Cleaning robot and task self-learning method and control method thereof

Abstract

The present application relates to the field of cleaning robots, and discloses a cleaning robot and its task self-learning method and control method. The task self-learning method includes: after receiving a task instruction, obtaining the current sensor data of the cleaning robot in real time; wherein the task instruction includes a cleaning task, a transport task or a return to a base station task; based on the sensor data, the task model controls the cleaning robot to perform corresponding actions; the task model performs behavioral learning based on the performance feedback of the cleaning robot in performing the action, until the task model is guided to control the cleaning robot to complete the task corresponding to the task instruction and achieve the task goal corresponding to the task. The present application increases the autonomous decision-making ability of the cleaning robot by performing task autonomous learning on the cleaning robot, so that the cleaning robot can autonomously complete complex tasks.

Description

Technical Field

[0001] The present invention relates to the field of cleaning robots, and in particular to a cleaning robot and a task self-learning method and a control method thereof. Background Art

[0002] For cleaning robots, it is usually necessary to use manually written algorithms and rules to train them in advance or input control so that the cleaning robots can complete their tasks. However, the existing manually written algorithms and rules cannot flexibly adapt to complex and changing environments, which makes the cleaning robots not very intelligent. Summary of the invention

[0003] In view of this, the present application proposes a cleaning robot and its task self-learning method and control method, which effectively increases the autonomous decision-making ability of the cleaning robot by performing autonomous task learning, so that the cleaning robot can complete complex tasks autonomously and flexibly adapt to complex and changing environments.

[0004] In a first aspect, an embodiment of the present application proposes a task self-learning method for a cleaning robot, comprising:

[0005] After receiving the task instruction, the current sensor data of the cleaning robot is obtained in real time; wherein the task instruction includes a cleaning task, a transport task or a return to base station task;

[0006] Based on the self-sensor data, the task model in the cleaning robot controls the cleaning robot to perform corresponding actions;

[0007] The task model performs behavioral learning based on the performance feedback of the cleaning robot in executing the action, until the task model is guided to control the cleaning robot to complete the task corresponding to the task instruction and achieve the task goal corresponding to the task.

[0008] In some embodiments, the task model is pre-constructed with a correspondence between executable actions and environmental states and / or operating states determined based on its own sensor data; the task model in the cleaning robot controls the cleaning robot to perform corresponding actions, including:

[0009] The task model in the cleaning robot determines the current environmental state or operating state according to the current acquired self-sensor data, and determines the executable action corresponding to the environmental state or the operating state according to the corresponding relationship, so that the task model controls the actuator of the cleaning robot to perform the executable action;

[0010] Wherein, the actuator includes any one or more combinations of walking wheels, cleaning parts, fans and pump bodies.

[0011] In some embodiments, the task model performs behavior learning based on the performance feedback of the cleaning robot performing the action, including:

[0012] Based on the updated sensor data of the cleaning robot itself after executing the action and / or the external sensor data in the current environment of the cleaning robot, the performance of executing the action is evaluated using behavior evaluation rules or behavior evaluation models, and the performance feedback is sent to the task model to enable the task model to perform behavioral learning.

[0013] In some embodiments, the using of a behavior evaluation rule or a behavior evaluation model to evaluate the performance of executing the action and sending the performance feedback to the task model so that the task model performs behavior learning includes:

[0014] If the behavior evaluation rule or the behavior evaluation model evaluates that the performance of the action belongs to the preset good performance, positive feedback is sent to the task model;

[0015] The task model records the action as a reference for subsequent action execution of the task, and updates the corresponding relationship between the executable action and the environmental state and / or operating state determined based on its own sensor data;

[0016] Among them, the preset good performance includes any one or more combinations of no collision, no getting stuck, no missed scans, completing the action within a preset time, a reasonable walking path, a reasonable posture, and local cleanliness meeting the standards.

[0017] In some embodiments, the using of a behavior evaluation rule or a behavior evaluation model to evaluate the performance of executing the action and sending the performance feedback to the task model so that the task model performs behavior learning includes:

[0018] If the behavior evaluation rule or the behavior evaluation model evaluates that the performance of the action belongs to a preset abnormal performance, negative feedback and behavior correction information corresponding to the abnormal performance are sent to the task model;

[0019] The task model performs action adjustment according to the behavior correction information, and updates the corresponding relationship between the executable action and the environmental state and / or operating state determined based on the self-sensor data based on the adjusted action;

[0020] The preset abnormal performance includes any one or more combinations of falling off a cliff, collision, getting stuck, missing a sweep, timeout for action completion, unreasonable walking path, unreasonable posture, and local cleanliness not meeting the standard;

[0021] The corresponding behavior correction information includes: adjusting the moving speed; adjusting the distance; adjusting the turning angle; adjusting the walking path; adjusting the rotation speed of the cleaning element; adjusting the suction force; adjusting the water volume; adjusting the state of the cleaning element; adjusting the self-cleaning frequency of the cleaning element; or one or more combinations thereof.

[0022] In some embodiments, the behavior evaluation rule is pre-constructed with a correspondence between the updated self-sensor data and / or the external sensor data after performing the action and the performance feedback;

[0023] or;

[0024] The behavior evaluation model is a large language model, which is used to perform semantic analysis based on the updated self-sensor data and / or the external sensor data after executing the action to obtain the performance feedback output to the task model.

[0025] In some embodiments, the step of guiding the cleaning robot to complete the task corresponding to the task instruction and achieve the task goal corresponding to the task includes:

[0026] After completing this task, if the task target corresponding to the task is not achieved, the cleaning robot is controlled to move to the starting position for executing the task, and then returns to the step of obtaining the current sensor data of the cleaning robot in real time until the task target is achieved.

[0027] In some embodiments, if the task instruction is the cleaning task, and the cleaning task is global cleaning, after receiving the task instruction, before the task model controls the cleaning robot to perform the corresponding action, the method further includes:

[0028] Decomposing the global cleaning into tasks to obtain a plurality of decomposed tasks; wherein each of the decomposed tasks corresponds to a decomposed task model; wherein the decomposed tasks include global edge cleaning and single room cleaning;

[0029] Each decomposed task model performs behavior learning according to the execution order of the decomposed tasks until the learning of all the decomposed task models is completed;

[0030] Accordingly, based on the self-sensor data, the task model controls the cleaning robot to perform corresponding actions, including:

[0031] Based on the self-sensor data, the task model controls the cleaning robot to sequentially combine the decomposed tasks and execute actions corresponding to the decomposed tasks according to each of the decomposed task models.

