Remote management method of livestock house manure cleaning robot, robot and electronic equipment

By analyzing regional data and power data, using deep learning and simulated annealing algorithms, intelligent scheduling of multiple manure cleaning robots is achieved, solving the problem of manure cleaning robot management in scattered livestock houses and improving the efficiency and safety of manure cleaning.

CN118761600BActive Publication Date: 2024-11-26北京佳沃天河智能科技有限公司
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411224507.5
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-09-03
Publication Date
2024-11-26
Estimated Expiration
2044-09-03

AI Technical Summary

Technical Problem

The prior art is difficult to effectively manage multiple livestock house manure cleaning robots scattered in different places, cannot assign cleaning tasks in a targeted manner, and fail to fully consider the robot's own energy conditions, resulting in insufficient cleaning efficiency and safety.

Method used

By receiving the regional data collected by the perception end, analyzing the dirt data and power data, transmitting it to the cloud platform, using deep learning algorithms to estimate the time and power consumption of cleaning tasks, and using simulated annealing algorithm to formulate intelligent scheduling strategies to achieve multiple cleaning Remote management and collaborative scheduling of feces robots.

Benefits of technology

It improves the efficiency and quality of manure cleaning, reduces the risk of insufficient power of the robot, improves the level of animal husbandry management, and ensures the efficient and orderly progress of cleaning tasks.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118761600B_ABST
    Figure CN118761600B_ABST
Patent Text Reader

Abstract

The present invention belongs to the technical field of robot management, and discloses a remote management method for a manure cleaning robot in a livestock house, a robot and an electronic device; the method comprises: receiving regional data collected by a sensing end; analyzing the regional data to obtain dirt data; collecting power data; transmitting the dirt data and power data to a cloud platform; receiving and executing an intelligent scheduling strategy formulated by the cloud platform; the present invention fully considers the energy status of the robot to realize unified remote management and coordinated scheduling of multiple manure cleaning robots; while greatly improving the manure cleaning efficiency and quality, it also improves the safety and hygiene level of manure cleaning operations, and further improves the level of animal husbandry management.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of robot management technology, and more specifically, to a remote management method for a manure-cleaning robot in a livestock house, a robot and electronic equipment. Background Art

[0002] Modern animal husbandry has put forward higher requirements for the management of livestock house environment and the treatment of feces. The traditional manual manure cleaning method is inefficient and has potential safety hazards, making it difficult to meet the growing demand for manure cleaning. The use of automated and intelligent livestock house manure cleaning robots can effectively improve manure cleaning efficiency, reduce work intensity, and reduce the risk of manual operation. However, as the scale of livestock enterprises continues to expand, there are often more than one livestock house group in livestock enterprises, such as cattle houses, sheep houses, pig houses, etc., and multiple livestock houses are scattered in different locations. At the same time, due to its large area, each livestock house needs to be equipped with multiple manure cleaning robots to coordinate manure cleaning to meet the cleaning needs of the livestock house. Therefore, how to remotely and centrally control and coordinate the manure cleaning robots scattered in various places has become an urgent problem to be solved.

[0003] The Chinese patent application with publication number CN116540612A discloses a robot real-time dispatch management system; including a server, a remote data acquisition terminal, a positioning module, a vehicle-mounted communication device, a vehicle-mounted camera, a scene camera and a dispatch control module; by transporting laboratory items through multiple robots, it solves the problem of unmanned, centralized and pollution-proof distribution management in highly intensive and partially contaminated or radioactive scenes such as medical and testing centers. By setting up a dispatch control module and a positioning module, the travel distances between multiple robots and the item sending requester are obtained, and the travel distances are intelligently sorted to obtain the robot with the shortest distance, and a work instruction is sent to the robot. The robot accepts the item transportation task and executes it. From the perspective of the travel distance, the optimal robot is selected to transport laboratory items, avoiding disordered item transportation.

[0004] The above technology only performs remote centralized control and coordination of the robot based on the travel distance between the robot and the party requesting the item. However, when the manure cleaning robot corresponding to each barn is stopped, the position is basically the same, that is, the distance between each manure cleaning robot and the manure in various places in the barn is basically the same. If only based on the distance, it is impossible to effectively control and coordinate multiple manure cleaning robots, and it is difficult to allocate cleaning tasks in a targeted manner. In addition, the energy of each manure cleaning robot is different, and the energy factor of the robot itself cannot be fully considered based on the distance alone.

[0005] In view of this, the present invention proposes a remote management method for a livestock house manure cleaning robot, a robot and an electronic device to solve the above problems. Summary of the invention

[0006] In order to overcome the above-mentioned defects of the prior art and achieve the above-mentioned purpose, the present invention provides the following technical solution: a remote management method of a livestock house manure cleaning robot, comprising:

[0007] Receive regional data collected by the sensing end;

[0008] Analyze regional data to obtain dirt data;

[0009] Collect power data;

[0010] Transmit dirt data and power data to the cloud platform;

[0011] Receive and execute intelligent scheduling strategies developed by the cloud platform.

[0012] Furthermore, the area data includes area images and dirt point cloud data; the area images are images of n areas in the livestock house to be cleaned; the dirt point cloud data are three-dimensional point cloud data of each pile of dirt in the n areas in the livestock house to be cleaned, and dirt means feces;

[0013] The dirt data includes dirt data of n areas; the dirt data includes dirt area, dirt volume, dirt type and dirt humidity;

[0014] The method for obtaining the pollution area includes:

[0015] The resolution of the regional image is obtained through the attributes of the regional image, and then the pixel area is obtained through the resolution of the regional image, where the pixel area is the area of ​​one pixel; an image segmentation algorithm is used to segment each regional image, and the dirty sub-regions in each regional image are segmented, and the number of pixels in each dirty sub-region is counted; the number of pixels in each dirty sub-region is multiplied by the pixel area to obtain the area of ​​each dirty sub-region; the areas of the dirty sub-regions corresponding to each regional image are added in sequence to obtain the dirty image area of ​​each regional image; a scaling factor is preset, and the dirty image area of ​​each regional image is multiplied by the scaling factor to obtain the dirty area of ​​each region.

[0016] Furthermore, the method for obtaining the pollution volume includes:

[0017] For each set of pollution point cloud data, a three-dimensional reconstruction algorithm is used to construct a pollution three-dimensional model corresponding to each area; for each pollution three-dimensional model, a numerical integration algorithm is used to obtain the pollution volume corresponding to each area.

[0018] Furthermore, the method for obtaining the type of pollution includes:

[0019] Mark the images of the area where the pollution area is not 0 as pollution images; use the trained type recognition model to recognize each pollution image, output digital labels, and obtain the corresponding pollution type according to the digital labels;

[0020] The training process of the category recognition model includes:

[0021] Collect a dirty images in advance, where a is an integer greater than 1, and label each dirty image as a dirty type; convert each dirty type into a digital label; divide the labeled dirty images into a training set and a test set; use the training set to train a type recognition model, and use the test set to test the type recognition model; preset an error threshold, and when the mean of the prediction errors of all dirty images in the test set is less than the error threshold, output a type recognition model; the type recognition model is a convolutional neural network model.

[0022] Furthermore, the method for obtaining the dirt humidity includes:

[0023] Read the RGB value of each pixel in each dirty sub-area and convert it to HSV value. The RGB value includes the value of the red channel, the value of the green channel and the value of the blue channel. The HSV value includes the hue value, the saturation value and the lightness value.

