Robotic Lawn Mower Boundary Detection Without Wires or Localization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for defining the boundary of a mobile robot's workspace, such as robotic lawn mowers, require manual installation of wires or user-guided virtual boundaries, which are time-consuming, prone to damage, and require continuous localization data, limiting adaptability and ease of use.
Innovation Solution
The use of multimodal sensing capabilities, including RGB cameras, depth-perception devices, and localization modules, to automatically detect and map the boundary of a workspace, allowing for fully automated exploration and mowing without the need for manual wire installation or continuous user guidance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If manual wire installation is used to define workspace boundary, then boundary definition is reliable, but setup time and manpower increase significantly
Solution Approach 1:
The patent replaces the mechanical wire-based boundary system with an optical sensing system using RGB cameras and depth-perception devices. The robotic mower captures images and depth information to automatically detect and map boundaries, eliminating the need for manual wire installation while maintaining reliable boundary definition through multimodal sensing and neural network processing.
Solution Approach 2:
The system enables the robotic mower to autonomously perform boundary detection and mapping without human intervention. The onboard sensors automatically capture environmental data, the neural network processes the information to identify boundaries, and the system self-adjusts its workspace definition, eliminating the need for user-guided setup processes.
2Reliability
If wire-based boundary definition is used, then boundary is clearly defined, but the wire can be damaged causing system failure
Solution Approach 1:
The patent eliminates the physical wire component by substituting it with optical and depth-sensing systems. The boundary is defined through processed image and depth data rather than physical barriers, making the system immune to wire damage from mower contact or animal interference while maintaining clear boundary definition through semantic segmentation.
3Measurement precision
If user-guided virtual boundary method is used, then boundary detection is achieved, but continuous localization data is required increasing system complexity
Solution Approach 1:
The patent extracts and removes the dependency on continuous localization data from the boundary detection process. Instead of relying on localization modules to track position for virtual boundary definition, the system uses onboard sensors to directly perceive and map boundaries in the environment, making boundary detection independent of localization system complexity.
4Reliability
If traditional boundary teaching process is used, then boundary is established, but it requires significant setup time and retraining when layout changes
Solution Approach 1:
The patent implements a dynamic boundary detection system that continuously adapts to environmental changes. The neural network processes real-time sensor data to automatically update boundary definitions when layout changes occur, eliminating the need for retraining processes while maintaining accurate boundary establishment through ongoing environmental perception and semantic segmentation.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach significantly reduces setup time and manpower, adapts to changes in the environment, and eliminates the need for continuous localization data, enabling efficient and adaptive boundary detection and mapping for mobile robots.
Implementation Method 1
depth-perception device (e.g., stereo cameras, time of flight (ToF) sensors, Lidar, etc.)
Implementation Method 2
capturing, by a camera device on the mobile robot, an image frame at the first location within the portion of the environment
Data Source
AI summary
The method and system disclosed herein presents a method and system for capturing, by a depth-perception device on a mobile robot moving in an environment, depth-perception information at a first location within a portion of the environment. The method includes capturing, by a camera device on the mobile robot, an image frame at the first location within the portion of the environment, generating a three-dimensional point cloud representing the portion of the environment based on the depth-perception information and the image frame, extracting semantic information by providing the frame and the point cloud to a neural network model configured to perform semantic segmentation, associating the semantic information to the three-dimensional point cloud, projecting the three-dimensional point cloud onto a two-dimensional plane to form a first two-dimensional map; and demarcating the first two-dimensional map with a plurality of different regions based on the semantic information.