[0032] In some embodiments, the behavior learning of each decomposed task model according to the execution order of the decomposed tasks includes:

[0033] According to the execution order of the decomposed tasks, each of the decomposed task models controls the cleaning robot to perform corresponding actions based on the self sensor data;

[0034] The decomposed task model performs behavioral learning based on the performance feedback of the cleaning robot executing the action, until the cleaning robot is guided to complete the corresponding decomposed task and achieve the decomposed task goal corresponding to the decomposed task.

[0035] In some embodiments, the task objectives include: overall cleanliness higher than a preset cleanliness, cleaning time lower than a preset time, cleaning coverage higher than a preset coverage, number of collisions lower than a first preset number, number of jams lower than a second preset number, and number of cliff falls lower than a third preset number, or one or more combinations thereof.

[0036] In some embodiments, the cleaning task includes one of edge cleaning, global cleaning, designated area cleaning, and fixed-point cleaning, and the transportation task includes replenishing water to external equipment or transporting household items.

[0037] In a second aspect, an embodiment of the present application provides a control method for a cleaning robot, comprising:

[0038] In response to a task instruction issued by a user, the task model in the cleaning robot controls the cleaning robot to perform the task corresponding to the task instruction based on the task instruction;

[0039] After the task is executed, user feedback information is received, so that the task model adjusts the action or the combination of actions when the task is executed next time according to the user feedback information;

[0040] Wherein, the task model is obtained based on the task self-learning method of the cleaning robot.

[0041] In some embodiments, the task model adjusts the action or combination of actions when performing the task next time according to the user feedback information, including:

[0042] Inputting the user feedback information into a large language model;

[0043] If the user feedback information includes negative feedback, the large language model outputs semantic text information to the task model; the semantic text information includes negative feedback and behavior correction information;

[0044] The task model adjusts the action or the combination of actions when the task is performed next time according to the negative feedback and the behavior correction information.

[0045] In a third aspect, an embodiment of the present application provides a cleaning robot, comprising:

[0046] An actuator is used to perform corresponding actions; the actuator includes any one or more combinations of a walking wheel, a cleaning member, a fan, and a pump body;

[0047] a memory storing a computer program;

[0048] A controller is used to execute the computer program to implement the task self-learning method or control method of the cleaning robot.

[0049] In a fourth aspect, an embodiment of the present application proposes a computer-readable storage medium storing a computer program, which, when executed on a processor, implements the task self-learning method or control method of the cleaning robot.

[0050] The embodiments of the present application have the following advantages:

[0051] The task self-learning method of the cleaning robot proposed in the embodiment of the present application includes: the cleaning robot responds after receiving the task instruction, first obtains the current sensor data of itself, and based on the sensor data of itself, the task model in the cleaning robot controls the cleaning robot to perform the corresponding action in an end-to-end manner; then, the task model performs behavior learning according to the performance feedback of the cleaning robot in performing the action, until the task model is guided to control the cleaning robot to complete the task corresponding to the task instruction and achieve the task goal corresponding to the task. The present application is guided by completing tasks and achieving task goals, so that the task model of the cleaning robot performs autonomous behavior learning, and uses the learned task model to autonomously complete complex tasks, so that the cleaning robot can not only perform tasks in a fixed manner according to pre-input algorithms and rules, but also effectively increases the autonomous decision-making ability of the cleaning robot, so that the cleaning robot can flexibly adapt to complex and changing environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] In order to more clearly illustrate the technical solution of the present invention, the following is a brief introduction to the drawings required for use in the embodiments. It should be understood that the following drawings only illustrate certain embodiments of the present invention and should not be regarded as limiting the scope of protection of the present invention. In each of the drawings, similar components are numbered similarly.

[0053] Figure 1 A schematic diagram of the structure of a cleaning robot according to an embodiment of the present application is shown;

[0054] Figure 2 A first flow chart of the task self-learning method of the cleaning robot according to an embodiment of the present application is shown;

[0055] Figure 3 A second flow chart of the task self-learning method of the cleaning robot according to an embodiment of the present application is shown;

[0056] Figure 4 A third flow chart showing the task self-learning method of the cleaning robot according to an embodiment of the present application is shown;

[0057] Figure 5 A schematic diagram of a global cleaning area including multiple rooms is shown;

[0058] Figure 6 A first flow chart showing a method for controlling a cleaning robot according to an embodiment of the present application is shown;

[0059] Figure 7 A schematic diagram showing a structure of a task self-learning device of a cleaning robot according to an embodiment of the present application is shown;

[0060] Figure 8 Another structural schematic diagram of the task self-learning device of the cleaning robot according to an embodiment of the present application is shown;

[0061] Fig. 9 A structural schematic diagram of a control device of a cleaning robot according to an embodiment of the present application is shown.

[0062] Description of main component symbols:

[0063] 10-cleaning robot; 11-processor; 12-memory; 13-actuator; 14-sensing unit;

[0064] 100-first task self-learning device; 110-task receiving module; 120-action execution module; 130-model learning module;

[0065] 200-second task self-learning device; 210-task receiving decomposition module; 220-decomposition task learning module; 230-global task execution module; 240-global task learning module;

[0066] 300 - control device; 310 - task execution module; 320 - feedback adjustment module. DETAILED DESCRIPTION

[0067] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0068] The components of the embodiments of the present invention generally described and shown in the drawings herein may be arranged and designed in a variety of different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0069] Hereinafter, the terms "including", "having" and their cognates, which may be used in various embodiments of the present invention, are intended only to indicate specific features, numbers, steps, operations, elements, components or combinations of the foregoing items, and should not be understood as first excluding the existence of one or more other features, numbers, steps, operations, elements, components or combinations of the foregoing items or adding the possibility of one or more features, numbers, steps, operations, elements, components or combinations of the foregoing items.

[0070] Furthermore, the terms “first”, “second”, “third”, etc. are merely used for distinguishing descriptions and are not to be understood as indicating or implying relative importance.

[0071] Unless otherwise defined, all terms (including technical terms and scientific terms) used herein have the same meanings as those generally understood by those skilled in the art to which the various embodiments of the present invention belong. The terms (such as those defined in generally used dictionaries) will be interpreted as having the same meanings as the contextual meanings in the relevant technical field and will not be interpreted as having idealized meanings or overly formal meanings unless clearly defined in the various embodiments of the present invention.

[0072] Example 1

[0073] Figure 1 Shown is a structural schematic diagram of the cleaning robot 10 proposed in an embodiment of the present application.

[0074] Exemplarily, the cleaning robot 10 includes a processor 11, a memory 12 and an actuator 13, wherein the actuator 13 is used to perform corresponding actions; the memory 12 stores a computer program, and the processor 11 runs the computer program, thereby enabling the cleaning robot 10 to perform the cleaning robot task self-learning method or the cleaning robot control method of the following embodiment.

[0075] Among them, the cleaning robot 10 may include a sweeping robot, a mopping robot and a sweeping and mopping robot.

