Robot Floor Plan Mapping from Overlapping Image Alignment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current mapping methods for autonomous robotic devices, such as SLAM using Extended Kalman Filter (EKF) techniques, face challenges including high computational power requirements, large memory needs, data association issues, and dependence on accurate measurement noise covariance matrices, which limit their efficiency and cost-effectiveness for consumer applications.
Innovation Solution
A method that involves obtaining raw pixel intensity values from overlapping fields of view using cameras, aligning sensor readings, and inferring environmental features to construct a floor plan, reducing computational costs by focusing on perimeter and object data rather than empty spaces, and using techniques like moving averages and k-nearest neighbors to improve accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If EKF-based SLAM methods are used to construct environmental maps, then mapping accuracy is improved, but computational power requirements and memory needs increase significantly
Solution Approach 1:
The patent extracts and processes only the most relevant features from sensor data (edges, corners, perimeter elements) rather than processing all raw data. This selective extraction reduces computational load while maintaining mapping accuracy by focusing on geometrically significant features that define the environment structure.
Solution Approach 2:
The mapping process is segmented into distinct stages: raw sensor data acquisition, feature detection, feature matching, and map construction. This segmentation allows each stage to be optimized independently, reducing overall computational requirements while preserving accuracy in the final map.
2Measurement precision
If EKF-based SLAM methods are used to construct environmental maps, then mapping accuracy is improved, but device complexity and implementation cost increase
Solution Approach 1:
The patent employs computationally inexpensive algorithms that can be implemented on low-cost microcontrollers rather than requiring sophisticated EKF-based systems. The approach uses simple geometric feature matching and probabilistic data association that are much less demanding in terms of hardware capabilities and implementation complexity.
Solution Approach 2:
The patent replaces complex mathematical filtering systems (EKF) with simpler geometric and probabilistic methods. By substituting the mechanical/mathematical complexity of EKF with more straightforward feature-based matching algorithms, the system achieves comparable accuracy with reduced implementation burden.
3Loss of information
If traditional mapping methods process all sensor data including empty spaces, then completeness of map is improved, but processing time and computational cost increase
Solution Approach 1:
The patent applies different processing strategies to different regions of the environment. High-detail feature processing is applied only to areas with relevant structures (walls, objects, perimeters), while empty spaces receive minimal or no processing. This local differentiation maintains map completeness for navigable areas while dramatically reducing processing time.
Solution Approach 2:
The patent processes only the necessary portion of sensor data required for effective mapping and navigation. Rather than processing all data uniformly, it selectively processes regions containing geometric features of interest, accepting that some minor details in empty spaces may be omitted without impacting overall map utility.
Data Source
AI summary
Provided is a process, including obtaining, with a robot, raw pixel intensity values of a first image and raw pixel intensity values of a second image, wherein the first image and the second image are taken from different positions; determining, with one or more processors, an overlapping area of a field of view of the first image and of a field of view of the second image by comparing the raw pixel intensity values of the first image to the raw pixel intensity values of the second image; spatially, with one or more processors, aligning values based on sensor readings of the robot based on the overlapping area; and inferring, with one or more processors, features of a working environment of the robot based on the spatially aligned sensor readings.