Road surface cleaning control system and method for environmental sanitation equipment

Abstract

The invention relates to a road surface cleaning control system and method for environmental sanitation equipment, and the system comprises a road surface condition data collection module, a cleaning parameter prediction module, an equipment operation parameter optimization module and an execution module, firstly, the road surface condition data is collected through a sensor and features are extracted, and then an improved SVR model is used for prediction; and then multi-objective optimization is carried out through NES, and finally cleaning parameters are dynamically adjusted according to a prediction result and optimization output. According to the invention, efficient and intelligent operation of environmental sanitation equipment under different road surface conditions is ensured.

Description

Technical Field

[0001] The present invention relates to the technical field of sanitation equipment control, and in particular to a road surface cleaning control system and method for sanitation equipment. Background Art

[0002] With the accelerating pace of urbanization, the demand for sanitation equipment is also increasing. However, most sanitation equipment currently on the market has significant limitations in its cleaning efficiency and effectiveness when dealing with different road conditions (including dry, wet, and oily environments). The main problem is that most of these devices rely on preset operating parameters and lack the ability to adapt to real-time road conditions, which directly leads to unsatisfactory cleaning results.

[0003] For example, after a rainfall, using traditional sanitation equipment to clean the streets presents numerous challenges. Because the equipment's operating parameters remain fixed, they struggle to remove accumulated water and oil, impacting overall cleaning quality. Specifically:

[0004] 1. On dry roads, although fixed-parameter equipment can effectively remove dust and small debris, it is helpless against larger debris.

[0005] 2. When encountering wet roads, the same equipment settings are likely to cause water splashing, which not only affects the safety of pedestrians, but also fails to effectively remove mud and slippery materials.

[0006] 3. These devices are even more difficult to deal with roads covered with oil stains because they cannot deeply remove the oil stains, leaving a slippery surface and posing potential safety hazards to pedestrians.

[0007] In addition to the aforementioned cleaning performance issues, the use of fixed parameters also creates additional costs. For example, when cleaning busy streets, the equipment's inability to flexibly adjust its operating mode to meet actual conditions necessitated repeated cleaning cycles to achieve satisfactory cleaning standards. This process not only consumed significant amounts of fuel and prolonged operation time, but also indirectly increased operating costs.

[0008] In summary, a common problem with existing sanitation equipment is that it typically uses fixed operating parameters and lacks the ability to adapt to varying road conditions. This not only affects cleaning effectiveness but also results in high energy consumption and long operation times. Therefore, developing new sanitation equipment that can intelligently identify and adapt to various road conditions is key to improving urban environmental sanitation management. Summary of the Invention

[0009] The object of the present invention is to provide a road cleaning control system and method for sanitation equipment to ensure efficient and intelligent operation of the sanitation equipment under different road conditions.

[0010] In a first aspect, a road cleaning control system for sanitation equipment is characterized by comprising:

[0011] A road condition data acquisition module is used to collect road condition data, wherein the road condition data includes road humidity, road temperature, road pollutant type and road pollutant distribution;

[0012] a cleaning parameter prediction module, which predicts cleaning parameters using a trained support vector regression model based on the equipment operating parameters and the road condition data collected by the road condition data collection module. The cleaning parameters include cleaning effect, energy consumption, and operation time;

[0013] The equipment operating parameter optimization module uses a multi-objective optimization function and combines it with the cleaning parameters predicted by the cleaning parameter prediction module to optimize the current equipment operating parameters. The multi-objective optimization function includes cleaning effect, energy consumption, and operation time as optimization targets. The module then performs secondary optimization through Pareto frontier exploration and selection of non-dominated solutions, as well as natural evolution strategy, to obtain the final optimized equipment operating parameters.

[0014] An execution module operates the device according to the device operating parameters optimized by the device operating parameter optimization module.

[0015] Optionally, the road condition data acquisition module includes:

[0016] Humidity sensor, non-contact installed at the front end of sanitation equipment, used to measure road humidity;

[0017] Temperature sensor, non-contact installed at the front end of sanitation equipment, used to measure road surface temperature;

[0018] The visual sensor is installed non-contactly above the front end of the sanitation equipment to capture road surface images and identify the type and distribution of pollutants.