[0024] The methods for converting RGB values ​​to HSV values ​​include:

[0025] Get the maximum value of the RGB value and mark it as , get the minimum value in RGB value and mark it as ;like is 0, the saturation value is also 0, if If it is not 0, the saturation value ; Lightness value ;

[0026] The saturation values ​​of each pixel in each dirty sub-region are sequentially added to obtain the total saturation value of each dirty sub-region, and the total saturation value of each dirty sub-region is divided by the corresponding number of pixels to obtain the average saturation value of each dirty sub-region; the brightness values ​​of each pixel in each dirty sub-region are sequentially added to obtain the total brightness value of each dirty sub-region, and the total brightness value of each dirty sub-region is divided by the corresponding number of pixels to obtain the average brightness value of each dirty sub-region;

[0027] The saturation mean and the brightness mean of each dirty sub-area are weightedly summed to obtain the dirt humidity of each dirty sub-area.

[0028] Furthermore, the steps of formulating the intelligent scheduling strategy by the cloud platform include:

[0029] Step 1: Analyze the received dirt data and calculate the power consumption and cleaning time corresponding to each dirty area, where the dirty area is the area corresponding to the dirty image;

[0030] Step 2: Based on the power consumption and cleaning time of each dirty area and combined with the power data, a simulated annealing algorithm is used to formulate an intelligent scheduling strategy; the power data is the remaining power of the robot corresponding to the livestock house to be cleaned.

[0031] Furthermore, the method for calculating the power consumption and cleaning time corresponding to each dirty area includes:

[0032] The power consumption is the power consumed by the robot when cleaning a dirty area, and the cleaning time is the time consumed by the robot when cleaning a dirty area; the dirt data corresponding to a dirty area is used as a group of analysis data, that is, the analysis data corresponds to the dirty area one by one; each group of analysis data is input into the trained power prediction model to predict the corresponding power consumption; each group of analysis data is input into the trained time prediction model to predict the corresponding cleaning time;

[0033] The training process of the power prediction model includes:

[0034] Collect b groups of analysis data in advance, set corresponding power consumption for each of the b groups of analysis data, where b is an integer greater than 1, and convert the analysis data and the corresponding power consumption into a corresponding set of feature vectors;

[0035] Each group of feature vectors is used as the input of the power prediction model. The power prediction model takes a group of predicted power consumption corresponding to each group of analysis data as output, and takes the actual power consumption corresponding to each group of analysis data as the prediction target, where the actual power consumption is the pre-set power consumption corresponding to the analysis data; minimizing the sum of prediction errors of all analysis data is used as the training target; the power prediction model is trained until the sum of prediction errors reaches convergence and the training is stopped; the power prediction model is a deep neural network model.

[0036] Furthermore, the step of formulating the intelligent scheduling strategy using the simulated annealing algorithm includes:

[0037] Step a: Preset initialization temperature , minimum temperature , Temperature reduction coefficient , maximum number of iterations And m area sets, and let the current temperature , m is an integer greater than 1;

[0038] Step b: Screen the m region sets and obtain A collection of regions, ;

[0039] Step c: Randomly set a feasible solution , a feasible solution That is, the region set;

[0040] Step d: determine the fitness function;

[0041] Step e: Calculate feasible solutions The corresponding fitness ; with feasible solutions For the current point, perform random perturbations in the neighborhood of the current point to obtain a new feasible solution , and calculate the new feasible solution The corresponding fitness ;

[0042] Step f: Calculate the fitness difference ; If the fitness difference , then let , the new feasible solution The value of is given to the feasible solution ; If the fitness difference , then calculate the probability , according to the probability make ;

[0043] Step g: Repeat steps e to f until the number of loops reaches the maximum number of iterations. When , the loop ends and enters step h;

[0044] Step h: Let the current temperature ; Set the maximum number of iterations ;

[0045] Step i: Repeat steps e to h until the current temperature reaches When , the loop ends and a feasible solution is obtained The corresponding area set; according to the obtained area set, the robot is dispatched to clean the livestock house to be cleaned.

[0046] Furthermore, in step a, the n areas of the livestock house to be cleaned are divided m times, each time dividing A collection of sub-regions, is the number of robots corresponding to the livestock house to be cleaned, according to the number of robots divided each time. Sub-region sets are used to construct region sets, and a total of m region sets are obtained. The sub-region sets correspond to the robots one by one.

[0047] In the step b, the power consumption of each dirty area in each sub-area set is added in sequence to obtain the sum of the power consumption corresponding to each sub-area set, and the sum of the power consumption corresponding to each sub-area set in each area set is compared with the power data of the corresponding robot respectively; if there is a sub-area set in the area set whose sum of power consumption is greater than or equal to the power data of the corresponding robot, the corresponding area set is deleted; if the sum of the power consumption of each sub-area set in the area set is less than the power data of the corresponding robot, the corresponding area set is retained.

[0048] Further, in step c, it is feasible to solve The range is A collection of regions;

[0049] In step d, the expression of the fitness function is: ;

[0050] In the formula, For fitness, is the maximum cleaning time;

[0051] According to the feasible solution Corresponding to the area set, obtain the corresponding sub-area set, add the cleaning time of each dirty area in each sub-area set in turn, obtain the total cleaning time of each sub-area set, compare the total cleaning time of each sub-area set, and take the total cleaning time with the largest value as the maximum cleaning time;

[0052] In step f, the fitness difference The expression is ; Probability The expression is: ; In the formula, e is a natural constant; according to probability make Right now The probability of .

[0053] Furthermore, the power data of some charging robots are transmitted to the cloud platform, and the formulated optimization scheduling strategy is received and executed;

[0054] The partial charging robot is the robot with the shortest total cleaning time in the intelligent scheduling strategy; the uncleaned dirty area in the sub-area set corresponding to the best robot is obtained and marked as the uncleaned area; the best robot is the robot with the longest total cleaning time in the intelligent scheduling strategy; the cleaning time of the uncleaned area is added in sequence to obtain the total cleaning time; when the power data of the partial charging robot is half of the total cleaning time, the corresponding power data is transmitted to the cloud platform;

[0055] The steps to develop an optimized scheduling strategy include:

[0056] Step A: Sort the cleaning time of the uncleaned areas from large to small to generate a sorting table;

[0057] Step B: constructing an allocation set, the allocation set includes a first set and a second set; wherein the first set corresponds to the best robot, and the second set corresponds to some charging robots;

[0058] Step C: respectively add the cleaning time in the first set and the second set to obtain the corresponding total time value, and compare them;

[0059] Step D: Add the cleaning time that ranks first in the sorting table to the allocation set with the smallest total time value. If the total time values ​​of the first set and the second set are the same, add the cleaning time that ranks first in the sorting table to the first set; and delete the cleaning time that has been added to the allocation set from the sorting table;

[0060] Step E: loop step C to step D until there is no cleaning time in the sorting table, and the loop ends; dispatch the best robot and some charging robots according to the allocation set to clean the uncleaned area.

[0061] A livestock house manure cleaning robot, running the livestock house manure cleaning robot remote management method, comprising:

[0062] A data receiving module, used to receive regional data collected by the sensing end;

[0063] Data analysis module, used to analyze regional data and obtain dirt data;

[0064] Data acquisition module, used to collect power data;

[0065] A data transmission module, used to transmit dirt data and power data to a cloud platform;

[0066] The strategy receiving module is used to receive and execute the intelligent scheduling strategy formulated by the cloud platform.

[0067] An electronic device comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the remote management method of a livestock house manure cleaning robot when executing the computer program.

