Comparison of Various SLAM Systems for Mobile Robot in an Indoor Environment
Maksim Filipenko, Ilya Afanasyev · 2025
Read Comparison of Various SLAM Systems for Mobile Robot in an Indoor Environment by Maksim Filipenko, Ilya Afanasyev (2025) 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-2501-09490.txt.
Introduction
Maksim Filipenko and Ilya Afanasyev race a dozen robot navigation systems against each other indoors. Their 2025 paper compares SLAM methods, the tools that let a mobile robot map a space and track its own place in it at the same time.
The pair built a robot carrying a laser scanner, a single camera, and a stereo camera, then drove it through a typical office and fed the same recordings to every system. The field split by sensor: three laser systems, three single-camera systems, and four stereo systems. Laser-based Cartographer, single-camera ORB SLAM, and stereo RTAB-Map came out on top in their classes. The shared data set keeps the contest fair, and the results guide sensor choices for indoor robots.
Catalog record
| CREATOR | Maksim Filipenko, Ilya Afanasyev |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 19 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2501-09490
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Comparison of Various SLAM Systems for Mobile Robot in an Indoor Environment Maksim Filipenko, Ilya Afanasyev Abstract This article presents a comparative analysis of a mobile robot trajectories computed by various ROS-based SLAM systems. For this reason we developed a prototype of a mobile robot with common sensors: 2D lidar, a monocular and ZED stereo cameras. Then we conducted experiments in a typical office environment and collected data from all sensors, running all tested SLAM systems based on the acquired dataset. We studied the following SLAM systems: (a) 2D lidar-based: GMapping, Hector SLAM, Cartographer; (b) monocular camera-based: Large Scale Direct monocular SLAM (LSD SLAM), ORB SLAM, Direct Sparse Odometry (DSO); and (c) stereo camera-based: ZEDfu, Real-Time Appearance-Based Mapping (RTAB map), ORB SLAM, Stereo Parallel Tracking and Mapping (S-PTAM). Since all SLAM methods were tested on the same dataset we compared results for different SLAM systems with appropriate metrics, demonstrating encouraging results for lidar-based Cartographer SLAM, Monocular ORB SLAM and Stereo RTAB Map methods. I Introduction The map building problem for an unknown environment with use of onboard sensors while solving the localization problem at the same time is known as Simultaneous Localization and Mapping (SLAM) [ 1 , 2 ] . One of the classical approaches to solve this problem is filter-based SLAM. This method is based on the Bayesian filtering theory. It has two main steps: (1) the prediction step, when robot localization and map state are updated using the previous information about the system state and input control commands; (2) the measurement update, where current sensor data are matched to predicted system state in order to make new system state prediction. This approach has various implementations. The earlier SLAM system were based on Extended Kalman filter (EKF-based SLAM, [ 3 ] ), or on Particle Filter (e.g., FastSLAM, [ 4 ] ). For introduction to the basics of SLAM system we direct to surveys [ 5 ] , for more detailed information about filter-based systems recommend [ 6 ] . The SLAM problem can be solved with the use of different sensors and the choice of suitable sensors plays a special role in the effective operation of Unmanned Ground Vehicle (UGV) due to the limited autonomous resource. Most mobile robots have Inertial Measurement Unit (IMU), which includes accelerometers, gyroscope, and magnetometer, which measure orientation, angular velocity, and acceleration correspondingly. This approach to getting mobile robot localization at using only Inertial Navigation System (INS) can give significant navigation errors [ 7 ] . Therefore the main sensor for indoor robot navigation and SLAM is usually lidar. 