Robot cleaner for recognizing stuck situation through artificial intelligence and method of operating the same
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional robot cleaners struggle to accurately detect and adapt to changes in their environment, particularly in stuck situations, as they rely on pre-mapped locations and fail to recognize changes such as removed obstacles, leading to inefficient cleaning and potential re-stuck situations.
Innovation Solution
A robot cleaner that converts 3D image data and bumper events into surrounding map data, uses a stuck situation recognition model to infer stuck situations, and continuously updates this model through self-validation and error detection, allowing it to dynamically adjust its cleaning route and avoid obstacles effectively.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If the robot cleaner uses location-based stuck area recognition, then it can avoid stuck areas efficiently, but it cannot adapt when the environment changes (e.g., obstacles are removed)
Solution Approach 1:
The patent transitions from static location-based stuck area recognition to dynamic real-time environment recognition. The robot uses a recognition module to continuously detect current obstacles and dynamically determine stuck situations based on actual environmental conditions rather than relying on pre-stored location data, enabling adaptation to environmental changes.
Solution Approach 2:
The patent replaces the mechanical/memory-based location storage system with an AI-based recognition system. Instead of storing and retrieving stuck area locations from memory, the robot uses machine learning models to recognize stuck situations in real-time based on current sensor data and environmental perception, substituting static data storage with dynamic intelligent recognition.
2Adaptability or versatility
If the robot cleaner continuously updates its stuck area map, then it improves environmental adaptability, but it increases system complexity and computational load
Solution Approach 1:
The patent implements self-service through automatic error detection and self-validation mechanisms. The robot autonomously validates its own recognition results, detects errors in AI model predictions, and performs self-correction without external intervention. This reduces the need for complex external validation systems while maintaining high adaptability.
Solution Approach 2:
The patent incorporates feedback loops where the robot's recognition results are validated against actual environmental data and operational outcomes. Error detection mechanisms provide feedback to the AI model, enabling continuous improvement and refinement of the recognition system, which simplifies the overall system by using intelligent feedback rather than complex manual control.
3Measurement precision
If the robot uses AI model for stuck situation recognition, then it improves recognition accuracy, but it increases power consumption
Solution Approach 1:
The patent applies partial action by using AI model inference only when necessary for stuck situation determination, rather than continuous full-model operation. The system selectively activates the AI recognition model based on operational context, sensor data thresholds, and navigation needs, reducing overall computational load and power consumption while maintaining high recognition accuracy when required.
Data Source
Figure 1~2
Figure 3
Figure 4
AI summary
A robot cleaner for recognizing a stuck situation through artificial intelligence includes a driving unit to drive the robot cleaner, a sensing unit configured to acquire three-dimensional (3D) image data and a bumper event, a memory configured to store a stuck situation recognition model for inferring the stuck situation of the robot cleaner, and a processor configured to convert the 3D image data and the bumper event into surrounding map image data, infer the stuck situation of the robot cleaner from the 3D image data and the bumper event using the stuck situation recognition model, and control the driving unit according to an inference result.