[0068] Technical effects and advantages of the remote management method of the livestock house manure cleaning robot, the robot and the electronic equipment of the present invention:

[0069] 1. By receiving the regional images and dirt point cloud data collected by the sensing end and analyzing them, the dirt data of each area is automatically obtained, including the dirt area, volume, type and humidity, etc.; the robot's own power data is collected and transmitted to the cloud platform together with the dirt data; the cloud platform uses deep learning algorithms based on the dirt data and power data to accurately estimate the time and power consumption of the cleaning task, and uses simulated annealing algorithms to formulate intelligent scheduling strategies. According to the intelligent scheduling strategies, the robots are remotely managed so that multiple robots coordinate and cooperate to complete the cleaning task of the livestock house, avoiding the problem of insufficient power when the robots are performing cleaning operations; the energy status of the robots is fully considered to achieve unified remote management and coordinated scheduling of multiple manure cleaning robots; while greatly improving the efficiency and quality of manure cleaning, it also improves the safety and hygiene level of manure cleaning operations, and further improves the level of animal husbandry management.

[0070] 2. Fully consider that when the area of ​​the livestock house is large, there will be a large cleaning time difference between different robots due to different power data; by transmitting the power data of some charging robots to the cloud platform, the cloud platform sorts the uncleaned areas from large to small according to the size of the uncleaned areas, and uses the greedy algorithm to quickly formulate an optimized scheduling strategy, and reasonably allocates the uncleaned areas to some charging robots and the best robot, so that the workload of the two is similar, thereby reducing the cleaning time of the best robot; the rescheduling of some charging robots is realized to maximize the use of the robot's own energy and complete the cleaning tasks within its capabilities, thereby improving cleaning efficiency, optimizing resource utilization, and reducing the cleaning time difference between robots, ensuring that the cleaning tasks are carried out efficiently and orderly. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] Figure 1 This is a schematic diagram of a livestock house manure cleaning robot according to Embodiment 1 of the present invention;

[0072] Figure 2 This is a schematic diagram of multi-port interaction in Example 1 of the present invention;

[0073] Figure 3 This is a schematic diagram of a livestock house manure cleaning robot according to Embodiment 2 of the present invention;

[0074] Figure 4 This is a flow chart of a remote management method for a livestock house manure cleaning robot according to Embodiment 3 of the present invention;

[0075] Figure 5 This is a schematic diagram of an electronic device according to Embodiment 4 of the present invention;

[0076] Figure 6 This is a schematic diagram of a storage medium according to Embodiment 5 of the present invention. DETAILED DESCRIPTION

[0077] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0078] Example 1

[0079] See also Figure 1 As shown, the livestock house manure cleaning robot described in this embodiment includes a data receiving module, a data analysis module, a data acquisition module, a data transmission module and a strategy receiving module; each module is connected by wired and / or wireless means to achieve data transmission between modules;

[0080] The data receiving module is used to receive the regional data collected by the sensing end.

[0081] The sensing end is the sensor equipment in the livestock house to be cleaned; the sensor equipment includes image sensors and lidars.

[0082] The regional data includes regional images and pollution point cloud data;

[0083] The regional images are images of n areas in the livestock house to be cleaned, that is, there are n regional images in total; the filth point cloud data are three-dimensional point cloud data of each pile of filth in the n areas in the livestock house to be cleaned, filth means feces, and there may be multiple piles of filth in one area; the n areas are obtained by technical personnel in this field by dividing the livestock house into regions according to the actual area of ​​the livestock house and the internal structure of the livestock house; the images of the n areas are obtained by image sensors installed above the n areas; and the three-dimensional point cloud data of filth in the n areas are obtained by a laser radar installed above the n areas.

[0084] The data analysis module is used to analyze the regional data and obtain the dirt data.

[0085] The dirt data includes dirt data of n areas; the dirt data includes dirt area, dirt volume, dirt type and dirt humidity.

[0086] Methods for obtaining the pollution area include:

[0087] The resolution of the regional image is obtained through the attributes of the regional image, and then the pixel area is obtained through the resolution of the regional image, where the pixel area is the area of ​​one pixel; wherein the attributes of each regional image are the same; an image segmentation algorithm (such as the Canny edge detection algorithm, the Sobel algorithm, etc.) is used to segment each regional image, and the dirty sub-regions in each regional image are segmented, and the number of pixels in each dirty sub-region is counted; the number of pixels in each dirty sub-region is multiplied by the pixel area to obtain the area of ​​each dirty sub-region; the areas of the dirty sub-regions corresponding to each regional image are added in sequence to obtain the dirty image area of ​​each regional image; a scaling factor is preset, and the dirty image area of ​​each regional image is multiplied by the scaling factor to obtain the dirty area of ​​each region;

[0088] The scaling factor is obtained by technicians in this field by measuring the regional image area and the actual area of ​​the region corresponding to multiple regional images. The regional image area is the area measured by the regional image, and the actual area is the area of ​​the region measured on the spot. The actual area of ​​the region is divided by the regional image area to obtain the area ratio, and the average of multiple area ratios is used as the scaling factor.

[0089] The methods for obtaining the filth volume include:

[0090] For each set of pollution point cloud data, a three-dimensional reconstruction algorithm (such as point cloud reconstruction algorithm, mesh generation algorithm, surface fitting algorithm, etc.) is used to construct a pollution three-dimensional model corresponding to each area; for each pollution three-dimensional model, a numerical integration algorithm (such as Gauss integration method, finite element method, etc.) is used to obtain the pollution volume corresponding to each area.

[0091] The methods of obtaining filth types include:

[0092] The images with non-zero dirt areas are marked as dirt images. The trained type recognition model is used to identify each dirt image, output digital labels, and obtain the corresponding dirt types according to the digital labels. Examples of dirt types include cow dung, pig dung, and sheep dung.

[0093] The specific training process of the category recognition model includes:

[0094] Collect a pieces of dirty images in advance, where a is an integer greater than 1, and label each dirty image as a type of dirty; convert each type of dirty into a digital label, for example, convert cow dung into 1, pig dung into 2, and sheep dung into 3; divide the labeled dirty images into a training set and a test set, with 70% of the dirty images as the training set and 30% of the dirty images as the test set; use the training set to train the type recognition model, and use the test set to test the type recognition model; preset an error threshold, and when the mean of the prediction errors of all dirty images in the test set is less than the error threshold, output the type recognition model; wherein the calculation formula for the mean prediction error is: ,in is the prediction error, is the number of the dirty image, For the The predicted digital labels corresponding to the dirty images, For the The actual digital labels corresponding to the dirty images in the group, U is the number of dirty images in the test set; the error threshold is pre-set according to the accuracy required by the type recognition model;

[0095] The above-mentioned category recognition model is specifically a convolutional neural network model.

[0096] Methods for obtaining dirt humidity include:

[0097] Read the RGB value of each pixel in each dirty sub-area and convert it to HSV value. The RGB value includes the value of the red channel (Red), the value of the green channel (Green) and the value of the blue channel (Blue). The HSV value includes the hue value (Hue), saturation value (Saturation) and brightness value (Value);

[0098] The methods for converting RGB values ​​to HSV values ​​include:

[0099] Get the maximum value of the RGB value and mark it as , get the minimum value in RGB value and mark it as ;like is 0, the saturation value is also 0, if If it is not 0, the saturation value ; Brightness value ; Since only saturation and lightness values ​​are needed to calculate the dirt humidity, there is no need to calculate the hue value.