[0076] Among them, the processor 11 can be an integrated circuit chip with signal processing capabilities. The processor 11 can be a general-purpose processor, including a central processing unit (CPU), a graphics processing unit (GPU), a network processor (NP), a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA) or at least one of other programmable logic devices, discrete gates or transistor logic devices, and discrete hardware components. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc., which can implement or execute the methods, steps and logic block diagrams disclosed in the embodiments of the present application.

[0077] The memory 12 may be, but is not limited to, a random access memory (RAM), a read only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable read-only memory (EEPROM), etc. The memory 12 is used to store a computer program, and the processor 11 may execute the computer program accordingly after receiving an execution instruction.

[0078] Among them, the actuator 13 is mainly used to enable the cleaning robot 10 to perform corresponding actions. For example, it may include but is not limited to any one or more combinations of walking wheels, cleaning parts, fans, and pump bodies. Among them, the walking wheels may include but are not limited to side wheels, track wheels, rollers, etc., which can be set according to the specific use scenario of the cleaning robot. Among them, the cleaning parts may include but are not limited to side brushes, roller brushes and rollers, etc. Different types of cleaning parts can perform different cleaning operations. The fan can be used for vacuuming, and the pump body can be used to pump water to the drum, etc. It can be understood that the specific structure of the actuator 13 can be set according to the actual needs of the cleaning robot 10, and no excessive restrictions are made here.

[0079] Furthermore, in order to realize the intelligence of the cleaning robot 10, the cleaning robot 10 also includes a sensing unit 14 arranged on the cleaning robot 10, and the sensing unit 14 is used to transmit necessary internal and external state information to the cleaning robot 10, so that the cleaning robot 10 performs corresponding actions and completes the corresponding tasks.

[0080] For example, if divided according to the sensing range, the sensing unit 14 includes an internal sensing unit and an external sensing unit. Among them, for the internal sensing unit, it is mainly used for the cleaning robot 10 to collect the status information of the equipment itself, such as motor current sensors, inertial measurement units (IMU, Inertial Measurement Unit), etc. For the external sensing unit, it is mainly used for the cleaning robot 10 to collect the environment information in which it is located, such as ranging sensors, ultrasonic sensors, visual devices, laser radars, collision sensors, etc. It can be understood that for different cleaning robots 10, the sensing unit 14 can be adaptively set according to the actual function, which is not limited here.

[0081] Example 2

[0082] In order to improve the intelligence level of the cleaning robot 10 in the above embodiment, the present application proposes to increase the autonomous decision-making ability of the cleaning robot 10 by enabling the task model in the cleaning robot 10 to perform autonomous behavior learning, so that the task model can control the cleaning robot 10 end-to-end to complete complex tasks and flexibly adapt to various complex environments.

[0083] It should be understood that the above-mentioned task model is a machine learning model that can run in the processor 11. By learning the mapping relationship from input to output, the task model can have end-to-end control capabilities. There is no limitation here on which models can be used to construct the machine learning model.

[0084] Among them, end-to-end means that the task model can directly control the actuator 13 according to the input of the perception unit 14, and no step-by-step processing is required in the middle. For the above-mentioned task model, on the one hand, since the task model can directly generate control output from the perception input, the processing time is reduced, thereby improving the response speed. On the other hand, since the intermediate step-by-step processing is reduced, the system design is also simplified.

[0085] The task self-learning method of the cleaning robot 10 proposed in the embodiment of the present application is described in detail below.

[0086] Figure 2 Shown is a flow chart of a task self-learning method of the cleaning robot 10 according to an embodiment of the present application.

[0087] Exemplarily, the task self-learning method of the cleaning robot 10 includes steps S110 - S130 .

[0088] S110, after receiving the task instruction, the cleaning robot's current self-sensor data is obtained in real time.

[0089] Among them, the task instruction can be sent to the cleaning robot 10 by an external object (such as a user) through a terminal device (such as a mobile phone, a tablet, a smart speaker, etc.) or through a button on the cleaning robot 10.

[0090] Among them, task instructions may include cleaning tasks, transportation tasks or return to base station tasks, etc.

[0091] In an optional embodiment, the cleaning task includes one of edge cleaning, global cleaning (cleaning the entire area), designated area cleaning, fixed-point cleaning, etc.; the transportation task includes replenishing water to external equipment or transporting household items.

[0092] Specifically, the external device can be a device that requires external water addition, such as a humidifier and a fish tank; the household objects can be smart home objects with specific functions, such as cameras, humidifiers, and air purifiers, or the household objects can be ordinary household objects to be moved, such as tableware, paper towels, umbrellas, and towels; the transportation of household objects in the transportation task is to transport the household objects to a designated target location, or to transport smart home objects in a home environment.

[0093] Specifically, the base station in the return to base station task is mainly used to dock the cleaning robot 10 so that the base station can perform operations such as charging, dust collection or water change on the cleaning robot.

[0094] Specifically, the self-sensor is a sensor of the perception unit 14 of the cleaning robot 10 .

[0095] Exemplarily, after receiving the task instruction, the cleaning robot 10 responds to the task instruction. First, the cleaning robot 10 obtains the current state of the device and / or the current environment information through the sensing unit 14.

[0096] For example, the cleaning robot 10 can collect visual images of the current environment through visual sensors (such as monocular or binocular cameras, etc.), collect whether there are obstacles in front through collision sensors (such as ultrasonic, infrared sensors, etc.), collect point cloud data of obstacles in the current environment through lidar, etc., collect current data of each motor in the equipment through current sensors, and collect equipment posture information through posture sensors (such as IMU, etc.).

[0097] S120, based on its own sensor data, the task model in the cleaning robot controls the cleaning robot to perform corresponding actions.

[0098] Exemplarily, after acquiring its own sensor data, the task model in the cleaning robot 10 will determine the current environmental state or operating state based on the acquired current self-sensor data; further, based on the correspondence between the pre-built executable actions and the environmental state and / or operating state determined by the self-sensor data, determine the executable actions under the current environmental state or operating state, so that the task model drives the actuator 13 of the cleaning robot 10 to execute the above executable actions. In other words, the task model determines the current executable action and attempts it based on the current sensor data and the above correspondence.

[0099] For example, the task model can determine whether there is a wall, table, chair, or other environmental state in front of it based on the sensor data of the visual sensor or the laser radar, and / or can determine whether it can walk or other operating states based on the current data of the motor, and then determine the current executable action based on the current environmental state, operating state, or a combination of the two. For example, these executable actions may include, but are not limited to, controlling the cleaning robot 10 to try to execute one of the actions of moving forward, moving backward, turning, slowing down, speeding up, extending and retracting the cleaning parts, rotating the cleaning parts, turning the fan on / off, turning the pump on / off, etc., and of course, it can also be a combination of the above multiple actions.