[0019] Optionally, the support vector regression model is trained by the following steps:

[0020] Obtain training data based on historical operation records, including road condition characteristics, equipment operating parameters and corresponding cleaning parameters;

[0021] Select the support vector regression model with RBF kernel function, and select the optimal model parameters through cross-validation and grid search method, including regularization parameters, RBF kernel function parameters and insensitive loss function parameters;

[0022] The support vector regression model is trained using the training data so that it can predict the cleaning parameters based on the input feature vector of the current road condition data and the equipment operating parameters.

[0023] Optionally, the equipment operation parameter optimization module adopts a method of dynamically adjusting weight coefficients to dynamically adjust the weights of cleaning effect, energy consumption and operation time in the multi-objective optimization function according to the current road condition characteristics, so as to find a balance point between multiple objectives.

[0024] Optionally, the Pareto front exploration includes:

[0025] Initial device operating parameter population generation: Using the device operating parameters to be optimized as a basis, a set of initial parameter settings is randomly generated;

[0026] Cleaning parameter evaluation: Use support vector regression model to predict the cleaning parameters under each set of parameter settings;

[0027] Select non-dominated solutions: Select non-dominated solutions based on the Pareto principle;

[0028] Mutation: Perform mutation operations on the selected non-dominated solutions to generate new parameter settings;

[0029] Iteration: Repeating the cleaning parameter evaluation, selecting non-dominated solutions, and mutation steps until a first convergence condition is met or a first maximum number of iterations is reached;

[0030] That is, a set of non-dominated solutions is obtained through Pareto front exploration, and the optimization is completed in one go.

[0031] Optionally, the natural evolution strategy is used to select a non-dominated solution from the first optimization result for secondary optimization, by calculating the natural gradient of the objective function with respect to the parameters and updating the parameters using an adaptive learning rate until a second convergence condition is met or a second maximum number of iterations is reached.

[0032] In a second aspect, the present invention provides a road cleaning control method for sanitation equipment, comprising the following steps:

[0033] Collect road condition data, including road moisture, road temperature, road pollutant types, and road pollutant distribution;

[0034] Based on the road condition data and equipment operating parameters, the trained support vector regression model is used to predict cleaning parameters, including cleaning effect, energy consumption, and operation time;

[0035] A multi-objective optimization function is used to optimize the current equipment operating parameters in combination with the predicted cleaning parameters. The multi-objective optimization function includes cleaning effect, energy consumption, and operation time as optimization objectives. Secondary optimization is performed through Pareto frontier exploration and selection of non-dominated solutions, as well as a natural evolution strategy, to obtain the final optimized equipment operating parameters.

[0036] Operate the equipment according to the optimized equipment operating parameters.

[0037] Optionally, the support vector regression model training step includes:

[0038] Obtain training data based on historical operation records, including road condition characteristics, equipment operating parameters and corresponding cleaning parameters;

[0039] Select the support vector regression model with RBF kernel function, and select the optimal model parameters through cross-validation and grid search method, including regularization parameters, RBF kernel function parameters and insensitive loss function parameters;

[0040] The support vector regression model is trained using the training data so that it can predict the cleaning parameters based on the input feature vector of the current road condition data and the equipment operating parameters.

[0041] Optionally, in the step of optimizing the equipment operating parameters, a method of dynamically adjusting the weight coefficient is adopted to dynamically adjust the weights of the cleaning effect, energy consumption and operation time in the multi-objective optimization function according to the current road condition characteristics, so as to find a balance point between multiple objectives.

[0042] Optionally, the step of exploring the Pareto frontier includes:

[0043] Initial device operating parameter population generation: Using the device operating parameters to be optimized as a basis, a set of initial parameter settings is randomly generated;

[0044] Cleaning parameter evaluation: Use support vector regression model to predict the cleaning parameters under each set of parameter settings;

[0045] Select non-dominated solutions: Select non-dominated solutions based on the Pareto principle;

[0046] Mutation: Perform mutation operations on the selected non-dominated solutions to generate new parameter settings;

[0047] Iteration: Repeating the steps of cleaning parameter evaluation, selecting non-dominated solutions, and mutation until a first convergence condition is met or a first maximum number of iterations is reached;

[0048] That is, a set of non-dominated solutions is obtained through Pareto front exploration;

[0049] The natural evolution strategy (NES) is used to select a non-dominated solution from the primary optimization results for secondary optimization by calculating the natural gradient of the objective function with respect to the parameters and updating the parameters using an adaptive learning rate until a second convergence condition is met or a second maximum number of iterations is reached.