[0100] The saturation values ​​of each pixel in each dirty sub-region are sequentially added to obtain the total saturation value of each dirty sub-region, and the total saturation value of each dirty sub-region is divided by the corresponding number of pixels to obtain the average saturation value of each dirty sub-region; the brightness values ​​of each pixel in each dirty sub-region are sequentially added to obtain the total brightness value of each dirty sub-region, and the total brightness value of each dirty sub-region is divided by the corresponding number of pixels to obtain the average brightness value of each dirty sub-region;

[0101] The saturation mean and brightness mean of each dirty sub-area are weighted and summed to obtain the dirt humidity of each dirty sub-area; the expression of dirt humidity is: ; is the dirt humidity of the jth dirty sub-area in the i-th dirty area, is the mean saturation value of the jth dirty sub-region in the i-th dirty region, is the average brightness of the jth dirty sub-region in the i-th dirty region, i ∈ [ 1 , N ] , j ∈ [ 1 , M i ] , N is the number of dirty areas, dirty areas are the areas corresponding to the dirty images, is the number of dirty sub-regions of the i-th dirty region, , is a preset proportional coefficient; the specific value of the proportional coefficient can be set according to the actual situation. The proportional coefficient reflects the influence of the saturation mean and the brightness mean on the pollution humidity. The technical personnel in this field can preset the corresponding proportional coefficient according to the actual influence of the saturation mean and the brightness mean on the pollution humidity, so as to accurately evaluate the humidity of the pollution in the pollution sub-area; the saturation mean and the brightness mean in the pollution humidity calculation process are both dimensionless values.

[0102] It should be noted that the larger the dirt area and dirt volume, the more dirt there is in the dirt area, and the more difficult it is for the robot to clean. Therefore, the power consumption and cleaning time are both greater, and vice versa. Different types of dirt have different levels of difficulty in robot cleaning. For example, sheep manure is dry and granular, relatively loose, not easy to adhere, and relatively easy to clean. Cow manure is semi-solid and slightly sticky, which is relatively difficult to clean. Pig manure is semi-liquid with high viscosity and is the most difficult to clean. Therefore, different types of dirt have different corresponding power consumption and cleaning time. The greater the dirt humidity and the more sticky the dirt, the more difficult it is for the robot to clean. Therefore, the power consumption and cleaning time are both greater, and vice versa.

[0103] The data acquisition module is used to collect power data.

[0104] The power data is the remaining power of the robot corresponding to the livestock house to be cleaned, and the robot is a livestock house manure cleaning robot; the power data is obtained from the robot's built-in battery management system.

[0105] The data transmission module is used to transmit the dirt data and power data to the cloud platform.

[0106] The cloud platform is a comprehensive cloud computing service platform; the robot and the cloud platform are connected by wireless network communication, Internet of Things protocol or cloud service provider's SDK to accurately and quickly transmit data; Internet of Things network communication such as WI-FI module, cellular network, etc., Internet of Things protocols such as MQTT, CoAP, LwM2M, etc., cloud service provider SDK such as IoT SDK provided by AWS, Azure, GCP, etc.

[0107] The strategy receiving module is used to receive and execute the intelligent scheduling strategy formulated by the cloud platform.

[0108] The steps for the cloud platform to formulate intelligent scheduling strategies include:

[0109] Step 1: Analyze the received dirt data and calculate the power consumption and cleaning time corresponding to each dirty area, where the dirty area is the area corresponding to the dirty image.

[0110] Methods for calculating power consumption and cleaning time for each dirty area include:

[0111] The power consumption is the power consumed by the robot when cleaning a dirty area, and the cleaning time is the time consumed by the robot when cleaning a dirty area; the dirt data corresponding to a dirty area is taken as a set of analysis data, that is, the analysis data corresponds to the dirty area one by one; each set of analysis data is input into the trained power prediction model to predict the corresponding power consumption; each set of analysis data is input into the trained time prediction model to predict the corresponding cleaning time.

[0112] The specific training process of the power prediction model includes:

[0113] Collect b groups of analysis data in advance, set corresponding power consumption for each of the b groups of analysis data, where b is an integer greater than 1, and convert the analysis data and the corresponding power consumption into a corresponding set of feature vectors; the power consumption corresponding to the analysis data is collected by a technician in this field during the process of cleaning the livestock house with a historical robot, and the corresponding power consumption is measured respectively under the condition of the b group of analysis data, and the corresponding power consumption is set for the b group of analysis data in sequence;

[0114] Each set of feature vectors is used as the input of the power prediction model. The power prediction model uses a set of predicted power consumption corresponding to each set of analysis data as output, and the actual power consumption corresponding to each set of analysis data as the prediction target. The actual power consumption is the pre-set power consumption corresponding to the analysis data. The training target is to minimize the sum of the prediction errors of all analysis data. The calculation formula of the prediction error is: ,in is the prediction error, K is the group number of the eigenvector corresponding to the analysis data, is the predicted power consumption corresponding to the Kth group of analysis data, is the actual power consumption corresponding to the Kth group of analysis data; the power prediction model is trained until the sum of the prediction errors reaches convergence and the training is stopped.

[0115] The above-mentioned electricity prediction model is specifically a deep neural network model; it includes an input layer, a hidden layer and an output layer; each hidden layer includes multiple neurons, each neuron is connected to the neurons in the next layer, and the connection contains weights, which determine the importance and influence of data transmission in the neural network; an activation function is applied to each neuron between the hidden layer and the output layer, and the activation function introduces nonlinearity, allowing the network to learn more complex patterns and features.

[0116] The specific training process of the time prediction model is consistent with the training process of the power prediction model, and both are deep neural network models.

[0117] Step 2: Based on the power consumption and cleaning time of each dirty area and combined with the power data, a simulated annealing algorithm is used to develop an intelligent scheduling strategy.

[0118] The steps of developing intelligent scheduling strategy using simulated annealing algorithm include:

[0119] Step a: Preset initialization temperature , minimum temperature , Temperature reduction coefficient , maximum number of iterations And m area sets, and let the current temperature , m is an integer greater than 1.

[0120] Step b: Screen the m region sets and obtain A collection of regions, .

[0121] Step c: Randomly set a feasible solution , a feasible solution That is the region collection.

[0122] Step d: Determine the fitness function.

[0123] Step e: Calculate feasible solutions The corresponding fitness ; with feasible solutions For the current point, perform random perturbations in the neighborhood of the current point to obtain a new feasible solution , and calculate the new feasible solution The corresponding fitness .

[0124] Step f: Calculate the fitness difference ; If the fitness difference , then let , the new feasible solution The value of is given to the feasible solution ; If the fitness difference , then calculate the probability , according to the probability make .

[0125] Step g: Repeat steps e to f until the number of loops reaches the maximum number of iterations. When , the loop ends and enters step h.

[0126] Step h: Let the current temperature , that is, cool down the current temperature in step a, and assign the value after cooling to the current temperature; let the maximum number of iterations , that is, the value of the reduced maximum number of iterations is assigned to the maximum number of iterations; if the reduced maximum number of iterations is not an integer, the reduced maximum number of iterations is rounded up so that the reduced maximum number of iterations is an integer.

[0127] Step i: Repeat steps e to h until the current temperature reaches When , the loop ends and a feasible solution is obtained The corresponding area set; according to the obtained area set, the robot is dispatched to clean the livestock house to be cleaned.