[0100] It can be understood that regarding the correspondence between executable actions and environmental states and / or operating states determined by their own sensor data, for example, in one embodiment, the task model can determine the corresponding device operating state or environmental state by pre-collecting sensor data, and determine which actions or action combinations can be executed based on the corresponding operating state and / or environmental state, thereby constructing an input-output relationship between each state and a single executable action, or between each state and the permutation and combination of executable actions.

[0101] S130, the task model performs behavior learning based on the performance feedback of the cleaning robot executing the action, until the task model is guided to control the cleaning robot to complete the task corresponding to the task instruction and achieve the task goal corresponding to the task.

[0102] It can be understood that in the present application, since the task model can only control the cleaning robot to perform corresponding actions when receiving the task instruction, it cannot control the cleaning robot to complete the task corresponding to the task instruction. Therefore, it is necessary to be guided by completing the task and achieving the task goal. By making the task model perform behavioral learning, the learned task model can control the cleaning robot 10 to complete the task autonomously and achieve the task goal. There is no need to control the cleaning robot 10 to complete the task through artificially pre-written algorithms and rules, thereby improving the autonomous decision-making ability of the cleaning robot 10.

[0103] Exemplarily, after the cleaning robot 10 completes an action or is performing an action, performance feedback of the cleaning robot 10 performing the action is obtained, and then the performance feedback is sent to the task model so that the task model can perform behavioral learning. That is, the task model adjusts its behavior after obtaining the performance feedback, so as to autonomously learn how to better complete the assigned task.

[0104] It should be understood that the action in the execution action described in step S130 is an executable action determined by step S120, and therefore, it may be a single action or multiple arranged coherent actions (such as back-rotate-forward, etc.). Based on this, there is no limitation on the timing of the performance feedback of the execution of the action and its evaluation and feedback to the task model. For example, the behavior evaluation may be performed while the cleaning robot 10 is executing the action, and the feedback may be given to the task model in real time, so that the task model can adjust the behavior immediately; or the behavior evaluation may be performed after the cleaning robot 10 has completed the action, and then the feedback may be given to the task model, so that the task model can adjust the next behavior, etc.

[0105] In this application, the task objectives corresponding to each task may be different, which can be determined by the user according to the execution configuration requirements of the current task. For example, the execution configuration requirements include but are not limited to: the overall cleanliness is higher than the preset cleanliness, the cleaning time is lower than the preset time, the cleaning coverage is higher than the preset coverage (such as the cleaning area is not lower than the preset area threshold, etc.), the number of collisions is lower than the first preset number, the number of jams is lower than the second preset number, the number of cliff falls is lower than the third preset number, etc. One or more combinations thereof.

[0106] Among them, the overall cleanliness can be determined by external sensors judging the degree of dirtiness of the entire cleaning area; the cleaning coverage rate is the ratio of the area cleaned by the cleaning robot to the entire cleaning area; the cleaning robot being stuck can be a state where it is stuck by an obstacle and cannot walk; the cleaning robot falling off a cliff is a state where the cleaning robot falls from a height (such as stairs, steps, etc.).

[0107] The implementation process of the task model of step S130 is to perform behavior learning according to the performance feedback of the cleaning robot performing the action. For example, in one embodiment, Figure 3 As shown, it includes the following sub-steps:

[0108] S131, based on the cleaning robot's updated self-sensor data after executing the action and / or the external sensor data in the cleaning robot's current environment, the action performance is evaluated using behavior evaluation rules or behavior evaluation models, and the performance is fed back and sent to the task model so that the task model can perform behavior learning.

[0109] Exemplarily, when using behavior evaluation rules or behavior evaluation models for evaluation, the rationality of the cleaning robot 10 executing the action can be evaluated in combination with the updated self-sensor data obtained after the cleaning robot 10 performs the action and / or the external sensor data in the current environment of the cleaning robot, and then the corresponding evaluation feedback information can be sent to the task model.

[0110] Among them, the external sensor is relative to the self-sensor on the cleaning robot 10, and refers to other sensors arranged in the space where the cleaning robot 10 is located, such as being set in a home area or a test environment, and there is usually a certain distance between them and the cleaning robot 10. For example, the external sensor can be set on an obstacle (such as a wall, table or chair) in the home area or test environment where the cleaning robot 10 is located, or it can be set on an intelligent device in the home area or test environment where the cleaning robot 10 is located.

[0111] External sensors include, but are not limited to, multiple fixed or steerable following cameras, depth cameras, inductive sensors (such as infrared sensors, photoelectric sensors, ultrasonic sensors), etc., which are arranged in the space around the cleaning robot 10 (such as above, left side, right side, etc.).

[0112] It should be noted that both the external sensor and the cleaning robot 10 have a wireless communication module, which may be a Bluetooth module or a WiFi module. The cleaning robot 10 can obtain external sensor data through wireless communication; the cleaning robot 10 can obtain external sensor data in real time, and can also obtain external sensor data while the cleaning robot 10 is performing an action or after performing an action.

[0113] It is understandable that as the cleaning robot 10 performs an action, the data collected by its own sensors may also change accordingly. Therefore, it can be judged in combination with the updated own sensor data and the existing external sensors. For example, these sensor data can detect whether the cleaning robot 10 falls off a cliff, whether a collision occurs, whether the posture is reasonable, whether it is stuck by an obstacle, whether it misses a sweep, whether the action is completed in timeout, whether the walking path is reasonable, whether the local cleanliness meets the standard, and other corresponding performances.

[0114] In one embodiment, the above-mentioned behavior evaluation rules can be obtained through a pre-written algorithm, which may include: which self-sensor data and / or external sensor data are customized good performance (i.e., preset good performance), and the corresponding relationship between good performance and positive feedback; which self-sensor data and / or external sensor data are customized abnormal performance (i.e., preset abnormal performance), and the corresponding relationship between abnormal performance and negative feedback. Furthermore, for abnormal performance, it can also include behavior correction information corresponding to the abnormal performance. In other words, the behavior evaluation rules also include the corresponding relationship between abnormal performance, negative feedback, and behavior correction information corresponding to the abnormal performance.

[0115] In another embodiment, the above-mentioned behavior evaluation model may be a pre-trained large language model (LLM), which is used to perform semantic analysis based on the updated self-sensor data and / or external sensor data information after the action is executed, and obtain performance feedback output to the task model. It can be understood that the performance feedback may be positive feedback sent to the task model, or negative feedback sent to the task model, as well as behavior correction information corresponding to the negative feedback, etc., and the performance feedback may be output text information.

