A multi-robot cooperative cleaning system and a scheduling method thereof

Abstract

The application discloses a multi-robot cooperative cleaning system and a scheduling method thereof, and relates to the technical field of intelligent cleaning robots, which comprises a cleaning robot cluster, a cleaning management and scheduling system, a large language model and an artificial interaction module. The robot cluster is composed of standard cleaning robots with environment perception and problem reporting functions, special robots for processing special tasks, and modular robots capable of automatically replacing cleaning modules. The cleaning management and scheduling system is deployed on an edge or a cloud server, and comprises an execution control, an intelligent analysis, a strategy library and a large language model interface, and is responsible for analyzing the information reported by the robots, calling the large language model for analysis, and generating a cleaning strategy. The application realizes multi-robot intelligent cooperation and dynamic scheduling, and improves the cleaning efficiency and the complex problem processing capability.

Description

Technical Field

[0001] This invention relates to the field of intelligent cleaning robot technology, specifically a multi-robot collaborative cleaning system and its scheduling method. Background Technology

[0002] With the popularization of intelligent robotic vacuum cleaners, their technology has evolved from early random collision-based cleaning to a stage with intelligent navigation, path planning, and automatic cleaning functions. Current mainstream technologies typically employ a navigation scheme that integrates LiDAR and visual sensors, and are equipped with multiple sensors to achieve functions such as obstacle avoidance, fall prevention, and stain recognition.

[0003] However, existing technologies still have significant limitations. In complex home or commercial environments, a single cleaning robot cannot handle all cleaning tasks: for standard tasks such as ordinary dust and stains, sweeping or mopping robots can complete the task; but for special tasks such as large pieces of trash, oil stains, and sticky stains, existing robots are usually powerless and still require human intervention. Although it is possible to handle more scenarios by enhancing the functions of individual robots (such as adding robotic arms, specialized cleaning solutions, etc.), this would lead to complex robot structures, high costs, and many special functions are low-frequency applications, making integration into a single body uneconomical.

[0004] Therefore, there is an urgent need for a multi-robot collaborative cleaning system and its scheduling method to solve the above problems. Summary of the Invention

[0005] The purpose of this invention is to provide a multi-robot collaborative cleaning system and its scheduling method to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: A multi-robot collaborative cleaning system includes: a cluster of cleaning robots, a cleaning management and scheduling system, a large language model (LLM), and a human interaction module; The cleaning robot cluster includes standard cleaning robots, special cleaning robots, and modular robots. The standard cleaning robots are equipped with environmental perception sensors, enabling them to perform daily cleaning tasks, identify cleaning problems, and report relevant information. The special cleaning robots have special cleaning functions to handle cleaning tasks that the standard cleaning robots cannot handle. The modular robots include a standardized chassis and automatically replaceable cleaning modules, which can switch cleaning functions by automatically loading different cleaning modules. The cleaning management and scheduling system is deployed on an edge server or cloud server and includes an execution control unit, an intelligent analysis unit, a strategy library, and an external large language model interface. The large language model is invoked by the cleaning management and scheduling system through an external large language model interface to assist in problem analysis and cleaning strategy generation. The manual interaction module is a dedicated terminal device used to receive notifications from the cleaning management and scheduling system and enable manual intervention.

[0007] As a further aspect of the present invention: the execution control unit in the cleaning management and scheduling system directly manages and schedules the cleaning robot cluster, issues execution instructions, and monitors the task execution status.

[0008] As a further aspect of the present invention: the intelligent analysis unit of the cleaning management and scheduling system can parse the parameterized description of the problems reported by the robot and generate problem description text, or perform in-depth problem analysis based on the raw data uploaded by the robot, and generate a cleaning strategy that includes the calling object, work objective and work method in combination with the strategy library.

[0009] As a further aspect of the present invention: the strategy library is used to store historical cleaning strategies, the intelligent analysis unit stores successfully executed cleaning strategies into the strategy library, and the strategy library can be optimized and adjusted manually, supporting strategy retrieval, optimization and reuse.

