17:15 〜 18:45
[HTT17-P02] 移動ロボットを使用した屋外不整地環境の3次元地図生成の研究
キーワード：3次元地図生成、不整地、SLAM、LIDAR、ロボット
In Japan, a lot of natural disasters such as huge earthquakes, volcanic eruptions and floods can happen and cause grate damages to human life and social activity. After the disasters, immediate investigation of surrounding terrains and situation of the disaster areas are required to help rescue and reconstruction works. Then, utilization of robotic technique is useful to investigate the environment without closing the dangerous areas. The data obtained by the sensors such as camera and LIDAR on a moving robot can be used to explore details of the surrounding terrains including the changes after the disaster. We research a method to construct an accurate 3-D map of outdoor irregular terrains such as disaster areas with high accuracy and resolution from onboard sensor data of the unmanned ground vehicle.
The SLAM (Simultaneous Localization and Mapping) algorithm is one of major methods to construct a 3-D map of a surrounding environment in robotic field. The onboard camera images and/or 3-D LIDAR point cloud data are used to construct a 3-D map and estimate the robot position simultaneously, and the algorithm is suitable for real-time mapping. We have developed the crawler-type robot which is the unmanned ground vehicle can run on irregular terrains. The robot has 3-D LIDAR and visible cameras, and the onboard sensor data can be applied to 3-D mapping by the SLAM. The SLAM algorithm usually utilizes the feature points such as edges of camera image and/or 3-D point cloud to construct a 3-D map. On the other hands, the featureless environments are often assumed in the disaster areas such as landslides after the earthquake and those around active volcanos. Therefore, we have performed the experiments of 3-D mapping using the facilities of the Fukushima Robot Test Field (RTF) in Minami-soma city, Fukushima. There are outdoor test fields that simulate the disaster areas such as landslides, cave-in and flooding in the RTF, and we can evaluate the accuracy of the 3-D map construction of the simulated terrains.
The [soil slope area] in the RTF has large slopes with 15 and 30 degrees, irregular ground and featureless surrounding environments. We have obtained the camera image and the 3-D LIDAR point cloud of the terrain using our developed robot and constructed the 3-D point cloud map from the LIDAR data. It has been found that the terrain was almost perfectly restored as the 3-D point cloud using the recent LIDAR SLAM algorithm. Then, we could provide the color information to the 3-D map by data fusion with the camera images. In this presentation, we will report the evaluation of the 3-D mapping of the irregular terrains and show the analysis results such as slope map. Then, we will also discuss combination of RTK/GNSS data and the SLAM for better positioning and 3-D map construction.
The SLAM (Simultaneous Localization and Mapping) algorithm is one of major methods to construct a 3-D map of a surrounding environment in robotic field. The onboard camera images and/or 3-D LIDAR point cloud data are used to construct a 3-D map and estimate the robot position simultaneously, and the algorithm is suitable for real-time mapping. We have developed the crawler-type robot which is the unmanned ground vehicle can run on irregular terrains. The robot has 3-D LIDAR and visible cameras, and the onboard sensor data can be applied to 3-D mapping by the SLAM. The SLAM algorithm usually utilizes the feature points such as edges of camera image and/or 3-D point cloud to construct a 3-D map. On the other hands, the featureless environments are often assumed in the disaster areas such as landslides after the earthquake and those around active volcanos. Therefore, we have performed the experiments of 3-D mapping using the facilities of the Fukushima Robot Test Field (RTF) in Minami-soma city, Fukushima. There are outdoor test fields that simulate the disaster areas such as landslides, cave-in and flooding in the RTF, and we can evaluate the accuracy of the 3-D map construction of the simulated terrains.
The [soil slope area] in the RTF has large slopes with 15 and 30 degrees, irregular ground and featureless surrounding environments. We have obtained the camera image and the 3-D LIDAR point cloud of the terrain using our developed robot and constructed the 3-D point cloud map from the LIDAR data. It has been found that the terrain was almost perfectly restored as the 3-D point cloud using the recent LIDAR SLAM algorithm. Then, we could provide the color information to the 3-D map by data fusion with the camera images. In this presentation, we will report the evaluation of the 3-D mapping of the irregular terrains and show the analysis results such as slope map. Then, we will also discuss combination of RTK/GNSS data and the SLAM for better positioning and 3-D map construction.