Indoor Robot Pose Estimation Through Topology Map Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The initial positioning of robots indoors is challenging due to the inefficiency of existing methods, which require traversing large datasets and are affected by lighting, rotation angles, and moving objects, leading to low efficiency and accuracy.
Innovation Solution
The robot employs a topology map matching method, where a current local topology map is matched with a pre-established full topology map to quickly and accurately determine its pose by selecting the search branch with the highest matching degree, using node associations and semantic segmentation to construct and compare maps.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the robot uses traditional positioning methods that traverse all data in the database, then the positioning can be performed without prior location information, but the positioning efficiency decreases significantly
Solution Approach 1:
The patent divides the positioning process into two distinct phases: an offline phase where a full topology map is pre-built and stored in the database, and an online phase where the robot performs rapid positioning by matching its current local topology map against the pre-built full map. This segmentation allows the computationally intensive map construction to be done once beforehand, while the actual positioning operation becomes much faster by only comparing against the stored topological structure rather than traversing all raw sensor data.
Solution Approach 2:
The patent performs preliminary action by pre-establishing the full topology map of the environment before the robot needs to perform positioning. The topology map includes all semantic information about objects, their positions, and relationships in the environment. When the robot needs to position itself, it only needs to match its current local topology map against this pre-built full map, avoiding the need to traverse and process all original sensor data during the positioning operation.
2Adaptability or versatility
If the robot uses traditional positioning methods, then it can work in various lighting conditions, but the positioning accuracy is affected by lighting, rotation angles, and moving objects
Solution Approach 1:
The patent replaces traditional vision-based positioning methods that rely on image matching and feature recognition with a topology-based positioning approach. Instead of comparing images that are sensitive to lighting, rotation, and moving objects, the system constructs and compares topological maps that represent the structural relationships between objects. This substitution of the positioning mechanism from image-based to topology-based methods eliminates the sensitivity to environmental variations while maintaining adaptability to different indoor environments.
Solution Approach 2:
The patent changes the parameters used for positioning from visual features (image pixels, colors, textures) to topological parameters (object relationships, spatial connections, semantic structures). By transforming the positioning problem from one of image matching to one of topological map matching, the system achieves immunity to lighting conditions, rotation angles, and moving objects while maintaining the ability to adapt to various indoor environments.
3Measurement precision
If the robot builds a detailed full topology map with semantic segmentation, then the positioning accuracy improves, but the computational complexity and data processing requirements increase
Solution Approach 1:
The patent segments the complex task of environment understanding into two parts: semantic segmentation for identifying and classifying objects in the environment, and topology construction for establishing relationships between these objects. By separating these functions, the system can build a rich semantic topology map that captures both what objects are present and how they are spatially related, enabling accurate positioning without requiring the robot to simultaneously perform all processing operations.
Solution Approach 2:
The patent performs the computationally intensive semantic segmentation and topology map construction as a preliminary action during the offline phase. The full topology map with all semantic information about objects and their relationships is built beforehand and stored in the database. During online positioning, the robot only needs to perform lightweight matching operations against this pre-built map, significantly reducing the computational complexity during actual positioning operations.
Data Source
AI summary
The present disclosure provides a positioning method, including: obtaining a current local topology map established based on objects in an environment currently observed by a robot, obtaining a full topology map pre-established based on objects in a full environment in a preset area, the current local topology map and the full topology map including nodes representing the objects; matching a node pair to be associated constructed by two nodes in the current local topology map and two nodes in the full topology map; if a degree of association of the node pair to be associated is greater than a threshold, determining that the node pair to be associated is an associated node pair; and determining a pose of the robot according to one of a plurality of search ranches, with a largest number of associated node pairs.