[0010] As a further aspect of the present invention, the information reported by the robot includes: the time of the problem, location coordinates, event type, priority, detailed description of the event, and raw data, wherein the raw data includes images, videos, and 3D point cloud data.

[0011] As a further aspect of the present invention: the process of switching cleaning modules by the modular robot does not require manual intervention, and it can automatically switch to the working mode that matches the module after loading the cleaning module.

[0012] A multi-robot collaborative cleaning method includes the following steps: S1: Standard cleaning robot performs daily cleaning tasks, monitors the cleaning area in real time, and identifies cleaning problems; S2: When a standard cleaning robot encounters a problem it cannot handle, it will report the identified cleaning problem information or the raw data that it cannot identify to the cleaning management and scheduling system. S3: The cleaning management and scheduling system calls the large language model, combines the strategy library to analyze the reported information, and generates corresponding cleaning strategies; S4: The execution control unit schedules the corresponding cleaning robot to perform cleaning tasks according to the cleaning strategy, and the modular robot automatically switches to the appropriate cleaning module; S5: The cleaning management and scheduling system monitors the task execution results and updates successful cleaning strategies to the strategy library; if the existing robots cannot solve the cleaning problem, human intervention is notified through the human interaction module.

[0013] Compared with the prior art, the beneficial effects of the present invention are: More comprehensive cleaning coverage and more flexible scene adaptation: Through the hierarchical configuration of "standard cleaning robot + special cleaning robot + modular robot", it not only meets the basic needs of daily periodic cleaning, but also can accurately deal with special cleaning tasks such as large debris and stubborn oil stains, effectively solving the problem of cleaning dead corners and complex obstacle handling; the modular design takes into account the diversity of functions and the utilization rate of equipment, avoiding the redundancy and waste of the functions of single robot integration, and adapting to a variety of complex scenarios such as home and commercial office.

[0014] More accurate problem identification and more intelligent strategy generation: By leveraging the data acquisition capabilities of multi-sensor systems and the deep analysis capabilities of large language models, the accuracy of problem identification in complex scenarios has been significantly improved, reducing false and missed identification. The collaborative application of the strategy library and the large language model not only enables the reuse of historical successful experiences and improves the efficiency of strategy generation, but also provides the innovative ability to deal with new scenarios and new problems, ensuring the relevance and effectiveness of cleaning strategies.

[0015] Higher degree of automation and less human intervention: A closed-loop mechanism of "problem identification - strategy generation - task execution - effect feedback" has been built, realizing automated operation from daily cleaning and problem reporting to special task handling; human intervention is only required when existing robots cannot solve the problem, minimizing human intervention for users and truly achieving a "seamless" cleaning experience.

[0016] The system boasts enhanced scalability and lower maintenance costs: the cleaning management and scheduling system supports both edge and cloud deployment modes, allowing for flexible adjustments based on the scale of the scenario; the strategy library supports automatic accumulation and manual optimization, continuously improving system performance as the number of usage scenarios increases; and the modular robot design reduces equipment maintenance costs, enabling functional upgrades simply by replacing modules without replacing the entire machine. Attached Figure Description

[0017] Figure 1 This is an architectural diagram of a multi-robot collaborative cleaning system according to an embodiment of the present invention.

[0018] Figure 2 This is a table explaining the parameters for reporting cleaning problems in this embodiment of the invention.

[0019] Figure 3 This is a flowchart illustrating the multi-robot collaborative cleaning task execution process in an embodiment of the present invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Example 1: Independent Configuration Scheme for Multiple Robots The system consists of the following components: Cleaning robot swarm: Standard cleaning robots (3 units): They adopt a sweeping and mopping design, and are equipped with visual sensors, lidar, dirt sensors and infrared anti-fall sensors. They have daily vacuuming and mopping functions, can detect the dust concentration and cleaning effect on the ground in real time, and identify and report cleaning problems that cannot be handled; the positioning accuracy is ±2cm, and they support multi-robot collaborative path planning to avoid overlapping operations.