[0050] Beneficial effects of the present invention:

[0051] The present invention combines the support vector regression model (SVR model) with the natural evolution strategy (NES) and introduces Pareto frontier exploration within a multi-objective optimization framework to achieve efficient optimization of sanitation equipment operating parameters to optimally control road clearing operations. The support vector regression model is used to predict the performance indicators of the equipment under different operating parameters, the Pareto frontier exploration screens out non-dominated solutions that achieve the optimal balance between multiple performance objectives, and the natural evolution strategy generates and optimizes multiple parameter combinations to form the optimal solution. This method not only improves the accuracy and efficiency of optimization, but also adapts to complex optimization scenarios, provides more reliable optimization results, and significantly improves the overall performance and work efficiency of sanitation equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 This is a principle block diagram of a road surface cleaning control system for sanitation equipment described in an embodiment of the present application;

[0053] Figure 2 is a flow chart of a road surface cleaning control method for sanitation equipment described in an embodiment of the present application;

[0054] Figure 3 is a flow chart of the steps for collecting road condition data described in an embodiment of the present application;

[0055] Figure 4 It is a flowchart of the steps of optimizing equipment operating parameters based on predicted cleaning parameters described in the embodiments of the present application. DETAILED DESCRIPTION

[0056] The following will describe the embodiments of the present invention with reference to the accompanying drawings and preferred embodiments. Those skilled in the art will be able to understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. It should be understood that the preferred embodiments are only for the purpose of illustrating the present invention and are not intended to limit the scope of protection of the present invention.

[0057] like Figure 1As shown in the embodiment of the present application, a road cleaning control system for sanitation equipment includes a road condition data acquisition module, a cleaning parameter prediction module, an equipment operating parameter optimization module, and an execution module. By introducing an improved support vector regression model (SVR model) and a natural evolution strategy (NES), this system achieves real-time perception and adaptive adjustment of road conditions, thereby improving cleaning results, reducing energy consumption, and shortening operation time.

[0058] The following is a detailed description of each module:

[0059] 1. Road condition data acquisition module

[0060] Used to collect road condition data, the road condition data (X) includes: road humidity (h), road temperature (T), road pollutant type (l) and road pollutant distribution (d).

[0061] 2. Clean up the parameter prediction module

[0062] Based on pavement condition data and equipment operating parameters (A), a support vector regression model is trained to predict cleaning parameters (B). The support vector regression model can predict cleaning parameters based on the input pavement condition characteristics. The predicted cleaning parameters include cleaning effect (R), energy consumption (E) and operation time (t).

[0063] 3. Equipment operating parameter optimization module

[0064] A multi-objective optimization function is used to optimize the current equipment operating parameters in combination with the cleaning parameters predicted by the cleaning parameter prediction module. The multi-objective optimization function includes cleaning effect, energy consumption and operation time as optimization objectives, and secondary optimization is performed through Pareto frontier exploration and selection of non-dominated solutions, as well as natural evolution strategy, to obtain the final optimized equipment operating parameters.

[0065] 4. Execution module

[0066] The equipment is operated according to the equipment operating parameters optimized by the equipment operating parameter optimization module.

[0067] This road cleaning control system for sanitation equipment solves the following problems:

[0068] 1. The cleaning effect is not ideal. The fixed parameter equipment cannot be adjusted adaptively, resulting in poor cleaning effect under specific road conditions (such as wet road splashing water and slippery oily road).

[0069] 2. High energy consumption. Fixed parameters lead to multiple operations, increasing fuel consumption and operating costs.

[0070] 3. The operation time is long and the parameters cannot be adjusted according to the road conditions, which prolongs the operation time.