[0128] It should be noted that the initialization temperature , minimum temperature , Temperature reduction coefficient And the maximum number of iterations As preset parameters, the preset parameters are obtained by a technician in this field during the process of the historical robot cleaning the livestock house, and multiple groups of different test data are obtained. The test data are the power consumption and cleaning time of each dirty area in the livestock house to be cleaned, as well as the power data. For the livestock houses to be cleaned with the same test data, multiple groups of different preset parameters are pre-set in turn, and the simulated annealing algorithm is used in turn to obtain the area sets. According to the multiple groups of area sets obtained, the corresponding fitness is obtained; the preset parameters corresponding to the area set with the largest fitness are used as the preset parameters corresponding to the group of test data, and the preset parameters corresponding to multiple groups of different test data are obtained by analogy, and the mean values ​​of multiple preset parameters (i.e., the initialization temperature mean, the minimum temperature mean, the cooling coefficient mean, and the maximum number of iterations mean) are used as the initialization temperature preset in step a. , minimum temperature , Temperature reduction coefficient And the maximum number of iterations .

[0129] In the above step a, the n areas of the livestock house to be cleaned are divided m times, and each time a A collection of sub-regions, is the number of robots corresponding to the livestock house to be cleaned, according to the number of robots divided each time. Sub-region sets are used to construct region sets, and a total of m region sets are obtained, and the sub-region sets correspond to the robots one by one; each sub-region set includes at least one region, and the regions in each sub-region set are clustered and continuous. For example, assuming that the plane figure of the livestock house to be cleaned is a square and is evenly divided into 4 regions, and 2 sub-region sets are divided each time, the region set should be one of the sub-region sets including region 1 and region 2, and the other sub-region set including region 3 or region 4, or one of the sub-region sets including region 1, and the other sub-region set including region 2, region 3 and region 4, etc.; that is, the regions in the sub-region sets are all clustered and continuous, and there is no situation where the regions included in the sub-region sets are scattered. For example, one of the sub-region sets includes region 1 and region 3, and the other sub-region set includes region 2 and region 4, region 1 and region 3 are scattered, and region 2 and region 4 are also scattered.

[0130] It should be noted that the reason why the areas in the sub-area set are clustered and continuous is to avoid collisions caused by cross-displacement of multiple robots during cleaning of the livestock house and to reduce the movement distance of the robots to improve cleaning efficiency and quality.

[0131] In the above step b, the power consumption of each dirty area in each sub-area set is added in turn to obtain the total power consumption corresponding to each sub-area set, and the total power consumption corresponding to each sub-area set in each area set is compared with the power data of the corresponding robot respectively; if there is a sub-area set in the area set whose total power consumption is greater than or equal to the power data of the corresponding robot, the corresponding area set is deleted; if the total power consumption of each sub-area set in the area set is less than the power data of the corresponding robot, the corresponding area set is retained to complete the screening of the area set.

[0132] In the above step c, the feasible solution is The range is A collection of regions.

[0133] In the above step d, the expression of the fitness function is: ;

[0134] In the formula, For fitness, The maximum cleanup time.

[0135] According to the feasible solution Corresponding to the area set, obtain the corresponding sub-area set, add up the cleaning time of each dirty area in each sub-area set in turn, obtain the total cleaning time of each sub-area set, compare the total cleaning time of each sub-area set, and take the total cleaning time with the largest value as the maximum cleaning time.

[0136] In the above step f, the fitness difference The expression is ; Probability The expression is: ; In the formula, e is a natural constant; according to probability make Right now The probability of .

[0137] It should be noted that the reason for formulating intelligent scheduling strategies based on dirt data and power data is that formulating intelligent scheduling strategies based on power data can effectively avoid robot shutdowns due to power exhaustion during cleaning tasks, thereby improving the safety and work efficiency of cleaning operations and reducing robot idle time; formulating intelligent scheduling strategies based on dirt data can accurately assess the complexity and workload of cleaning tasks, thereby reasonably allocating robots and improving cleaning effects.

[0138] The interaction process between the sensor, robot and cloud platform is as follows: Figure 2 shown.

[0139] This embodiment receives the regional images and dirt point cloud data collected by the sensing end, analyzes them, and automatically obtains the dirt data of each area, including the dirt area, volume, type and humidity. It also collects the power data of the robot itself and transmits it to the cloud platform together with the dirt data. The cloud platform uses a deep learning algorithm based on the dirt data and power data to accurately estimate the time and power consumption of the cleaning task, and adopts a simulated annealing algorithm to formulate an intelligent scheduling strategy. The robot is remotely managed according to the intelligent scheduling strategy, so that multiple robots coordinate and cooperate to complete the cleaning task of the livestock house, avoiding the problem of insufficient power when the robot is performing the cleaning operation. The energy status of the robot is fully considered to realize unified remote management and coordinated scheduling of multiple manure cleaning robots. While greatly improving the efficiency and quality of manure cleaning, it also improves the safety and hygiene level of manure cleaning operations, and further improves the management level of animal husbandry.

[0140] Example 2

[0141] See also Figure 3 As shown, this embodiment further improves the design on the basis of embodiment 1. When the area of ​​the livestock house to be cleaned is large, multiple robots take a long time to perform the cleaning task. If the time difference between the robot that takes the longest time and the robot that takes the shortest time is too large, when the robot that takes the longest time is still performing the cleaning task, the robot that takes the shortest time has completed the corresponding cleaning task and been charged, and the corresponding power data is sufficient to clean multiple dirty areas; if the robot whose power data is partially restored is not remotely controlled and re-entered into the cleaning task, it is impossible to effectively improve the cleaning efficiency and maximize the use efficiency of the robot; therefore, this embodiment provides a livestock house manure cleaning robot, which also includes a resource utilization module;

[0142] The resource utilization module is used to transmit the power data of some charging robots to the cloud platform, and receive and execute the formulated optimization scheduling strategy; to maximize the utilization efficiency of the robots, make full use of their restored power to complete as many cleaning tasks as possible, improve cleaning efficiency, and effectively manage and optimize resource utilization.

[0143] The partial charging robot is the robot with the shortest total cleaning time in the intelligent scheduling strategy; the uncleaned dirty area in the sub-area set corresponding to the best robot is obtained and marked as the uncleaned area; the best robot is the robot with the longest total cleaning time in the intelligent scheduling strategy; the cleaning time of the uncleaned area is added up in sequence to obtain the total cleaning time; when the power data of the partial charging robot is half of the total cleaning time, the corresponding power data is transmitted to the cloud platform.

[0144] The steps to develop an optimized scheduling strategy include:

[0145] Step A: Sort the cleaning time of the uncleaned areas from large to small to generate a sorting table;

[0146] Step B: constructing an allocation set, the allocation set includes a first set and a second set; wherein the first set corresponds to the best robot, and the second set corresponds to some charging robots;

[0147] Step C: respectively add the cleaning time in the first set and the second set to obtain the corresponding total time value, and compare them;

[0148] Step D: Add the cleaning time that ranks first in the sorting table to the allocation set with the smallest total time value. If the total time values ​​of the first set and the second set are the same, add the cleaning time that ranks first in the sorting table to the first set; and delete the cleaning time that has been added to the allocation set from the sorting table;

[0149] Step E: looping steps C to D until there is no cleaning time in the sorting table, and the loop ends; dispatching the best robot and some charging robots according to the allocation set to clean the uncleaned area;

[0150] This embodiment fully considers that when the area of ​​the livestock house is large, there will be a large cleaning time difference between different robots due to different power data; by transmitting the power data of some charging robots to the cloud platform, the cloud platform sorts the uncleaned areas from large to small according to the size of the uncleaned areas, and uses a greedy algorithm to quickly formulate an optimized scheduling strategy, and reasonably allocates the uncleaned areas to some charging robots and the best robot, so that the workload of the two is similar, thereby reducing the cleaning time of the best robot; it realizes the rescheduling of some charging robots to maximize the use of the robots' own energy and complete cleaning tasks within their capabilities, thereby improving cleaning efficiency, optimizing resource utilization, and reducing the cleaning time difference between robots, ensuring that cleaning tasks are carried out efficiently and orderly.