[0022] Special cleaning robots (2 units): including 1 foreign object grasping robot and 1 oil stain removal robot; the foreign object grasping robot is equipped with a robotic arm assembly (maximum grasping weight 2kg), a high-definition image sensor and an ultrasonic proximity detection sensor, which can identify and grasp large debris such as toys, documents, and slippers; the oil stain removal robot is equipped with an alkaline cleaning solution storage tank (capacity 5L), a high-pressure spray assembly (maximum spray pressure 0.3MPa) and a rotating scrubbing module (rotation speed 300 rpm), which is suitable for cleaning oil stains on floor materials such as tiles and marble.

[0023] Human-computer interaction module: A dedicated mobile APP is used, which supports Android and iOS systems. It can receive system push problem reports (including location coordinates, on-site pictures and problem descriptions). Staff can view task progress and provide feedback on processing results through the APP.

[0024] Cleaning management and scheduling system: Deployed on a local edge server, it interacts with each robot via a Wi-Fi 6 communication network. The core components of the system include: Execution control unit: Supports simultaneous scheduling of up to 10 robots, can receive robot status feedback in real time, and issue commands such as coordinate positioning, task start / pause, and mode switching.

[0025] Intelligent Analysis Unit: Built-in parameter parsing algorithm to convert structured data reported by the robot into natural language problem description; the response time of the search strategy library is ≤0.5 seconds. When no similar strategy is matched, the GPT-4 large language model is called to assist in generating a strategy, with a generation time of ≤3 seconds.

[0026] Strategy Library: Pre-stores common cleaning scenario strategies, categorized by "cleaning type - area material - priority", supports quick retrieval and batch optimization, and allows for manual addition, modification and deletion of strategies through the backend.

[0027] Large Language Model: It adopts an external interface call mode to communicate with the cleaning management and scheduling system, and assists in completing in-depth analysis of complex problems and generating innovative strategies.

[0028] Work process: Three standard cleaning robots start daily cleaning tasks according to preset paths, respectively responsible for sweeping and mopping different areas such as office area, meeting room, and corridor. The cleanliness of the floor is monitored in real time by dirt sensor, and abnormal situations are identified by visual sensor during the cleaning process.

[0029] One of the standard cleaning robots detected an insoluble oily stain on the conference room floor and immediately recorded the problem parameters: reporting time "2025.10.20; 13:30:46", coordinates "123,134", event type "oil stain", priority "medium", and detailed description "the stain area is about 0.05㎡, the surface is sticky, and it is located under the conference table". It also took 3 high-definition pictures (before cleaning) and uploaded them to the cleaning management and scheduling system.

[0030] After receiving the data, the intelligent analysis unit of the cleaning management and scheduling system first parses the parameters to generate a problem description text: "There is approximately 0.05㎡ of viscous oily stains under the conference table in the conference room (coordinates 123, 134), which requires deep cleaning." Then, it searches the strategy library and matches the strategy "treatment of cooking oil stains on the ceramic tile surface of the conference room." At the same time, it calls the large language model to verify the adaptability of the strategy and finally generates the execution strategy: "Dispatch the oil stain treatment robot to coordinates 123, 134, and use the 'high-pressure spray + 30-second rest + rotation scrubbing' mode to treat the oil stains. After cleaning, take comparison pictures through the image sensor to verify that the stain removal rate is ≥95%."

[0031] The execution control unit sends the strategy to the oil stain removal robot. The robot navigates to the target location using LiDAR (navigation accuracy ±2cm), activates the high-pressure spray component to spray alkaline cleaning solution, and after standing for 30 seconds, activates the rotary scrubbing module to clean continuously for 2 minutes.