[0071] like Figure 2 As shown, in an embodiment of the present application, a road surface cleaning control method for sanitation equipment is provided, which is based on an improved support vector regression (SVR) and natural evolution strategy (NES). First, road surface condition data is collected and features are extracted through sensors. Then, an improved SVR model is used for prediction. Then, multi-objective optimization is performed through NES. Finally, cleaning parameters are dynamically adjusted according to the prediction results and optimization outputs to ensure efficient and intelligent operation of sanitation equipment under different road conditions.

[0072] The following is a detailed description of the various steps of the road cleaning control method for sanitation equipment:

[0073] 1. Collect road condition data

[0074] like Figure 3 As shown, road condition data are collected: road moisture, road temperature, road pollutant type and road pollutant distribution.

[0075] 1.1 Sensor Installation

[0076] Humidity sensor: non-contact, installed at the front end of sanitation equipment, used to measure road humidity.

[0077] Temperature sensor: non-contact, installed at the front end of sanitation equipment, used to measure road surface temperature.

[0078] Visual sensor: non-contact, installed above the front end of sanitation equipment, used to capture road surface images and identify pollutant types and distribution.

[0079] Among them, pollutant types are divided into the following categories:

[0080] Dust: fine particles of matter.

[0081] Mud: A sticky mixture of water and soil.

[0082] Oil pollution: grease pollution.

[0083] Garbage: Large solid waste.

[0084] Pollutant distribution: Images captured by visual sensors are used to determine the distribution of pollutants using image processing techniques (such as edge detection and region segmentation) and quantify them into a numerical value d.

[0085] 1.2 Sensor Data Processing

[0086] The sensor data processing step formats and extracts features based on the data collected by the sensors. Specifically, conventional denoising, filtering, and normalization are performed to form the road condition feature X = [h, t, l, d].

[0087] 2. Predict cleaning parameters based on road condition data and equipment operating parameters

[0088] This step trains a support vector regression (SVR) model so that it can predict cleaning parameters (including cleaning effect, energy consumption, and operation time) based on the input road condition characteristics and equipment operating parameters.

[0089] 2.1 SVR model training

[0090] This step trains an SVR model to predict cleaning effect, energy consumption, and operation time based on the input feature X.

[0091] ① Source of training data

[0092] Based on the historical operation records of the sanitation equipment obtained in the previous step of road condition data collection (for example, one month's data), including the road condition characteristics X at each operation i (i-th time), and the corresponding equipment operating parameters A i and cleanup parameter B i .

[0093] Equipment operating parameters A i Including: brush speed v i , suction i , water pressure p i , obtained from the device's built-in display recording device.

[0094] Cleanup parameter B i Includes: Cleaning Effect R i Energy consumption E i and operation time t i , cleaning effect R i The visual sensor installed at the rear of the equipment takes the image of the cleaned road surface and uses image processing technology to evaluate the cleanliness. i The energy consumption of the equipment is monitored in real time through the equipment's energy consumption sensors (such as current sensors and voltage sensors), and the operation time t i The time required for the equipment to complete one operation is recorded by the built-in timer of the equipment. i =[v i ,s i ,p i ], clean up parameter B i =[R i ,E i ,t i].

[0095] The road condition feature X i and equipment operating parameters A i Combined into a comprehensive feature vector Z i :Z i =[X i ,A i ], and Z i =[X i ,A i ] and B i Corresponding.

[0096] ②Model selection

[0097] The SVR model of the RBF kernel function is selected, and its mathematical expression is:

[0098]

[0099] Among them, K(Z,Z i )=exp(-γ‖ZZ i ‖ 2 ) is the RBF kernel function, Z is the current real-time data, Z i is the comprehensive feature vector of the i-th training sample containing road conditions and equipment operating parameters, α i and b are model parameters, α i It is a Lagrange multiplier used to solve the optimization problem of SVR. During the training process, each training sample Z is obtained by solving a quadratic programming problem. i The corresponding α i , b is the bias term of the SVR model, which is used to adjust the intercept of the regression function.

[0100] ③Model parameter optimization

[0101] Cross-validation: Divide historical data into a training set and a validation set, use the training set to train the model, and use the validation set to evaluate the model performance.

