Robot State Estimation Using Landmark Candidate Areas
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous robots using SLAM struggle to estimate their position accurately due to the limited observable range and directivity of sensors like RGB cameras and range finders, leading to increased uncertainty and processing complexity, which can hinder stable autonomous movement.
Innovation Solution
A moving object controller with an observation obtaining unit, landmark detection unit, candidate area obtaining unit, state estimation unit, and storage unit processes image and distance data to narrow down candidate positions using particle filtering and environmental map updates, effectively estimating the robot's position with high accuracy and speed by utilizing landmark information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If an RGB camera or range finder is used to obtain observation data, then the robot can acquire landmark information, but the position estimation accuracy deteriorates due to limited observable range and directivity
Solution Approach 1:
The patent divides the environment into multiple sectors around the robot and assigns different observation tasks to different sensors based on their directional characteristics. The RGB camera focuses on detecting landmark features within its field of view, while the range finder measures distances to landmarks in specific directions. This segmentation of observation tasks resolves the contradiction by enabling accurate position estimation through coordinated multi-sensor data fusion without requiring any single sensor to have omnidirectional capability.
2Loss of information
If multiple landmarks are arranged around the robot, then more observation data can be obtained, but the number of candidate positions increases making position estimation more difficult
Solution Approach 1:
The patent implements a feedback mechanism where the robot's current position estimate is used to predict which landmarks should be visible and at what positions. This predicted landmark configuration is then compared with actual observations to evaluate candidate positions. The feedback loop eliminates inconsistent candidates and refines the position estimate, resolving the contradiction by systematically reducing the candidate position space even when multiple landmarks are present.
3Loss of information
If the robot uses image data from RGB camera to determine landmark positions, then visual information can be obtained, but the actual position of the robot cannot be easily estimated with high accuracy
Solution Approach 1:
The patent merges data from multiple sources: image data from the RGB camera for landmark detection, distance data from the range finder for radial positioning, and angular information from the sector-based observation model. By combining these different types of observations through data fusion, the system overcomes the limitation of using image data alone and achieves high-accuracy robot position estimation that resolves the contradiction between visual information quality and position accuracy.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A landmark detection unit obtains a distance between the moving object and each of two or more landmarks as landmark distance information based on observation data obtained by an observation obtaining unit. A candidate area obtaining unit determines a candidate area for a position of the moving object based on the landmark distance information obtained by the landmark detection unit, and obtains candidate area information indicating the determined candidate area. A state estimation unit estimates an internal state of the moving object based on the observation data, the landmark distance information, and the candidate area information to obtain moving object internal state estimation data, and estimates the environmental map based on the candidate area information and the landmark distance information to obtain environmental map data.