Cleaning Robot Scene Identification for Automatic Task Execution

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing cleaning robots lack the ability to adapt to specific scenes and require manual user intervention for cleaning tasks, leading to complex user operations and delayed cleaning of stained zones.

Innovation Solution

A method and apparatus for controlling a robot that utilizes real-time image collection and scene identification to automatically execute cleaning tasks tailored to specific scenes, such as target site staining or object scattering, enhancing user experience by simplifying operations and ensuring timely cleaning.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Ease of operation

If cleaning robots execute fixed cleaning tasks based on manual user instructions, then user control over cleaning operations is maintained, but user operation complexity increases and cleaning timeliness decreases

Engineering Contradiction:
Improveuser operation simplicityVSAvoidmanual intervention requirement
Core Design Contradiction:
Ease of operationVSDevice complexity

Solution Approach 1:

The cleaning robot autonomously identifies specific scenes (staining scenes, scattering scenes) through image collection and scene identification models, automatically triggers corresponding cleaning tasks without manual user intervention, and executes the cleaning operations itself. This self-service capability directly resolves the contradiction by eliminating the need for complex manual operations while maintaining simple user experience.

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The system pre-establishes scene identification models trained on various staining and scattering scenarios, and pre-configures corresponding cleaning task parameters (cleaning modes, zones, times) for different scene types. When a specific scene is detected, the pre-prepared cleaning protocol is automatically activated, enabling rapid response without complex real-time decision-making or manual intervention.

Inventive Principle:
Principle #10Preliminary action

2Reliability

If cleaning robots wait for manual user instructions to clean stained zones, then cleaning operations can be controlled, but cleaning timeliness is delayed

Engineering Contradiction:
Improvecleaning timelinessVSAvoiddelay in cleaning execution
Core Design Contradiction:
ReliabilityVSLoss of time

Solution Approach 1:

The cleaning robot continuously collects real-time images of the environment, feeds these images into the scene identification model, and automatically triggers cleaning tasks when specific scenes (staining or scattering) are detected. This closed-loop feedback mechanism ensures immediate response to cleaning needs without waiting for manual user instructions, thereby improving cleaning timeliness and reducing time loss.

Inventive Principle:
Principle #23Feedback

Solution Approach 2:

The system pre-trains scene identification models on various staining and scattering scenarios and pre-configures cleaning task parameters for different scene types. This preliminary preparation enables the robot to immediately execute appropriate cleaning operations upon detecting a specific scene, eliminating the delay associated with waiting for manual user assessment and instruction.

Inventive Principle:
Principle #10Preliminary action

3Adaptability or versatility

If cleaning robots perform generic cleaning tasks, then device operation is simple, but adaptability to specific scenes is reduced

Engineering Contradiction:
Improvescene-specific cleaning capabilityVSAvoidscene identification system
Core Design Contradiction:
Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The cleaning robot system segments the cleaning domain into specific scene types (staining scenes, scattering scenes) and develops dedicated scene identification models and cleaning task parameters for each type. This segmentation allows the robot to adapt to different scenes with specialized protocols while maintaining overall system manageability, resolving the contradiction between adaptability and complexity.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The scene identification model serves multiple functions: it identifies both staining scenes and scattering scenes, determines appropriate cleaning modes, identifies cleaning zones, and estimates cleaning times. This multi-functionality enables the robot to adapt to various specific scenes using a single unified system, achieving high adaptability without proportionally increasing device complexity.

Inventive Principle:
Principle #6Universality (Multi-functionality)

Data Source

PatentEP4686443A1Method for controlling robot, computer program product, robot, and storage medium
Publication Date: 2026.02.04 BEIJING XIAOMI MOBILE SOFTWARE CO LTD
  • EP4686443A1 patent drawingFigure 1~3
  • EP4686443A1 patent drawingFigure 4~5
  • EP4686443A1 patent drawing

AI summary

The disclosure provides a method and apparatus for controlling a robot, a computer program product, a robot, and a storage medium. The method includes: (S101) obtaining an image collected in real time; (S102) determining whether a current scene is a specific scene according to the image collected in real time; and (S103) controlling, in a case where determining that the current scene is the specific scene, the robot to execute a cleaning task corresponding to the specific scene.