[0102] Grid search: Perform grid search within a predefined parameter range to find the best C, γ, and ∈, where C is the regularization parameter that controls the complexity and overfitting risk of the model, γ is the parameter of the RBF kernel function that controls the width of the kernel function, and ∈ is the parameter of the insensitive loss function that controls the model's tolerance to training errors. For example, set C∈[0.1,1,10], γ∈[0.01,0.1,1], ∈∈[0.01,0.1,1], use the mean square error (MSE) as the evaluation metric, select the parameter combination with the smallest MSE, and obtain the optimal parameter combination C, γ, ∈. Finally, train the SVR model with the optimal parameter combination C, γ, ∈ to obtain the final model parameter α iSpecifically, the optimal parameter combination C,γ,∈ is used to train the SVR model to obtain the final model parameter α i , by selecting the support vector (i.e. α i >0) to calculate b.

[0103] After the above training, the SVR model has the optimal model parameters and can predict the cleaning parameters, namely cleaning effect, energy consumption and operation time, based on the input feature vector of the current road condition data and equipment operating parameters.

[0104] 2.2SVR model prediction

[0105] Use the trained model parameters α i and b, predicting the cleaning parameters based on the feature vector of the current pavement condition data and the equipment operating parameters.

[0106] Based on the road condition data collection step, the road surface humidity, road surface temperature, road surface pollutant type, and road surface pollutant distribution are collected in real time. The current brush speed v, suction force s, and water pressure p of the device are obtained through the built-in display and recording device of the device. Finally, the cleaning effect, energy consumption, and operation time are calculated using the following formula:

[0107]

[0108] in, is the eigenvector of the support vector. During the training process, only those corresponding Lagrange multipliers α i Non-zero training samples Z i will be selected as support vectors. S is the index set of support vectors. The other letters in the formula have the same meanings as above.

[0109] 3. Optimize equipment operating parameters based on predicted cleaning parameters

[0110] like Figure 4 As shown, 3.1 Multi-objective optimization function

[0111] The cleaning effect, energy consumption and operation time are combined into a multi-objective optimization problem, and the objective function is:

[0112] J(θ)=w1R+w2E+w3t

[0113] Where w1, w2, and w3 are weight coefficients, representing the importance of cleaning effect, energy consumption, and operation time, respectively, and their sum equals 1. θ represents the equipment operating parameters to be optimized, namely, brush speed v, suction force s, and water pressure p.

[0114] By setting w1, w2, and w3, a balance is found between cleaning effect, energy consumption, and operation time. In theory, cleaning effect should be maximized to ensure road cleanliness, energy consumption should be minimized to improve energy efficiency, and operation time should be minimized to improve work efficiency. An optimal point needs to be found among the three. Specifically, the weight coefficients w1, w2, and w3 are dynamically adjusted according to the current road condition characteristics. The following is a detailed explanation:

[0115] Assuming that the road pollution level after data standardization and normalization is y1, the temperature is y2, and the humidity is y3, then w1, w2, and w3 are calculated according to the following formula.

[0116]

[0117] w3=1-w2-w3

[0118] The road contamination level is the product or weighted average of the road contaminant type and its distribution. Temperature refers to the road surface temperature, and humidity refers to the road surface humidity. β1, β2, and β3 are adjustment parameters used to control the magnitude of weight changes. They were determined through historical data analysis, expert experience, and experimental verification. w1 is proportional to the road contamination level; higher levels of contamination are weighted more heavily by the cleaning effect, R. w2 is proportional to the temperature; higher temperatures are weighted more heavily by energy consumption, E. w3 is the remainder, representing the weight of the operating time. This allows for dynamic adjustment of the weight coefficients, paving the way for optimized equipment operating parameters.

[0119] 3.2 One-time optimization

[0120] Pareto front exploration is used to optimize the predicted cleaning parameters to obtain a set of non-dominated solutions that strike a balance between multiple objectives. Each non-dominated solution represents a different trade-off, allowing one to be selected based on specific needs. Pareto front exploration is a multi-objective optimization method used to find a set of non-dominated solutions (also known as Pareto optimal solutions) across multiple objectives. The specific steps are as follows:

[0121] ①Generation of initial equipment operating parameter population:

[0122] Using the device operating parameters θ to be optimized as a basis, a set of initial parameter settings is randomly generated. In other words, a set of initial parameters is generated by randomly deforming θ.

[0123] ② Cleaning parameter evaluation: Use the SVR model to predict the cleaning parameters for each parameter setting. This means predicting the cleaning parameters for each parameter set and the current road condition characteristic X.