[0032] After cleaning is completed, the oil stain removal robot takes three high-resolution images and uploads them to the system. The intelligent analysis unit compares the images before and after cleaning using a grayscale analysis algorithm and determines that the stain removal rate is 98%, which meets the cleaning requirements. The task is deemed successful, and the execution strategy is optimized and updated to the strategy library. The standard cleaning robot continues to perform the cleaning task for the remaining areas.

[0033] If the oil stain removal robot only achieves a 75% stain removal rate (not meeting the standard) after cleaning, the cleaning management and scheduling system will send a problem report to the staff via a mobile app, notifying them to intervene manually. After the staff has finished cleaning, they will report "Cleaned with special cleaning agent" via the app. The system will then save the manual handling solution to the strategy library for reference in similar scenarios in the future.

[0034] Example 2: Modular Robot Configuration Scheme The system consists of the following components: Cleaning robot cluster: Employs one modular robot, including a standardized chassis and three automatically switchable cleaning modules. Standardized chassis: Equipped with lidar, vision sensor, infrared anti-fall sensor and navigation and positioning module, it supports automatic return to the module storage base station for module switching. The switching process does not require manual intervention. The mechanical interface adopts a foolproof design, the electrical interface supports hot-swapping, and the single switching time is ≤60 seconds.

[0035] Cleaning modules include Module 1 (standard cleaning module, integrating a vacuum cleaner fan, mop roller and dirt sensor to realize daily sweeping and mopping functions), Module 2 (object grabbing module, including a robotic arm, clamping components and image recognition sensor, which can grab objects weighing ≤1kg), and Module 3 (oil stain treatment module, including a small cleaning liquid storage tank, low-pressure spray components and scrubbing components, suitable for wood flooring, tile and other common household floor materials).

[0036] Human-computer interaction module: It adopts a combination of "mobile APP + smart bracelet". The smart bracelet is used to receive vibration alerts for emergency problems, and the mobile APP displays detailed problem information, processing progress and history.

[0037] Cleaning management and scheduling system: Deployed on a cloud server, it communicates with modular robots via home Wi-Fi. The core components are the same as in Example 1. The policy library pre-stores home scene cleaning policies and supports automatic optimization of policy priorities based on user habits.

[0038] Large Language Model: Calls domestic open-source large language models (such as Tongyi Qianwen), communicates with the system through API interfaces, and adapts to the problem analysis needs of complex household cleaning scenarios.

[0039] Working process: The modular robot, equipped with module 1 (standard cleaning module), starts daily cleaning tasks, traversing areas such as the living room, bedroom, and kitchen according to a preset path, and identifying abnormal situations during the cleaning process through visual sensors.

[0040] The robot detected a child's toy (building blocks, 5cm x 3cm) in the bedroom area, determined it to be a solid foreign object that it could not handle, and then reported the following parameters: reporting time "2025.10.20; 14:15:22", coordinates "118,142", event type "solid obstacle (toy)", priority "high", detailed description "movable foreign object, no sharp edges", and uploaded 3D point cloud data.

[0041] After the intelligent analysis unit of the cleaning management and scheduling system parses the data, it calls the large language model to generate a strategy: "The modular robot returns to the module storage base station, switches to module 2 (foreign object grasping module), goes to coordinates 118,142 to grasp the building block and place it in the living room storage box (coordinates 105,120), and after completion, switches back to module 1 (standard cleaning module) to continue to perform the remaining cleaning tasks."

[0042] The modular robot executes the strategy instructions issued by the control unit and returns to the base station via LiDAR navigation (navigation time is about 2 minutes). The base station automatically unlocks module 1 and loads module 2. After loading is completed, the robot automatically switches to the grasping mode and heads to the target coordinates.

[0043] The modular robot uses image recognition sensors to accurately locate the building blocks (positioning error ≤1cm), the robotic arm extends to grip the building blocks (gripping force is adjustable to avoid damaging the objects), moves them to the living room storage box and releases them, then returns to the base station to switch back to module 1 to continue the unfinished daily cleaning tasks.