[0116] For example, the text message of negative feedback sent by a unidirectional task model may be: "Turn now, it's too far away from the table leg, and there are still many places that have not been swept!" Correspondingly, the behavior correction information is that when the cleaning robot 10's own sensor data detects the table leg, it needs to continue moving forward for a distance before rotating, etc., or, the text message of negative feedback may be: "The right side of the cleaning robot 10 hit the table leg. Next time you see a table leg, remember to brake and change direction in advance to avoid collision!" Correspondingly, the behavior correction information is that when the cleaning robot 10's own sensor data detects the table leg, it needs to rotate and change the direction of travel in advance; for another example, the text message of positive feedback sent by a unidirectional task model may be: "Great! The cleaning robot 10 did not collide with any obstacles this time, nor was it too far away from the obstacles to cause missed scanning."

[0117] It is understood that when evaluating behavior, the updated self-sensor data and / or the external sensor data after executing the corresponding action are used. The above-mentioned behavior evaluation rules or behavior evaluation models can be run outside the cleaning robot 10 (such as a cloud server, etc.), or can be run on the processor 11 inside the cleaning robot 10, depending on the main control computing power of the cleaning robot 10. For example, when the internal main control computing power of the cleaning robot 10 is insufficient, it can be run outside the cleaning robot 10 to speed up the model training process.

[0118] It should be noted that the behavior evaluation rules or behavior evaluation models can be adjusted according to the task objectives, so that the task model can perform behavior learning for different task objectives through the behavior evaluation rules or behavior evaluation models, so that the control module can control the cleaning robot 10 to achieve different task objectives.

[0119] It can be understood that the behavior correction information is used to guide the cleaning robot 10 to adjust its actions when abnormal performance occurs, so as to guide the task model to control the cleaning robot 10 to complete the task and achieve the task goal.

[0120] Therefore, for step S130, if the performance of the action is a good performance included in the above-mentioned behavior evaluation rules or behavior evaluation models, steps S132 and S133 are executed. On the contrary, if the performance of the action is an abnormal performance included in the behavior evaluation rules or behavior evaluation models, steps S134 and S135 are executed.

[0121] S132: If the performance of the action is a preset good performance, positive feedback is sent to the task model.

[0122] Among them, good behavior can be customized according to its own sensor data and / or external sensor data, for example, including but not limited to, detecting according to its own sensor data and / or external sensor data that the cleaning robot 10 has not collided, is not stuck, has not missed a sweep, completes the action within a preset time, has a reasonable walking path, a reasonable posture, and local cleanliness meets any one or more combinations thereof.

[0123] Among them, no missed sweeping means that the local area passed by the cleaning robot when performing the action is not missed; the local in the local cleanliness standard refers to the local cleaning area passed by the cleaning robot 10 when performing the action.

[0124] S133, the task model records the action as a reference for subsequent action execution of the task, and updates the correspondence between the executable action and the environmental state and / or operating state determined based on its own sensor data.

[0125] It can be understood that if the performance of the action performed by the cleaning robot 10 is a preset good performance, the task model will be recorded, so that when the same situation is encountered when the same task is performed next time, that is, the same self-sensor data is detected, it will be executed according to this action, that is, the behavior of this time is learned. Optionally, the correspondence between the pre-stored self-sensor data and the executable action can be updated according to the currently executed action, that is, the task model is updated.

[0126] It should be noted that before updating the task model, one self-sensor data may correspond to multiple executable actions or action combinations. After updating the task model according to positive feedback, the self-sensor data detected this time corresponds to this action or action combination.

[0127] S134, if the performance of the action belongs to the preset abnormal performance, negative feedback and behavior correction information corresponding to the preset abnormal performance are sent to the task model.

[0128] Among them, abnormal performance can also be customized according to its own sensor data and / or external sensor data, for example, including but not limited to, detecting according to its own sensor data and / or external sensor data that the cleaning robot 10 has collided, got stuck, missed a sweep, timed out from the action, had an unreasonable walking path, had an unreasonable posture (for example, when the cleaning robot 10 is docked with the base station, it is not facing the docking direction, but docking with an unreasonable posture, etc.), local cleanliness does not meet the standard, falls off a cliff, etc., any one or more combinations thereof.

[0129] Exemplarily, the behavior correction information may include, but is not limited to, one or more combinations of: adjusting the moving speed; adjusting the distance; adjusting the steering angle; adjusting the posture; adjusting the walking path; adjusting the rotation speed of the cleaning member; adjusting the suction force; adjusting the water volume; adjusting the state of the cleaning member; adjusting the self-cleaning frequency of the cleaning member, etc. Among them, the above-mentioned adjustment distance is to adjust the distance between the cleaning robot 10 and the obstacle or cliff. Among them, the above-mentioned adjustment of the steering angle may include not only the angle size when turning, but also the steering direction, such as adjusting from left turn to right turn, etc. Among them, the above-mentioned adjustment of the state of the cleaning member may include controlling whether the cleaning member is extended or retracted, such as adjusting the extension of the side brush to improve the cleaning efficiency of the corners; the above-mentioned adjustment of the self-cleaning frequency of the cleaning member may include adjusting the self-cleaning frequency of the drum, etc.

[0130] For example, if there is an abnormal performance of the cleaning robot 10 colliding with an obstacle (such as a table, chair, sofa, wall, etc.) or falling off a cliff (such as falling off a step, etc.), the behavior correction information corresponding to the abnormal performance may be: adjusting the moving speed, adjusting the distance, adjusting the steering angle, etc. One or more combinations thereof. For example, the behavior correction information is specifically to reduce the moving speed to avoid being unable to avoid due to excessive speed; and / or, increase the distance between the obstacle so that the cleaning robot 10 turns at a farther position from the obstacle or cliff, so as to avoid it in time; and / or, increase the steering angle to solve the collision caused by insufficient steering angle, etc. It can be understood that the amplitude of each adjustment in this embodiment can be preset, for example, the amplitude of reducing the moving speed is reduced by 0.1m / s, the amplitude of increasing the preset distance between the obstacle is increased by 0.01m, and the amplitude of increasing the steering angle is increased by 10 degrees, etc. Similar adjustment amplitudes in other embodiments can also be preset.

[0131] If the cleaning robot 10 is stuck in an abnormal state, the behavior correction information corresponding to the abnormal state may be: controlling the cleaning part to retract, or adjusting the walking path, etc., which may be determined according to the current environmental state.

[0132] If an abnormal performance occurs in which the cleaning robot 10 times out to complete an action, the behavior correction information corresponding to the abnormal performance may be: speeding up the speed at which the cleaning robot 10 performs the action.

