Color 3D Map Completion for Undetected Robot Areas
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing robots' color three-dimensional maps are incomplete due to restricted viewing angles and limited navigable regions, making it difficult for users to identify obstacles and specific objects in the robot's working environment, thereby degrading the display effect and user experience.
Innovation Solution
A method and apparatus for displaying a color three-dimensional map that involves acquiring initial map information, inferring structural and color information of undetected regions using algorithms like NERF-VAE, SPSG, and Scan2CAD, and rendering a completed map for display.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If the robot uses its sensor to directly detect and display the working environment, then the display process is simple and fast, but the displayed map is incomplete due to restricted viewing angles and limited navigable regions
Solution Approach 1:
The system performs preliminary actions by training a completion model in advance using large amounts of three-dimensional map data. During actual operation, the pre-trained model quickly infers and completes undetected regions, avoiding the need for complex real-time multi-angle sensing while achieving complete map coverage.
Solution Approach 2:
The patent introduces a completion model as an intermediary between the robot's sensor data and the final displayed map. This model acts as a bridge that infers missing information from partial observations, resolving the contradiction between simple sensing and complete information display.
2Loss of information
If the robot navigates to more regions to collect complete map data, then the map completeness improves, but the time and energy consumption increase significantly
Solution Approach 1:
The system creates a virtual copy or completion of the undetected regions by training an inference model on comprehensive three-dimensional map data. This virtual copying allows the system to display complete maps without physically navigating to all regions, saving time and energy.
Solution Approach 2:
The completion model is trained in advance on complete three-dimensional map data, performing the information gathering task preliminarily. During actual use, the pre-trained model instantly infers missing regions without requiring the robot to physically visit those areas, eliminating time loss.
3Reliability
If the robot displays only the initially detected regions, then the display process is fast and simple, but users cannot identify obstacles and specific objects in undetected regions
Solution Approach 1:
The completion model serves as an intermediary that processes the robot's partial sensor data and generates a complete representation of the working environment. This intermediary layer enables accurate obstacle identification in undetected regions without requiring complex real-time multi-sensor systems.
Solution Approach 2:
The system performs preliminary learning of environmental structures and obstacle patterns by training on complete three-dimensional map data. This preliminary action enables the model to accurately infer and identify obstacles in undetected regions without complex real-time analysis.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The present disclosure provides a color three-dimensional map display method and apparatus. The method comprises: acquiring initial color three-dimensional map information of a space where a robot is located, wherein the space where the robot is located represents a space corresponding to an area that a robot sensor can detect (S 101); complementing the color three-dimensional map on the basis of the initial color three-dimensional map information, so as to infer and complement structural information and color information of an area that is not detected by the robot sensor (S102); and displaying a complemented color three-dimensional map obtained after completion (S103).