Optimization RTAB-Map Based on TORO Graph to Filter Wrong Loop Closure Detection for Search and Rescue Robot Application
Syahri Muharom, Tukadi, Riza Agung Firmansyah, Andy Rachman, Yuliyanto Agung Prabowo, Muchamad Kurniawan, Ilmiatul Masfufiah, Pratama Sandi Alala, Trisna Wati · 2023
This paper discusses 3D area mapping using RTAB-Map and TORO Optimizer, where visual-based mapping is becoming increasingly important because of the need to represent an area from the air. The problem is optimizing the process of creating an area map by utilizing loop closure in RTAB-Map. However, loop closure detection is still a challenge due to errors in detection in making area maps. To solve it, this research introduces an approach that integrates TORO Graph for optimization in 3D area mapping using ROS, and this study focuses on real-time optimization of visual-based mapping. This system works after the loop closure process, and the TORO graph optimizes the system to reduce errors in the loop closure process. From the system developed, results were obtained based on the analysis of over 650 data points, indicating that when using only loop closure, the number of matching IDs for map creation ranges from 294 to 677. However, when TORO Optimizer is employed, the number of matching IDs becomes more limited, ranging from 182 to 538. The number of matching IDs decreases because TORO Optimizer removes some IDs deemed erroneous in the feature-matching process for map creation. From the results obtained, the TORO graph can reduce errors in loop closure, so making 3D area maps is better. It is hoped that by utilizing TORO Optimizer, the quality of 3D area mapping results can be enhanced for aerial mapping in search and rescue robot applications.