[0133] If the cleaning robot 10 shows abnormal performance of missing a sweep, the behavior correction information corresponding to the abnormal performance may be one or more combinations of: adjusting the walking path, adjusting the distance, adjusting the steering angle, etc. It can be understood that for the same task instruction, when different walking paths are used, the time required, the coverage area, etc. may be different. Therefore, the missing sweep behavior can be avoided by adjusting the walking path of the cleaning robot 10. Of course, the cleaning robot 10 can also be controlled to reduce the distance from the obstacle or cliff or reduce the steering angle, etc.

[0134] If the cleaning robot 10 docks with the base station in an unreasonable posture, the behavior correction information corresponding to the abnormal behavior may be: adjust the body posture (for example, adjust the angle between the body center line of the cleaning robot and the body center line of the base station), etc., so that it docks with the base station as directly as possible.

[0135] If there is an abnormal performance that the cleaning robot 10 does not meet the local cleanliness standard, the behavior correction information corresponding to the abnormal performance may be: adjusting the rotation speed of the cleaning element, adjusting the suction force, adjusting the water volume, etc. For example, the behavior correction information specifically increases the rotation speed or suction force of the cleaning element to increase the cleaning force; of course, the behavior correction information may also specifically include increasing the water spray volume to make the wet roller easier to clean.

[0136] S135, the task model adjusts the action according to the behavior correction information, and updates the corresponding relationship between the executable action and the environmental state and / or the operating state determined based on the sensor data based on the adjusted action.

[0137] It should be noted that after updating the correspondence between the pre-stored executable actions and the environmental states and / or operating states determined based on its own sensor data, that is, after updating the task model, when the same task is subsequently executed and the same situation is encountered, that is, when the cleaning robot is in the same environmental state and / or operating state, it will be executed according to the adjusted actions.

[0138] It can be understood that each time the cleaning robot 10 performs an action, the task model is guided by completing the task and achieving the task goal, and behavior learning is performed based on the performance feedback obtained each time the action is executed.

[0139] In step S130, until the cleaning robot is guided to complete the task corresponding to the task instruction and reaches the task target corresponding to the task, steps S136 to S138 are included.

[0140] S136, after completing the task, determine whether the task goal corresponding to the task is achieved. If the task goal is achieved, execute step S138 to stop learning the task model. As an optional solution, if the task goal is not achieved, execute step S137.

[0141] S137, if the task target corresponding to the task is not achieved, the cleaning robot is controlled to move to the starting position of the task, and the process returns to the step of obtaining the current sensor data of the cleaning robot in real time.

[0142] This allows the cleaning robot to continue the next round of task execution and task model learning until the task goal is achieved.

[0143] It should be noted that when the next round of task execution and task model learning is carried out, the task model used is the task model after the previous round of learning, that is, an iterative learning method is adopted.

[0144] S138: If the task goal corresponding to the task is reached, stop the task model learning.

[0145] In other words, if the corresponding task goal is not achieved after executing a task once, it needs to be executed again. Each execution will use the correspondence between its own sensor data and executable actions recorded and updated in the previous execution to iteratively learn, that is, the updated task model, until the task goal is finally completed.

[0146] It can be understood that the above-mentioned task model learning process can be carried out in various different environments so that the task model can adapt to different environments.

[0147] The task self-learning method proposed in the embodiment of the present application is guided by the task goal, and realizes autonomous learning of the task model by outputting positive feedback, negative feedback and behavior correction information corresponding to the negative feedback to the task model. The task model controls the cleaning robot 10 to complete complex tasks autonomously based on the task goal, and can adapt to the complex and changeable home environment, making the cleaning robot 10 more intelligent.

[0148] In another embodiment, if a complex task is received, the complex task can be further decomposed, and then each decomposed task is executed in sequence and the decomposed task is learned, and finally the autonomous learning of the complex task is completed.

[0149] The following takes the task instruction as a cleaning task and the cleaning task as global cleaning as an example to illustrate the task self-learning method of the cleaning robot 10. After receiving the task instruction in step S110 and before the task model controls the cleaning robot to perform the corresponding action in step S120, Figure 4 As shown, the task self-learning method of the cleaning robot 10 includes the following steps:

[0150] S210, decomposing the global cleaning task to obtain a plurality of decomposed tasks; wherein each decomposed task corresponds to a decomposed task model.

[0151] For example, global cleaning can be decomposed into two subtasks: global edge cleaning and single room cleaning. Figure 5 As shown, for the cleaning of the entire area, the task can be decomposed into four decomposition tasks, namely global edge cleaning, cleaning of room A, cleaning of room B and cleaning of room C. The decomposition task models corresponding to the four decomposition tasks are the global edge cleaning model, the room A cleaning task model, the room B cleaning task model and the room C cleaning task model.

[0152] S220, each decomposed task model performs behavior learning in accordance with the decomposed task execution order until all decomposed task models have completed learning.

[0153] In one embodiment, exemplary, according to the execution order of all decomposed tasks, when each decomposed task model performs behavior learning, the current sensor data of the cleaning robot can be obtained in real time, so that each decomposed task model controls the cleaning robot to perform corresponding actions based on the current sensor data of the cleaning robot. Then, each decomposed task model will perform behavior learning based on the performance feedback of the cleaning robot performing the action, until the cleaning robot is guided to complete the corresponding decomposed task and achieve the decomposed task goal corresponding to the decomposed task.

[0154] For example, still Figure 5 Taking the global cleaning shown as an example, the global edge cleaning model, room A cleaning task model, room B cleaning task model and room C cleaning task model can control the cleaning robot to complete global edge cleaning, cleaning of room A, cleaning of room B and cleaning of room C respectively through learning, and achieve the task goals of global edge cleaning, cleaning of room A, cleaning of room B and cleaning of room C respectively.

[0155] Accordingly, in step S120, based on its own sensor data, the task model controls the cleaning robot to perform corresponding actions, including:

[0156] S230, based on its own sensor data, the task model controls the cleaning robot to sequentially combine the decomposed tasks according to each decomposed task model, and execute the actions corresponding to the decomposed tasks.

[0157] After acquiring each decomposed task model, the cleaning robot 10 starts to perform the task execution and self-learning of global cleaning. Exemplarily, the task model will control the cleaning robot to sequentially combine the decomposed tasks involved according to each decomposed task model, and then perform the actions corresponding to each decomposed task, so that the subsequent task model can perform behavioral learning according to the performance feedback of the cleaning robot performing the actions corresponding to each decomposed task, so that the task model can control the cleaning robot to complete the global cleaning task in the optimal sequential combination manner.

[0158] For example, still Figure 5 Taking the global cleaning shown as an example, if the task goal corresponding to the global cleaning is that the cleaning time is lower than the preset time, the task model first uses the room A cleaning task model to clean room A, and then uses the room B cleaning task model through the performance feedback guidance task model to clean room B, and then uses the room C cleaning task model through the performance feedback guidance task model to clean room C, and finally uses the global edge cleaning model to perform global edge cleaning starting from room C, thereby completing the global cleaning at the fastest speed.