[0151] Example 3

[0152] See also Figure 4 As shown, the parts not described in detail in this embodiment are described in Example 1 and Example 2, and a remote management method for a livestock house manure cleaning robot is provided, the method comprising:

[0153] Receive regional data collected by the sensing end;

[0154] Analyze regional data to obtain dirt data;

[0155] Collect power data;

[0156] Transmit dirt data and power data to the cloud platform;

[0157] Receive and execute intelligent scheduling strategies developed by the cloud platform.

[0158] Furthermore, the area data includes area images and dirt point cloud data; the area images are images of n areas in the livestock house to be cleaned; the dirt point cloud data are three-dimensional point cloud data of each pile of dirt in the n areas in the livestock house to be cleaned, and dirt means feces;

[0159] The dirt data includes dirt data of n areas; the dirt data includes dirt area, dirt volume, dirt type and dirt humidity;

[0160] The method for obtaining the pollution area includes:

[0161] The resolution of the regional image is obtained through the attributes of the regional image, and then the pixel area is obtained through the resolution of the regional image, where the pixel area is the area of ​​one pixel; an image segmentation algorithm is used to segment each regional image, and the dirty sub-regions in each regional image are segmented, and the number of pixels in each dirty sub-region is counted; the number of pixels in each dirty sub-region is multiplied by the pixel area to obtain the area of ​​each dirty sub-region; the areas of the dirty sub-regions corresponding to each regional image are added in sequence to obtain the dirty image area of ​​each regional image; a scaling factor is preset, and the dirty image area of ​​each regional image is multiplied by the scaling factor to obtain the dirty area of ​​each region.

[0162] Furthermore, the method for obtaining the pollution volume includes:

[0163] For each set of pollution point cloud data, a three-dimensional reconstruction algorithm is used to construct a pollution three-dimensional model corresponding to each area; for each pollution three-dimensional model, a numerical integration algorithm is used to obtain the pollution volume corresponding to each area.

[0164] Furthermore, the method for obtaining the type of pollution includes:

[0165] Mark the images of the area where the pollution area is not 0 as pollution images; use the trained type recognition model to recognize each pollution image, output digital labels, and obtain the corresponding pollution type according to the digital labels;

[0166] The training process of the category recognition model includes:

[0167] Collect a dirty images in advance, where a is an integer greater than 1, and label each dirty image as a dirty type; convert each dirty type into a digital label; divide the labeled dirty images into a training set and a test set; use the training set to train a type recognition model, and use the test set to test the type recognition model; preset an error threshold, and when the mean of the prediction errors of all dirty images in the test set is less than the error threshold, output a type recognition model; the type recognition model is a convolutional neural network model.

[0168] Furthermore, the method for obtaining the dirt humidity includes:

[0169] Read the RGB value of each pixel in each dirty sub-area and convert it to HSV value. The RGB value includes the value of the red channel, the value of the green channel and the value of the blue channel. The HSV value includes the hue value, the saturation value and the lightness value.

[0170] The methods for converting RGB values ​​to HSV values ​​include:

[0171] Get the maximum value of the RGB value and mark it as , get the minimum value in RGB value and mark it as ;like is 0, the saturation value is also 0, if If it is not 0, the saturation value ; Lightness value ;

[0172] The saturation values ​​of each pixel in each dirty sub-region are sequentially added to obtain the total saturation value of each dirty sub-region, and the total saturation value of each dirty sub-region is divided by the corresponding number of pixels to obtain the average saturation value of each dirty sub-region; the brightness values ​​of each pixel in each dirty sub-region are sequentially added to obtain the total brightness value of each dirty sub-region, and the total brightness value of each dirty sub-region is divided by the corresponding number of pixels to obtain the average brightness value of each dirty sub-region;

[0173] The saturation mean and the brightness mean of each dirty sub-area are weightedly summed to obtain the dirt humidity of each dirty sub-area.

[0174] Furthermore, the steps of formulating the intelligent scheduling strategy by the cloud platform include:

[0175] Step 1: Analyze the received dirt data and calculate the power consumption and cleaning time corresponding to each dirty area, where the dirty area is the area corresponding to the dirty image;

[0176] Step 2: Based on the power consumption and cleaning time of each dirty area and combined with the power data, a simulated annealing algorithm is used to formulate an intelligent scheduling strategy; the power data is the remaining power of the robot corresponding to the livestock house to be cleaned.

[0177] Furthermore, the method for calculating the power consumption and cleaning time corresponding to each dirty area includes:

[0178] The power consumption is the power consumed by the robot when cleaning a dirty area, and the cleaning time is the time consumed by the robot when cleaning a dirty area; the dirt data corresponding to a dirty area is used as a group of analysis data, that is, the analysis data corresponds to the dirty area one by one; each group of analysis data is input into the trained power prediction model to predict the corresponding power consumption; each group of analysis data is input into the trained time prediction model to predict the corresponding cleaning time;

[0179] The training process of the power prediction model includes:

[0180] Collect b groups of analysis data in advance, set corresponding power consumption for each of the b groups of analysis data, where b is an integer greater than 1, and convert the analysis data and the corresponding power consumption into a corresponding set of feature vectors;

[0181] Each group of feature vectors is used as the input of the power prediction model. The power prediction model takes a group of predicted power consumption corresponding to each group of analysis data as output, and takes the actual power consumption corresponding to each group of analysis data as the prediction target, where the actual power consumption is the pre-set power consumption corresponding to the analysis data; minimizing the sum of prediction errors of all analysis data is used as the training target; the power prediction model is trained until the sum of prediction errors reaches convergence and the training is stopped; the power prediction model is a deep neural network model.

[0182] Furthermore, the step of formulating the intelligent scheduling strategy using the simulated annealing algorithm includes:

[0183] Step a: Preset initialization temperature , minimum temperature , Temperature reduction coefficient , maximum number of iterations And m area sets, and let the current temperature , m is an integer greater than 1;

[0184] Step b: Screen the m region sets and obtain A collection of regions, ;

[0185] Step c: Randomly set a feasible solution , a feasible solution That is, the region set;

[0186] Step d: determine the fitness function;

[0187] Step e: Calculate feasible solutions The corresponding fitness ; with feasible solutions For the current point, perform random perturbations in the neighborhood of the current point to obtain a new feasible solution , and calculate the new feasible solution The corresponding fitness ;

[0188] Step f: Calculate the fitness difference ; If the fitness difference , then let , the new feasible solution The value of is given to the feasible solution ; If the fitness difference , then calculate the probability , according to the probability make ;

[0189] Step g: Repeat steps e to f until the number of loops reaches the maximum number of iterations. When , the loop ends and enters step h;

[0190] Step h: Let the current temperature ; Set the maximum number of iterations ;

[0191] Step i: Repeat steps e to h until the current temperature reaches When , the loop ends and a feasible solution is obtained The corresponding area set; according to the obtained area set, the robot is dispatched to clean the livestock house to be cleaned.