[0124] ③ Selection: Select non-dominated solutions according to the Pareto principle, that is, select these solutions when no other solution is better than the current solution in all objectives.

[0125] ④ Mutation: Perform mutation operation on the selected non-dominated solution to generate new parameter settings.

[0126] ⑤ Iteration: Repeat the evaluation, selection, and mutation steps until the convergence condition is met or the maximum number of iterations is reached.

[0127] In this way, a set of non-dominated solutions such as θ1, θ2…θ are obtained through Pareto front exploration. n , prepare for the next step of secondary optimization.

[0128] 3.3 Secondary Optimization

[0129] A secondary optimization is performed based on the primary optimization. Specifically, a non-dominated solution is selected from the primary optimization result and a secondary optimization is performed using the Natural Evolution Strategy (NES) to obtain the final optimized equipment operating parameters. The specific steps are as follows:

[0130] ①Select the non-dominated solution on the Pareto front

[0131] According to the dynamically adjusted weight coefficients w1, w2, w3, an initial parameter is selected from the non-dominated solutions on the Pareto frontier. Assuming that w1 is the largest among w1, w2, w3, then in θ1, θ2…θ n Select θ1 which has the best cleaning effect as the initial parameter, which is also the starting point of the secondary optimization. Assuming that w3 is the largest among w1, w2, and w3, which means that the operation time requirement is high, then θ1, θ2…θ n The best performing θ2 in terms of operation time is selected as the initial parameter, which is also the starting point of the secondary optimization.

[0132] ②Natural gradient calculation: Calculate the natural gradient of the objective function J(θ) with respect to the parameter θ At the same time, the adaptive learning rate η is used to update the parameters:

[0133]

[0134] ③Evaluation: Use the SVR model to evaluate the updated parameter settings:

[0135] ④ Iteration: Repeat the calculation of natural gradient and parameter update steps until the convergence condition is met or the maximum number of iterations is reached.

[0136] In the multi-objective optimization of sanitation equipment, the combination of Pareto frontier exploration and natural evolution strategy can efficiently find the optimal combination of equipment operating parameters. Pareto frontier exploration ensures that the optimization process not only considers a single objective but also comprehensively evaluates the performance of multiple objectives by identifying non-dominated solutions that achieve the optimal balance between multiple objectives (such as cleaning effect, energy consumption and operation time). The natural evolution strategy simulates the natural selection mechanism to generate and optimize multiple equipment operating parameter combinations, gradually approaching the global optimal solution. Therefore, through this step, it is possible to accurately predict the actual cleaning effect, and dynamically adjust the equipment operating parameters according to the prediction results to optimize the operation effect.

[0137] 4. Dynamically adjust equipment operating parameters

[0138] According to the optimization results of the previous step, the operating parameters of the equipment, such as brush speed, suction force and water pressure, are dynamically adjusted, and the optimized parameters are sent to the equipment controller to adjust the operating parameters of the equipment.

[0139] The present invention innovatively combines the support vector regression model and the natural evolution strategy, and introduces Pareto frontier exploration under the multi-objective optimization framework.

[0140] 1. Application of Support Vector Regression Models: By collecting historical data, cleaning it, and preprocessing it, we train an SVR model to predict equipment performance indicators under different operating parameters, such as cleaning effectiveness, energy consumption, and operation time. The SVR model selects optimal hyperparameters through cross-validation and grid search methods, ensuring high accuracy and reliability of prediction results.

[0141] 2. Application of Pareto frontier exploration and natural evolution strategy: In the multi-objective optimization framework, Pareto frontier exploration is used to screen out non-dominated solutions that achieve the optimal balance between multiple objectives from the parameter combination, and the weights are dynamically adjusted according to the current road state characteristics. Based on the adjusted weights, one of the non-dominated solutions optimized by Pareto frontier exploration is selected, and then the natural evolution strategy is used to obtain the optimal equipment operating parameters.

[0142] The above embodiments are preferred implementation modes of the present invention, but the implementation modes of the present invention are not limited to the above embodiments. Any other changes, modifications, substitutions, combinations, and simplifications that do not deviate from the spirit and principles of the present invention should be considered as equivalent replacement methods and are included in the scope of protection of the present invention.