[0044] The cleaning management and scheduling system monitors the module switching process and task execution results in real time. After confirming that the building blocks are successfully stored, the strategy is stored in the strategy library. If the module switching fails or the grabbing fails, the system immediately notifies the user to handle the issue via wristband vibration and APP. After the user handles the issue and provides feedback, the system updates the strategy library.

[0045] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

[0046] Furthermore, it should be understood that although this specification describes embodiments, not every embodiment contains only one independent technical solution. This narrative style is merely for clarity. Those skilled in the art should consider the specification as a whole, and the technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.

Claims

1. A multi-robot collaborative cleaning system, characterized in that, include: Cleaning robot clusters, cleaning management and scheduling system, large language model and human interaction module; The cleaning robot cluster includes standard cleaning robots, special cleaning robots, and modular robots. The standard cleaning robots are equipped with environmental perception sensors, enabling them to perform daily cleaning tasks, identify cleaning problems, and report relevant information. The special cleaning robots have special cleaning functions to handle cleaning tasks that the standard cleaning robots cannot handle. The modular robots include a standardized chassis and automatically replaceable cleaning modules, which can switch cleaning functions by automatically loading different cleaning modules. The cleaning management and scheduling system is deployed on an edge server or cloud server and includes an execution control unit, an intelligent analysis unit, a strategy library, and an external large language model interface. The large language model is invoked by the cleaning management and scheduling system through an external large language model interface to assist in problem analysis and cleaning strategy generation. The manual interaction module is a dedicated terminal device used to receive notifications from the cleaning management and scheduling system and enable manual intervention.

2. The multi-robot collaborative cleaning system according to claim 1, characterized in that, The execution control unit in the cleaning management and scheduling system directly manages and schedules the cleaning robot cluster, issues execution instructions, and monitors task execution.

3. The multi-robot collaborative cleaning system according to claim 2, characterized in that, The intelligent analysis unit of the cleaning management and scheduling system can parse the parameterized description of the problems reported by the robot and generate problem description text, or perform in-depth problem analysis based on the raw data uploaded by the robot, and generate cleaning strategies that include calling objects, work objectives and work methods in combination with the strategy library.

4. The multi-robot collaborative cleaning system according to claim 1, characterized in that, The strategy library is used to store historical cleaning strategies. The intelligent analysis unit stores successfully executed cleaning strategies into the strategy library, and the strategy library can be optimized and adjusted manually, supporting strategy retrieval, optimization and reuse.

5. A multi-robot collaborative cleaning system according to claim 1, characterized in that, The information reported by the robot includes: the time of the problem, location coordinates, event type, priority, detailed description of the event, and raw data, including images, videos, and 3D point cloud data.

6. The multi-robot collaborative cleaning system according to claim 1, characterized in that, The modular robot can switch cleaning modules without human intervention. After loading the cleaning module, it can automatically switch to the working mode that matches the module.

7. A multi-robot collaborative cleaning method, characterized in that, Includes the following steps: S1: Standard cleaning robot performs daily cleaning tasks, monitors the cleaning area in real time, and identifies cleaning problems; S2: When a standard cleaning robot encounters a problem it cannot handle, it will report the identified cleaning problem information or the raw data that it cannot identify to the cleaning management and scheduling system. S3: The cleaning management and scheduling system calls the large language model, combines the strategy library to analyze the reported information, and generates corresponding cleaning strategies; S4: The execution control unit schedules the corresponding cleaning robot to perform cleaning tasks according to the cleaning strategy, and the modular robot automatically switches to the appropriate cleaning module; S5: The cleaning management and scheduling system monitors the execution effect of tasks and updates successful cleaning strategies to the strategy library; If the existing robots cannot solve the cleaning problem, human intervention will be notified through the human interaction module.

Images