[0159] In this embodiment, complex tasks are decomposed and then each decomposed task is subjected to the above-mentioned autonomous behavioral learning guided by the task objectives of the decomposed task to obtain various decomposed task models; finally, the task model uses these decomposed task models to guide the cleaning robot to complete the execution of complex tasks, so that the cleaning robot 10 can cope with more complex tasks through the task model.

[0160] Example 3

[0161] Figure 6 The figure is a flow chart of a control method of the cleaning robot 10 according to an embodiment of the present application. Exemplarily, the control method of the cleaning robot 10 includes:

[0162] S310, in response to a task instruction issued by a user, a task model in the cleaning robot controls the cleaning robot to perform the task corresponding to the task instruction based on the task instruction.

[0163] The task model in this embodiment is obtained based on the task self-learning method of the cleaning robot 10 in the above-mentioned embodiment 2, so it will not be described again here.

[0164] S320, after completing the task, receiving user feedback information, so that the task model adjusts the action or combination of actions when executing the task next time according to the user feedback information.

[0165] In one embodiment, the user feedback information can be input into the large language model for semantic analysis and corresponding feedback information can be obtained. If the user feedback information includes negative feedback, for example, "the cleaning time needs to be further shortened this time", the large language model outputs semantic text information to the task model; wherein the semantic text information includes negative feedback and behavior correction information.

[0166] Furthermore, the task model adjusts the action or combination of actions when performing the task next time based on the negative feedback and behavior correction information.

[0167] Alternatively, in another case, if the user feedback information includes positive feedback, for example, "the number of collisions during this cleaning is small", the task model will also record all the behaviors of this execution as a reference for the next execution of the same task.

[0168] It is understandable that in actual use, after the cleaning robot 10 completes the cleaning task, the user can rate the cleaning robot 10 for executing this task through the application APP on the terminal device, and input relevant suggestions, questions and affirmations, etc. Therefore, the behavior evaluation model can use the user's feedback information, so that the task model will adjust the action of the next execution of the same task according to the user's suggestions, that is, update according to the user's feedback, so that the cleaning robot 10 has the ability to continue learning.

[0169] Example 4

[0170] Figure 7 Shown is a structural schematic diagram of a first task self-learning device 100 of the cleaning robot 10 according to an embodiment of the present application.

[0171] Exemplarily, the first task self-learning device 100 of the cleaning robot 10 includes:

[0172] The task receiving module 110 is used to obtain the current sensor data of the cleaning robot in real time after receiving the task instruction; wherein the task instruction includes a cleaning task, a transport task or a return to base station task.

[0173] The action execution module 120 is used to control the cleaning robot to perform corresponding actions based on the self-sensor data and the task model in the cleaning robot.

[0174] The model learning module 130 is used for the task model to perform behavioral learning based on the performance feedback of the cleaning robot executing the action, until the task model is guided to control the cleaning robot to complete the task corresponding to the task instruction and achieve the task goal corresponding to the task.

[0175] It can be understood that the device of this embodiment corresponds to the task self-learning method of the cleaning robot 10 in the above-mentioned embodiment 2, and the optional items in the above-mentioned embodiment are also applicable to this embodiment, so they will not be repeated here.

[0176] Example 5

[0177] Figure 8 Shown is a structural schematic diagram of the second task self-learning device 200 of the cleaning robot 10 according to an embodiment of the present application.

[0178] Exemplarily, the second task self-learning device 200 of the cleaning robot 10 includes:

[0179] The task receiving and decomposing module 210 is used to decompose the global cleaning task after receiving the task instruction of the global cleaning, and obtain a plurality of decomposed tasks; wherein each decomposed task corresponds to a decomposed task model.

[0180] The decomposed task learning module 220 is used to perform behavior learning on each decomposed task model according to the decomposed task execution order until the learning of all decomposed task models is completed.

[0181] The global task execution module 230 is used to control the cleaning robot to sequentially combine the decomposed tasks and execute actions corresponding to the decomposed tasks based on the self-sensor data and the task model according to each of the decomposed task models.

[0182] It can be understood that the device of this embodiment corresponds to the task self-learning method of the cleaning robot 10 in the above-mentioned embodiment 2, wherein the optional items in the above-mentioned embodiment 2 are also applicable to this embodiment, so they will not be described repeatedly here.

[0183] Example 6

[0184] Fig. 9 Shown is a structural schematic diagram of the control device 300 of the cleaning robot 10 according to an embodiment of the present application.

[0185] Exemplarily, the control device 300 of the cleaning robot 10 includes:

[0186] The task execution module 310 is used to respond to the task instruction issued by the user. The task model controls the cleaning robot 10 to execute the task corresponding to the task instruction based on the task instruction.

[0187] The task model is obtained based on the task self-learning method of the cleaning robot 10 described in the second embodiment.

[0188] The feedback adjustment module 320 is used to receive user feedback information after the task is executed, so that the task model adjusts the action or the combination of actions when the task is executed next time according to the user feedback information.

[0189] It can be understood that the device of this embodiment corresponds to the application method of the task model of the cleaning robot 10 in the above-mentioned embodiment 3, and the optional items in the above-mentioned embodiment are also applicable to this embodiment, so they will not be repeated here.

[0190] Example 7

[0191] The embodiment of the present application also proposes a computer-readable storage medium, which stores a computer program. When the computer program is executed on the processor 11, it implements the task self-learning method of the cleaning robot 10 in the above-mentioned embodiment 2 or the control method of the cleaning robot 10 in the above-mentioned embodiment 3.

[0192] For example, the computer-readable storage medium includes: a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and other media that can store program codes.

[0193] In several embodiments provided in the present application, it should be understood that the disclosed devices and methods can also be implemented in other ways. The device embodiments described above are merely schematic. For example, the flowcharts and structure diagrams in the accompanying drawings show the possible architecture, functions and operations of the devices, methods and computer program products according to multiple embodiments of the present invention. In this regard, each box in the flowchart or block diagram can represent a module, a program segment or a part of a code, and the module, a program segment or a part of a code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in an alternative implementation, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the structure diagram and / or the flow chart, and the combination of boxes in the structure diagram and / or the flow chart, can be implemented with a dedicated hardware-based system that performs a specified function or action, or can be implemented with a combination of dedicated hardware and computer instructions.

[0194] In addition, the functional modules or units in the various embodiments of the present invention may be integrated together to form an independent part, or each module may exist independently, or two or more modules may be integrated to form an independent part.

[0195] If the functions are implemented in the form of software function modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art or the part of the technical solution can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a smart phone, a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention.

