Robot Lawnmower Local Mapping With Semantic Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Robot lawnmowers often require complex systems and sensors to navigate effectively, leading to increased costs without a balance between functionality and complexity.
Innovation Solution
A computer-implemented method using an imaging sensor to apply semantic segmentation and perspective transformation to create a local map, allowing the robot lawnmower to navigate by classifying regions as grass or non-grass, with additional sensors like odometry and IMU data for efficient memory management and real-time updates.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If complex sensors and navigation systems are used, then navigation accuracy is improved, but device complexity and cost increase
Solution Approach 1:
The patent divides the navigation system into modular components: imaging sensor for visual data collection, semantic segmentation algorithm for feature identification, perspective transform for viewpoint correction, and local map construction for spatial representation. Each module performs a specific function, allowing the system to achieve accurate navigation through coordinated simple components rather than a single complex sensor system.
Solution Approach 2:
The patent replaces complex mechanical navigation systems (such as laser rangefinders, radar, and multiple specialized sensors) with a vision-based system using standard imaging sensors and computer vision algorithms. The imaging sensor captures images that are then processed through semantic segmentation and perspective transformation to extract navigation-relevant information, substituting mechanical sensing with optical sensing and computational processing.
2Reliability
If a complete local map is stored in memory, then navigation reliability is improved, but memory requirements increase
Solution Approach 1:
The patent constructs a local map that focuses specifically on navigation-relevant features (grass regions, non-grass obstacles, boundaries) within the robot's operational vicinity rather than storing complete environmental data. The map is built by transforming images to an overhead viewpoint and segmenting relevant regions, creating a compact representation that contains only the information necessary for reliable navigation decisions.
Solution Approach 2:
The patent extracts only the essential navigation information from captured images by applying semantic segmentation to identify and separate grass regions, non-grass features, and boundaries. This extraction process removes irrelevant visual data, retaining only the critical features needed for navigation while significantly reducing the data volume that must be stored in memory.
Data Source
AI summary
A robot lawnmower and method for controlling the robot lawnmower based on the generation of a local map. The method including receiving an image from an imaging sensor onboard the robot lawnmower, the image including an area of ground in an upcoming path, applying a semantic segmentation algorithm to produce a segmented image from the received image, the segmented image including regions corresponding to features in the image, applying a perspective transform to the segmented image to obtain an overhead view transformed image, wherein the regions are preserved in the transformed image, determining, from the transformed image, positions of the regions relative to the current position of the robot lawnmower, plotting a local map of the environment of the robot lawnmower based on positions of the regions relative to a current position of the robot lawnmower; and controlling the robot lawnmower to navigate a lawn area using the local map.