Robot Obstacle Labeling Using Pose Interpolation and Confidence Maps
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Robots equipped with laser positioning struggle to accurately detect and label invisible obstacles like transparent glass or low-reflectivity objects, leading to inefficient navigation and potential collisions due to limitations in optical and physical collision sensors.
Innovation Solution
A method combining discrete positioning poses, pose interpolation, and positioning confidence coefficients to improve obstacle labeling accuracy by defining coverage areas, calculating confidence coefficients, and constructing closed graphs on a grid map to modify obstacle labels.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If physical collision sensors are used to detect invisible obstacles, then the robot can detect obstacles that optical sensors cannot detect, but the obstacle labeling accuracy remains low and navigation efficiency decreases
Solution Approach 1:
The patent introduces an intermediary processing system that combines laser positioning data with physical collision sensor data. The laser positioning system provides accurate position information before collision occurs, serving as a mediator between the optical detection phase and physical collision detection phase, thereby improving obstacle labeling accuracy while maintaining detection reliability
Solution Approach 2:
The patent applies preliminary action by using laser positioning to predict and pre-label obstacles before physical collision occurs. The system performs preliminary obstacle detection and labeling using laser data, then refines these labels when physical collision sensors confirm the obstacle, improving overall labeling accuracy without waiting for collision to occur
2Reliability
If physical collision sensors are used as the last detection method, then the robot can detect invisible obstacles, but the robot must label more area as obstacles to avoid repeated collisions, reducing navigation efficiency
Solution Approach 1:
The patent implements feedback by continuously comparing laser positioning data with physical collision sensor data. When physical collision occurs, the system feeds back this information to refine the obstacle labels, adjusting the labeled area to match the actual obstacle boundaries more closely, thereby improving navigation efficiency while maintaining detection reliability
Solution Approach 2:
The patent applies parameter changes by dynamically adjusting the obstacle label parameters (area, shape, position) based on the combination of laser positioning data and physical collision sensor data. The system modifies the labeled obstacle area to be more precise, reducing unnecessary conservative labeling while ensuring collision avoidance, thus improving navigation efficiency
Data Source
AI summary
The invention discloses a method for robots to improve the accuracy of an obstacle labeling, which comprises: making two positionings according to set moments, and then acquiring positioning poses of the two positionings on a grid map respectively at a first moment and a second moment; defining coverage areas of the first and the second moments according to positions of the two positionings, acquiring confidence coefficients of the two positionings, and processing the coverage areas through the confidence coefficients; interpolating the positioning poses, and constructing a closed graph according to the positioning poses, the pose interpolation, and the processed coverage areas; and obtaining a grid occupied by the closed graph on the grid map and modifying the obstacle labeling according to the grid occupied by the closed graph on the grid map and the area of the closed graph.