2D lidar SLAM systems are currently presented in different packages like GMapping [ 8 ] (which uses Rao-Blackwellized particle filer [ 9 ] to learn grid maps from 2D lidar data), Hector SLAM [ 10 ] (that is another very popular ROS-based SLAM), and Cartographer [ 11 ] (which is one of the most recent systems). On the other hand, monocular and stereo cameras are good low-cost passive sensors, which can effectively solve SLAM problem working as a single source of information about an environment that referring to Visual SLAM (V-SLAM [ 12 , 13 ] ). All methods of V-SLAM can be divided into two groups: (1) feature-based, which use selected features for map building, and (2) direct that work with entire images as a whole. The earlier investigations on visual navigation were carried out with a binocular stereo camera [ 14 , 15 ] and a monocular camera (that was called MonoSLAM [ 16 ] ). Over the past decade, numerous methods have been proposed, including Parallel Tracking and Mapping (PTAM, [ 17 , 18 ] ), REgularized MOnocular Depth Estimation (REMODE, [ 19 ] ), Oriented FAST [ 20 ] and Rotated BRIEF [ 21 ] (ORB-SLAM, [ 22 , 23 ] ), Dense Tracking and Mapping in Real-Time (DTAM, [ 24 ] ), Large-Scale Direct monocular SLAM (LSD-SLAM, [ 25 ] ), Large-Scale Direct SLAM with Stereo Cameras (Stereo LSD-SLAM, [ 26 ] ), Fast semi-direct monocular visual odometry (SVO, [ 27 ] ), Real-Time Appearance-Based Mapping (RTAB map, [ 28 ] ), Dense Piecewise Parallel Tracking and Mapping (DPPTAM, [ 29 ] ), Direct Sparse Odometry (DSO, [ 30 ] ), ElasticFusion (Dense SLAM without a Pose Graph [ 31 ] ), Convolutional Neural Networks SLAM (CNN-SLAM, [ 32 ] ), Stereo Parallel Tracking and Mapping (S-PTAM, [ 33 ] ). Each of these methods has its advantages and disadvantages. Common features include weak resistance to unfavorable conditions, experimentation, poor performance with a weak geometric diversity of the environment, sensitivity to pure rotations [ 34 ] , [ 35 ] . These methods do not provide metric information that is required in some applications (this problem is also known as Scale ambiguity). The solution of this problem is discussed in computer vision and robotics society. Thus, e.g., the paper [ 36 ] suggested to use additional sensors like IMU, whereas the authors [ 37 ] proposed the approach for restoring metric information using information about the geometric sensor arrangement. The other important things, which can contribute to SLAM quality, are map initialization [ 38 ] , [ 39 ] , a loop closure [ 40 ] (that compensates errors accumulated during laps running by UGV) and a type of camera shutter, which is used [ 41 ] . More detailed description of features for Visual SLAM system [ 42 ] , which we investigated in this research, is presented in the Table I . TABLE I : DESCRIPTION OF FEATURES FOR VISUAL SLAM SYSTEMS INVESTIGATED IN THIS RESEARCH System Method Feature type Feature descriptor Map density Loop Closure PTAM feature FAST local patch of pixels sparse none SVO semi-direct FAST local patch of pixels sparse none DPPTAM direct intensity gradient local patch of pixels dense none LSD SLAM direct intensity gradient local patch of pixels dense Bag of Words place recognition ORB SLAM feature FAST ORB semi-dense FabMap DSO direct intensity gradient local patch of pixels dense none RTAB map feature GFTT, FAST, … SIFT, ORB, … sparse Bag of Words place recognition S-PTAM feature GFTT, FAST, … SIFT, ORB, … sparse Bag of Words place recognition In this paper, we mainly focus on a real robotics application for existing SLAM systems, which are implemented in Robot Operating System (ROS, [ 43 ] ). For this reason, we investigate the following SLAM systems: (a) 2D lidar-based: GMapping, Hector SLAM, Cartographer; (b) monocular camera-based: LSD SLAM, ORB SLAM, DSO; and (c) stereo camera-based: ZEDfu, RTAB map, ORB SLAM, S-PTAM. The brief information about ROS-based SLAM methods which were studied in this research are presented in the Table II . TABLE II : ROS-BASED SIMULTANEOUS LOCALIZATION AND MAPPING (SLAM) SYSTEMS STUDIED IN THIS RESEARCH Year System Sensor Ref. 2007 GMapping 2D lidar [ 8 ] 2007 Parallel Tracking and Mapping (PTAM) mono [ 17 ] 2011 Hector SLAM 2D lidar [ 10 ] 2014 Semi-direct Visual Odometry (SVO) mono [ 27 ] 2014 Large Scale Direct monocular SLAM (LSD SLAM) mono [ 25 ] 2014 Real-Time Appearance-Based Mapping (RTAB map) stereo [ 28 ] 2015 ORB SLAM mono, stereo [ 22 ] 2015 Dense Piecewise Parallel Tracking and Mapping (DPPTAM)