Embedded Files
Fully Automatic SLAM Using Hand-Drawn Maps in Indoor Environments
Fully Automatic SLAM Using Hand-Drawn Maps in Indoor Environments
In recent years, autonomous mobile robots have increasingly been introduced into indoor environments such as factories and airports for transportation, guidance, and other purposes. These robots use SLAM, or Simultaneous Localization and Mapping, which simultaneously estimates the robot's position and constructs a map. Providing prior information about the operating environment can improve mapping accuracy. In this research, we propose a SLAM method that uses hand-drawn maps, which can be easily created from human prior knowledge.
Improving SLAM by Using Potentially Dynamic Objects in Indoor Environments
Improving SLAM by Using Potentially Dynamic Objects in Indoor Environments
In this research, we improve SLAM in dynamic indoor environments by defining potentially dynamic objects in advance and using an object detector. In general, dynamic point clouds negatively affect self-localization accuracy and mapping results. Previous methods removed all objects that could potentially move, but this can cause degeneracy problems in environments with few features. We therefore newly define potentially dynamic objects and aim to improve the degeneracy problem by retaining as many point clouds as possible.
Detection of Approaching Objects at Indoor Corners Using Curved Mirrors for Collision Avoidance
Detection of Approaching Objects at Indoor Corners Using Curved Mirrors for Collision Avoidance
When autonomous mobile robots are used for transportation, cleaning, security, or similar tasks, they must be able to avoid collisions with people. In general, mobile robots do not move at very high speeds or accelerations for safety reasons, making it difficult to avoid people who suddenly appear from blind spots such as corners. We propose a system that detects approaching people in advance from images reflected in curved mirrors and uses the measured position and moving speed of the person for collision avoidance.
Three-Dimensional Human Tracking by Integrating Tracking Results from Two Cameras
Three-Dimensional Human Tracking by Integrating Tracking Results from Two Cameras
Tracking a specific person is one of the functions required for autonomous mobile robots in a wide range of applications. This function enables a robot to follow a specific person while carrying luggage or materials. We are building a method that performs target tracking using two cameras and integrates the results to obtain three-dimensional tracking information.
-関連論文-
本多 明彦, 梅田 和昇, "二台のカメラのトラッキングの融合による移動ロボットの3次元人物追跡," 日本機械学会ロボティクス・メカトロニクス講演会2018講演論文集, 1A1-M09, June 2018.
Conceptual View of Tracking and Tracking Results Observed from Two Cameras
Map Generation Considering the Physical Attributes of the Environment
Map Generation Considering the Physical Attributes of the Environment
Because it is difficult for people to enter damaged nuclear power plants, decommissioning work using robots has attracted attention. Environmental maps are essential for collecting environmental information, understanding the situation after a disaster, and controlling robots. In this research, we aim to generate environmental maps inside nuclear power plants using technologies such as SLAM. Damaged nuclear power plants may contain areas with distinctive physical characteristics, such as puddles and heat sources. By incorporating such information into SLAM, we aim to improve the accuracy of environmental maps.
Left-top image : Thermal Camera Image
Center image : 3D Map and Visualized Temperature Distribution of the Environment
Human Tracking by Mobile Robots: Robust Target Tracking Under Illumination Changes
Human Tracking by Mobile Robots: Robust Target Tracking Under Illumination Changes
We are developing a human-tracking system for applications such as luggage-carrying robots. The main challenge is that the target's features obtained from color images are easily affected by changes in illumination. Using a stereo camera, we combine color features and position features while changing their weights according to illumination conditions. This approach has enabled successful human tracking by a robot in real-world environments.
-関連論文-
Page updated
Report abuse