[0196] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A task self-learning method for a cleaning robot, characterized in that: After receiving the task instruction, the current sensor data of the cleaning robot is obtained in real time; wherein the task instruction includes a cleaning task, a transport task or a return to base station task; Based on the self-sensor data, the task model in the cleaning robot determines the current environmental state or operating state according to the acquired current self-sensor data, and determines the executable action corresponding to the environmental state or the operating state according to the correspondence between the pre-constructed executable action and the environmental state and / or operating state determined based on the self-sensor data, so that the task model controls the actuator of the cleaning robot to execute the executable action; Using a behavior evaluation model to evaluate the performance of executing the action according to the updated self-sensor data of the cleaning robot after executing the action and the external sensor data in the current environment of the cleaning robot, and sending the performance feedback to the task model so that the task model can perform behavior learning; The behavior evaluation model is a large language model, which is used to obtain the performance feedback output to the task model based on the updated self-sensor data and external sensor data after the action is performed; the external sensor is set in the home area or test environment where the cleaning robot is located, including a plurality of fixed or steerable follow-up cameras and / or inductive sensors set in the space around the cleaning robot; the cleaning robot obtains the external sensor data through wireless communication; Until the task model is guided to control the cleaning robot to complete the task corresponding to the task instruction and achieve the task goal corresponding to the task.

2. The method according to claim 1, characterized in that The actuator includes any one or more combinations of a walking wheel, a cleaning member, a fan, and a pump body.

3. The method according to claim 1, characterized in that: The step of evaluating the performance of executing the action using the behavior evaluation model and sending the performance feedback to the task model so that the task model performs behavior learning includes: If the behavior evaluation model evaluates that the performance of the action is a preset good performance, positive feedback is sent to the task model; The task model records the action as a reference for subsequent action execution of the task, and updates the corresponding relationship between the executable action and the environmental state and / or operating state determined based on its own sensor data; Among them, the preset good performance includes any one or more combinations of no collision, no getting stuck, no missed scanning, completing the action within the preset time, reasonable walking path, reasonable posture, and local cleanliness meeting the standards.

4. The method according to claim 1, characterized in that The step of evaluating the performance of executing the action using the behavior evaluation model and sending the performance feedback to the task model so that the task model performs behavior learning includes: If the behavior evaluation model evaluates that the performance of the action belongs to a preset abnormal performance, negative feedback and behavior correction information corresponding to the preset abnormal performance are sent to the task model; The task model performs action adjustment according to the behavior correction information, and updates the corresponding relationship between the executable action and the environmental state and / or operating state determined based on the self-sensor data based on the adjusted action; The preset abnormal performance includes any one or more combinations of falling off a cliff, collision, getting stuck, missing a sweep, timeout for action completion, unreasonable walking path, unreasonable posture, and local cleanliness not meeting the standard; The corresponding behavior correction information includes: adjusting the moving speed; adjusting the distance; adjusting the turning angle; adjusting the walking path; adjusting the rotation speed of the cleaning element; adjusting the suction force; adjusting the water volume; adjusting the state of the cleaning element; adjusting the self-cleaning frequency of the cleaning element, or one or more combinations thereof.

5. The method according to claim 1, characterized in that The large language model is used to perform semantic analysis based on the self-sensor data and the external sensor data updated after performing the action.

6. The method according to claim 1, characterized in that The step of guiding the cleaning robot to complete the task corresponding to the task instruction and achieve the task target corresponding to the task includes: After completing this task, if the task target corresponding to the task is not achieved, the cleaning robot is controlled to move to the starting position for executing the task, and then returns to the step of obtaining the current sensor data of the cleaning robot in real time until the task target is achieved.

7. The method according to any one of claims 1 to 6, characterized in that If the task instruction is the cleaning task, and the cleaning task is global cleaning, after receiving the task instruction, before the task model controls the cleaning robot to perform a corresponding action, the method further includes: Decomposing the global cleaning task to obtain a plurality of decomposed tasks; wherein each decomposed task corresponds to a decomposed task model; wherein the decomposed tasks include global edge cleaning and single room cleaning; Each decomposed task model performs behavior learning according to the execution order of the decomposed tasks until the learning of all the decomposed task models is completed; Accordingly, the task model controls the actuator of the cleaning robot to execute the executable action, including: Based on the self-sensor data, the task model controls the cleaning robot to sequentially combine the decomposed tasks according to each of the decomposed task models, and to execute the action corresponding to each of the decomposed tasks.

8. The method according to claim 7, characterized in that According to the execution order of the decomposed tasks, each decomposed task model performs behavior learning, including: According to the execution order of the decomposed tasks, each of the decomposed task models controls the cleaning robot to perform corresponding actions based on the self sensor data; The decomposed task model performs behavioral learning based on the performance feedback of the cleaning robot executing the action, until the cleaning robot is guided to complete the corresponding decomposed task and achieve the decomposed task goal corresponding to the decomposed task.

9. The method according to any one of claims 1 to 6, characterized in that: The task objectives include: overall cleanliness higher than a preset cleanliness, cleaning time lower than a preset time, cleaning coverage higher than a preset coverage, number of collisions lower than a first preset number, number of jams lower than a second preset number, and number of cliff falls lower than a third preset number, or one or more combinations thereof.

10. The method according to any one of claims 1 to 6, characterized in that The cleaning task includes one of edge cleaning, global cleaning, designated area cleaning, and fixed-point cleaning; the transportation task includes replenishing water to external equipment or transporting household items.

11. A control method for a cleaning robot, characterized in that: include: In response to a task instruction issued by a user, the task model in the cleaning robot controls the cleaning robot to perform the task corresponding to the task instruction based on the task instruction; After the task is executed, user feedback information is received, so that the task model adjusts the action or the combination of actions when the task is executed next time according to the user feedback information; Wherein, the task model is obtained based on the task self-learning method of the cleaning robot according to any one of claims 1 to 10.

12. The method according to claim 11, characterized in that The task model adjusts the action or combination of actions when the task is executed next time according to the user feedback information, including: Inputting the user feedback information into a large language model; If the user feedback information includes negative feedback, the large language model outputs semantic text information to the task model; the semantic text information includes negative feedback and behavior correction information; The task model adjusts the action or the combination of actions when the task is performed next time according to the negative feedback and the behavior correction information.

13. A cleaning robot, characterized in that: include: An actuator is used to perform corresponding actions; the actuator includes any one or more combinations of walking wheels, cleaning parts, fans, and pump bodies; a memory storing a computer program; A controller, configured to execute the computer program to implement the method according to any one of claims 1 to 12.

14. A computer-readable storage medium, characterized in that: The device stores a computer program, which, when executed on a processor, implements the method according to any one of claims 1 to 12.

Images