[0192] Furthermore, in step a, the n areas of the livestock house to be cleaned are divided m times, each time dividing A collection of sub-regions, is the number of robots corresponding to the livestock house to be cleaned, according to the number of robots divided each time. Sub-region sets are used to construct region sets, and a total of m region sets are obtained. The sub-region sets correspond to the robots one by one.

[0193] In the step b, the power consumption of each dirty area in each sub-area set is added in sequence to obtain the sum of the power consumption corresponding to each sub-area set, and the sum of the power consumption corresponding to each sub-area set in each area set is compared with the power data of the corresponding robot respectively; if there is a sub-area set in the area set whose sum of power consumption is greater than or equal to the power data of the corresponding robot, the corresponding area set is deleted; if the sum of the power consumption of each sub-area set in the area set is less than the power data of the corresponding robot, the corresponding area set is retained.

[0194] Further, in step c, it is feasible to solve The range is A collection of regions;

[0195] In step d, the expression of the fitness function is: ;

[0196] In the formula, For fitness, is the maximum cleaning time;

[0197] According to the feasible solution Corresponding to the area set, obtain the corresponding sub-area set, add the cleaning time of each dirty area in each sub-area set in turn, obtain the total cleaning time of each sub-area set, compare the total cleaning time of each sub-area set, and take the total cleaning time with the largest value as the maximum cleaning time;

[0198] In step f, the fitness difference The expression is ; Probability The expression is: ; In the formula, e is a natural constant; according to probability make Right now The probability of .

[0199] Furthermore, the power data of some charging robots are transmitted to the cloud platform, and the formulated optimization scheduling strategy is received and executed;

[0200] The partial charging robot is the robot with the shortest total cleaning time in the intelligent scheduling strategy; the uncleaned dirty area in the sub-area set corresponding to the best robot is obtained and marked as the uncleaned area; the best robot is the robot with the longest total cleaning time in the intelligent scheduling strategy; the cleaning time of the uncleaned area is added in sequence to obtain the total cleaning time; when the power data of the partial charging robot is half of the total cleaning time, the corresponding power data is transmitted to the cloud platform;

[0201] The steps to develop an optimized scheduling strategy include:

[0202] Step A: Sort the cleaning time of the uncleaned areas from large to small to generate a sorting table;

[0203] Step B: constructing an allocation set, the allocation set includes a first set and a second set; wherein the first set corresponds to the best robot, and the second set corresponds to some charging robots;

[0204] Step C: respectively add the cleaning time in the first set and the second set to obtain the corresponding total time value, and compare them;

[0205] Step D: Add the cleaning time that ranks first in the sorting table to the allocation set with the smallest total time value. If the total time values ​​of the first set and the second set are the same, add the cleaning time that ranks first in the sorting table to the first set; and delete the cleaning time that has been added to the allocation set from the sorting table;

[0206] Step E: loop step C to step D until there is no cleaning time in the sorting table, and the loop ends; dispatch the best robot and some charging robots according to the allocation set to clean the uncleaned area.

[0207] Example 4

[0208] See also Figure 5 As shown, the present application also provides an electronic device 500. The electronic device 500 may include one or more processors and one or more memories. The memories store computer readable codes, and when the computer readable codes are executed by one or more processors, the remote management method of the livestock house manure cleaning robot as described above can be executed.

[0209] The method or system according to the embodiment of the present application can also be used by Figure 5 The electronic device architecture shown in FIG. Figure 5 As shown, the electronic device 500 may include a bus 501, one or more CPUs 502, a ROM 503, a RAM 504, a communication port 505 connected to a network, an input / output 506, a hard disk 507, etc. The storage device in the electronic device 500, such as the ROM 503 or the hard disk 507, may store the remote management method of the livestock house manure cleaning robot provided in the present application. Furthermore, the electronic device 500 may also include a user interface 508. Of course, Figure 5 The architecture shown is only exemplary and can be omitted according to actual needs when implementing different devices. Figure 5 One or more components of an electronic device are shown.

[0210] Example 5

[0211] See also Figure 6 As shown, one embodiment of the present application discloses a storage medium 600. Computer-readable instructions are stored on the storage medium 600. When the computer-readable instructions are executed by the processor, the remote management method of the livestock house manure cleaning robot according to the embodiment of the present application described with reference to the above figures can be executed. The storage medium 600 includes, but is not limited to, for example, volatile memory and / or non-volatile memory. Volatile memory may include, for example, random access memory (RAM) and cache memory (cache), etc. Non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc.

[0212] In addition, according to the embodiments of the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the present application provides a non-transitory machine-readable storage medium, which stores machine-readable instructions, and the machine-readable instructions can be executed by a processor to execute instructions corresponding to the method steps provided by the present application, such as: a remote management method for a livestock house manure cleaning robot. When the computer program is executed by a central processing unit (CPU), the above functions defined in the method of the present application are executed.

[0213] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

[0214] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A remote management method for a livestock house manure cleaning robot, characterized in that: include: Receive regional data collected by the sensing end; the regional data includes regional images and pollution point cloud data; The area images are images of n areas in the livestock house to be cleaned; Analyze the regional data to obtain dirt data; the dirt data includes dirt data of n regions; the dirt data includes dirt area, dirt volume, dirt type and dirt humidity; mark the image of the region where the dirt area is not 0 as a dirt image; Collect power data; Transmit dirt data and power data to the cloud platform; Receive and execute the intelligent scheduling strategy formulated by the cloud platform; The steps of formulating the intelligent scheduling strategy on the cloud platform include: Step 1: Analyze the received dirt data and calculate the power consumption and cleaning time corresponding to each dirty area, where the dirty area is the area corresponding to the dirty image; Step 2: Based on the power consumption and cleaning time of each dirty area, combined with the power data, a simulated annealing algorithm is used to formulate an intelligent scheduling strategy; the power data is the remaining power of the robot corresponding to the livestock house to be cleaned; The n areas of the livestock house to be cleaned are divided m times respectively, and each time A collection of sub-regions, is the number of robots corresponding to the livestock house to be cleaned, according to the number of robots divided each time. Sub-region sets are used to construct region sets, and a total of m region sets are obtained. The sub-region sets correspond to the robots one by one. The cleaning time of each dirty region in each sub-region set is added up in sequence to obtain the total cleaning time of each sub-region set. Transmit the power data of some charging robots to the cloud platform, and receive and execute the formulated optimization scheduling strategy; The partial charging robot is the robot with the shortest total cleaning time in the intelligent scheduling strategy; the uncleaned dirty area in the sub-area set corresponding to the best robot is obtained and marked as the uncleaned area; the best robot is the robot with the longest total cleaning time in the intelligent scheduling strategy; the cleaning time of the uncleaned area is added in sequence to obtain the total cleaning time; when the power data of the partial charging robot is half of the total cleaning time, the corresponding power data is transmitted to the cloud platform; The steps to develop an optimized scheduling strategy include: Step A: Sort the cleaning time of the uncleaned areas from large to small to generate a sorting table; Step B: constructing an allocation set, the allocation set includes a first set and a second set; wherein the first set corresponds to the best robot, and the second set corresponds to some charging robots; Step C: respectively add the cleaning time in the first set and the second set to obtain the corresponding total time value, and compare them; Step D: Add the cleaning time that ranks first in the sorting table to the allocation set with the smallest total time value. If the total time values ​​of the first set and the second set are the same, add the cleaning time that ranks first in the sorting table to the first set; and delete the cleaning time that has been added to the allocation set from the sorting table; Step E: loop step C to step D until there is no cleaning time in the sorting table, and the loop ends; dispatch the best robot and some charging robots according to the allocation set to clean the uncleaned area.

