Robot Vision Positioning for Repeating Indoor Structures
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Robots operating in dynamic environments like airports and large public spaces face challenges in accurately setting their position due to repeating structures and limitations of GPS information, leading to reduced positioning accuracy.
Innovation Solution
A method and robot system that utilize vision information to set an initial position, combining lidar sensing results with vision-based candidate positions to enhance accuracy and speed, by controlling sensors to generate and match image and lidar information with stored map data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If GPS information is used for positioning in indoor spaces, then positioning can be provided, but positioning accuracy is reduced due to limitations of GPS in indoor environments
Solution Approach 1:
The patent replaces GPS-based positioning with a vision-based positioning system that uses image recognition and matching techniques. The system captures images of the environment, extracts feature points, and matches them with pre-stored map data to determine position, thereby eliminating reliance on GPS which does not function properly in indoor spaces.
Solution Approach 2:
The patent introduces vision information and feature point matching as an intermediary mechanism between the robot and the environment. Instead of directly using GPS signals, the system uses visual features of the environment as intermediaries to infer position, enabling accurate indoor positioning through image-based recognition and comparison with map data.
2Measurement precision
If only structure information from maps is used for position setting, then positioning can be performed, but positioning accuracy is reduced in spaces with repeating structures
Solution Approach 1:
The patent transitions from two-dimensional map structure comparison to three-dimensional visual feature matching by incorporating depth information and spatial relationships. The system uses stereo vision or depth sensors to capture 3D environmental data, which is then matched with 3D map representations, adding a dimensional aspect that resolves ambiguities in repeating structures.
Solution Approach 2:
The patent utilizes visual features such as color, texture, and appearance variations in the environment as distinguishing characteristics. By analyzing color patterns and visual properties of objects and surfaces, the system can differentiate between repeating structures that appear identical in simple 2D maps, thereby improving position setting accuracy in corridors and rectangular spaces.
3Measurement precision
If vision information is used to set initial position, then positioning accuracy is improved, but processing time increases due to image analysis requirements
Solution Approach 1:
The patent performs preliminary processing by pre-storing processed map data with extracted feature points and visual characteristics before the robot enters the environment. This pre-processing creates a ready-to-match reference database, so that during actual positioning, the system only needs to compare current sensor data against pre-processed maps, significantly reducing real-time computation time while maintaining high accuracy.
Solution Approach 2:
The patent extracts only the most salient and distinctive visual features from the environment for positioning purposes, rather than processing entire images. By identifying and extracting key feature points, edges, corners, and distinctive visual markers, the system reduces the data volume requiring processing while preserving sufficient information for accurate position determination.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
Enables the robot to quickly and accurately set its initial position using vision information, improving navigation and operation in large, complex spaces by reducing reliance on GPS and enhancing position recognition.
Implementation Method 1
controlling a vision sensor of a sensor module of the robot to generate first vision information using image information of a sensed object
Implementation Method 2
controlling a lidar sensor of the sensor module to generate an around map using lidar information of a sensed object
Data Source
AI summary
The present invention relates to a method for positioning on the basis of vision information and a robot implementing the method. The method for positioning on the basis of vision information, according to an embodiment of the present invention, comprises the steps of: generating, by a control unit of a robot, first vision information by using image information of an object sensed by controlling a vision sensing unit of a sensor module of the robot; generating, by the control unit of the robot, a vision-based candidate position by matching the first vision information with second vision information stored in a vision information storage unit of a map storage unit; and generating, by the control unit, the vision-based candidate position as the position information of the robot when there is one vision-based candidate position.