex environments [ 23 , 24 ] . For our case, unlike traditional methods that rely on fixed scanning patterns, we want to incorporate adaptive control strategies, such as adjusting motor speeds, to achieve a balance between accuracy and efficiency. This adaptability allows the system to respond intelligently to varying environmental conditions and task-specific requirements, significantly enhancing its performance [ 25 ] . This idea is similar to the work from [ 26 ] , which proposes an adaptive control of a motorized LiDAR system considering the region of interest but without the accuracy of the LO. In challenging scenarios, such as feature-sparse, improper motor operation can lead to data degeneracy, adversely affecting both odometry accuracy and mapping reliability [ 15 ] . To address these limitations, it is essential to explore methods for dynamically optimizing motor control to maximize system accuracy while maintaining efficiency. This necessitates a comprehensive understanding of the relationship between motor behavior, scene characteristics, and system outcomes. In light of these considerations, we identify two critical issues for motorized LiDAR systems that warrant further investigation: I-A How to design a proper control strategy to balance the accuracy and efficiency? To the best of our knowledge, limited work has been done on developing effective motor control strategies for smarter motorized LiDAR systems. Intuitively, motorized LiDAR systems could improve LO accuracy by prioritizing areas with rich features, thereby enhancing feature association quality and overall localization precision. However, this targeted tracking strategy introduces potential trade-offs. Focusing on specific regions may lead to data holes in the point cloud, where parts of the environment are insufficiently scanned, negatively impacting global mapping consistency. Additionally, such strategies risk reducing scanning efficiency, as more time and resources are allocated to high-feature-density areas at the expense of overall coverage. Addressing these challenges requires a comprehensive approach that not only maximizes feature utilization but also ensures uniform rotating coverage and operational efficiency [ 27 ] . Developing smarter control strategies that dynamically balance these competing demands is crucial for advancing the capabilities of motorized LiDAR systems and unlocking their full potential in active SLAM applications. I-B How to evaluate the different control strategies? Existing simulation environments (e.g., Gazebo [ 28 ] , Airsim [ 29 ] , Isaac [ 30 ] ) have played a crucial role in advancing research on LiDAR SLAM and motor control strategies. However, these platforms exhibit significant limitations, particularly in complex scenarios. A primary drawback is the considerable manual effort required to create detailed and realistic simulation environments. Designing complex scenes with diverse features, varying densities, and dynamic elements often demands substantial human labor, time, and expertise, limiting scalability and adaptability for broader applications.