2. The remote management method of the livestock house manure cleaning robot according to claim 1, characterized in that: The dirt point cloud data is three-dimensional point cloud data of each pile of dirt in n areas in the livestock house to be cleaned, where dirt means feces; The method for obtaining the pollution area includes: The resolution of the regional image is obtained through the attributes of the regional image, and then the pixel area is obtained through the resolution of the regional image, where the pixel area is the area of ​​one pixel; an image segmentation algorithm is used to segment each regional image, and the dirty sub-regions in each regional image are segmented, and the number of pixels in each dirty sub-region is counted; the number of pixels in each dirty sub-region is multiplied by the pixel area to obtain the area of ​​each dirty sub-region; the areas of the dirty sub-regions corresponding to each regional image are added in sequence to obtain the dirty image area of ​​each regional image; a scaling factor is preset, and the dirty image area of ​​each regional image is multiplied by the scaling factor to obtain the dirty area of ​​each region.

3. The remote management method of the livestock house manure cleaning robot according to claim 2, characterized in that: The method for obtaining the pollution volume includes: For each set of pollution point cloud data, a three-dimensional reconstruction algorithm is used to construct a pollution three-dimensional model corresponding to each area; for each pollution three-dimensional model, a numerical integration algorithm is used to obtain the pollution volume corresponding to each area.

4. The remote management method of the livestock house manure cleaning robot according to claim 3, characterized in that: The method for obtaining the type of filth includes: Use the trained category recognition model to identify each dirty image, output digital labels, and obtain the corresponding dirt category based on the digital labels; The training process of the category recognition model includes: Collect a dirty images in advance, where a is an integer greater than 1, and label each dirty image as a dirty type; convert each dirty type into a digital label; divide the labeled dirty images into a training set and a test set; use the training set to train a type recognition model, and use the test set to test the type recognition model; preset an error threshold, and when the mean of the prediction errors of all dirty images in the test set is less than the error threshold, output a type recognition model; the type recognition model is a convolutional neural network model.

5. The remote management method of the livestock house manure cleaning robot according to claim 4, characterized in that: The method for obtaining the dirt humidity comprises: Read the RGB value of each pixel in each dirty sub-area and convert it to HSV value. The RGB value includes the value of the red channel, the value of the green channel and the value of the blue channel. The HSV value includes the hue value, the saturation value and the lightness value. The methods for converting RGB values ​​to HSV values ​​include: Get the maximum value of the RGB value and mark it as , get the minimum value in RGB value and mark it as ;like is 0, the saturation value is also 0, if If it is not 0, the saturation value ; Lightness value ; The saturation values ​​of each pixel in each dirty sub-region are sequentially added to obtain the total saturation value of each dirty sub-region, and the total saturation value of each dirty sub-region is divided by the corresponding number of pixels to obtain the average saturation value of each dirty sub-region; the brightness values ​​of each pixel in each dirty sub-region are sequentially added to obtain the total brightness value of each dirty sub-region, and the total brightness value of each dirty sub-region is divided by the corresponding number of pixels to obtain the average brightness value of each dirty sub-region; The saturation mean and the brightness mean of each dirty sub-area are weightedly summed to obtain the dirt humidity of each dirty sub-area.

6. The remote management method of the livestock house manure cleaning robot according to claim 5, characterized in that: The method for calculating the power consumption and cleaning time corresponding to each dirty area includes: The power consumption is the power consumed by the robot when cleaning a dirty area, and the cleaning time is the time consumed by the robot when cleaning a dirty area; the dirt data corresponding to a dirty area is used as a group of analysis data, that is, the analysis data corresponds to the dirty area one by one; each group of analysis data is input into the trained power prediction model to predict the corresponding power consumption; each group of analysis data is input into the trained time prediction model to predict the corresponding cleaning time; The training process of the power prediction model includes: Collect b groups of analysis data in advance, set corresponding power consumption for each of the b groups of analysis data, where b is an integer greater than 1, and convert the analysis data and the corresponding power consumption into a corresponding set of feature vectors; Each group of feature vectors is used as the input of the power prediction model. The power prediction model takes a group of predicted power consumption corresponding to each group of analysis data as output, and takes the actual power consumption corresponding to each group of analysis data as the prediction target, where the actual power consumption is the pre-set power consumption corresponding to the analysis data; minimizing the sum of prediction errors of all analysis data is used as the training target; the power prediction model is trained until the sum of prediction errors reaches convergence and the training is stopped; the power prediction model is a deep neural network model.

7. The remote management method of the livestock house manure cleaning robot according to claim 6, characterized in that: The steps of formulating the intelligent scheduling strategy by using the simulated annealing algorithm include: Step a: Preset initialization temperature , minimum temperature , Temperature reduction coefficient , maximum number of iterations And m area sets, and let the current temperature , m is an integer greater than 1; Step b: Screen the m region sets and obtain A collection of regions, ; Step c: Randomly set a feasible solution , a feasible solution That is, the region set; Step d: determine the fitness function; Step e: Calculate feasible solutions The corresponding fitness ; with feasible solutions For the current point, perform random perturbations in the neighborhood of the current point to obtain a new feasible solution , and calculate the new feasible solution The corresponding fitness ; Step f: Calculate the fitness difference ; If the fitness difference , then let , the new feasible solution The value of is given to the feasible solution ; If the fitness difference , then calculate the probability , according to the probability make ; Step g: Repeat steps e to f until the number of loops reaches the maximum number of iterations. When , the loop ends and enters step h; Step h: Let the current temperature ; Set the maximum number of iterations ; Step i: Repeat steps e to h until the current temperature reaches When , the loop ends and a feasible solution is obtained The corresponding area set; according to the obtained area set, the robot is dispatched to clean the livestock house to be cleaned.

8. The remote management method of the livestock house manure cleaning robot according to claim 7, characterized in that: In the step b, the power consumption of each dirty area in each sub-area set is added in sequence to obtain the sum of the power consumption corresponding to each sub-area set, and the sum of the power consumption corresponding to each sub-area set in each area set is compared with the power data of the corresponding robot respectively; if there is a sub-area set in the area set whose sum of power consumption is greater than or equal to the power data of the corresponding robot, the corresponding area set is deleted; if the sum of the power consumption of each sub-area set in the area set is less than the power data of the corresponding robot, the corresponding area set is retained.

9. The remote management method of the livestock house manure cleaning robot according to claim 8, characterized in that: In step c, the feasible solution is The range is A collection of regions; In step d, the expression of the fitness function is: ; In the formula, For fitness, is the maximum cleaning time; According to feasible solution Corresponding to the area set, obtain the corresponding sub-area set, compare the total cleaning time of each sub-area set, and take the total cleaning time with the largest value as the maximum cleaning time; In step f, the fitness difference The expression is ; Probability The expression is: ; In the formula, e is a natural constant; according to probability make Right now The probability of .

10. A livestock house manure cleaning robot, running the livestock house manure cleaning robot remote management method according to any one of claims 1 to 9, characterized in that: include: A data receiving module, used to receive regional data collected by the sensing end; Data analysis module, used to analyze regional data and obtain dirt data; Data acquisition module, used to collect power data; A data transmission module, used to transmit dirt data and power data to a cloud platform; The strategy receiving module is used to receive and execute the intelligent scheduling strategy formulated by the cloud platform.

11. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the remote management method for the livestock house manure cleaning robot according to any one of claims 1 to 9 is implemented.