Claims

1. A road cleaning control system for sanitation equipment, characterized in that: include: A road condition data collection module, used to collect road condition data, wherein the road condition data includes road humidity, road temperature, road pollutant type and road pollutant distribution; A cleaning parameter prediction module, which predicts cleaning parameters through a trained support vector regression model based on equipment operating parameters and road condition data collected by the road condition data collection module, wherein the cleaning parameters include cleaning effect, energy consumption and operation time; The equipment operation parameter optimization module uses a multi-objective optimization function and combines the cleaning parameters predicted by the cleaning parameter prediction module to optimize the current equipment operation parameters, wherein the multi-objective optimization function includes cleaning effect, energy consumption and operation time as optimization targets, and performs secondary optimization through Pareto frontier exploration and selection of non-dominated solutions, as well as natural evolution strategy, to obtain the final optimized equipment operation parameters; The execution module operates the equipment according to the equipment operation parameters optimized by the equipment operation parameter optimization module.

2. The road cleaning control system for sanitation equipment according to claim 1, characterized in that: The road condition data acquisition module comprises: Humidity sensor, non-contact installed at the front end of sanitation equipment, used to measure road humidity; Temperature sensor, non-contact installed at the front end of sanitation equipment, used to measure road surface temperature; The visual sensor is installed non-contactly above the front end of the sanitation equipment to capture road surface images and identify the type and distribution of pollutants.

3. The road surface cleaning control system for sanitation equipment according to claim 1, characterized in that: The support vector regression model is trained by the following steps: Obtain training data based on historical operation records, including road condition characteristics, equipment operating parameters and corresponding cleaning parameters; Select the support vector regression model of RBF kernel function, and select the best model parameters including regularization parameter, RBF kernel function parameter and insensitive loss function parameter through cross validation and grid search method; The support vector regression model is trained using the training data so that it can predict the cleaning parameters based on the feature vector of the input current road condition data and the equipment operating parameters.

4. The road surface cleaning control system for sanitation equipment according to claim 1, characterized in that: The equipment operation parameter optimization module adopts a method of dynamically adjusting weight coefficients to dynamically adjust the weights of cleaning effect, energy consumption and operation time in a multi-objective optimization function according to the current road condition characteristics, so as to find a balance point between multiple objectives.

5. The road surface cleaning control system for sanitation equipment according to claim 1, characterized in that: The Pareto frontier exploration includes: Initial equipment operating parameter population generation: Taking the equipment operating parameters to be optimized as the basis, a set of initial parameter settings are randomly generated; Cleaning parameter evaluation: Use support vector regression model to predict the cleaning parameters under each set of parameter settings; Select non-dominated solutions: Select non-dominated solutions according to the Pareto principle; Mutation: Perform mutation operations on the selected non-dominated solutions to generate new parameter settings; Iteration: repeat the steps of cleaning parameter evaluation, selecting non-dominated solutions and mutation until the first convergence condition is met or the first maximum number of iterations is reached, and the optimization is completed; That is, a set of non-dominated solutions is obtained through Pareto front exploration.

6. The road surface cleaning control system for sanitation equipment according to claim 5, characterized in that: The natural evolution strategy is used to select a non-dominated solution from the primary optimization result for secondary optimization by calculating the natural gradient of the objective function to the parameter and updating the parameter with an adaptive learning rate until a second convergence condition is met or a second maximum number of iterations is reached.

7. A road cleaning control method for sanitation equipment, characterized in that: The following steps are involved: Collect road condition data, including road moisture, road temperature, road pollutant type, and road pollutant distribution; Based on the road condition data and equipment operating parameters, the cleaning parameters, including cleaning effect, energy consumption and operation time, are predicted by a trained support vector regression model; A multi-objective optimization function is used to optimize the current equipment operating parameters in combination with the predicted cleaning parameters, wherein the multi-objective optimization function includes cleaning effect, energy consumption and operation time as optimization targets, and secondary optimization is performed through Pareto frontier exploration and selection of non-dominated solutions, as well as natural evolution strategy, to obtain the final optimized equipment operating parameters; Operate the equipment according to the optimized equipment operating parameters.

8. The road surface cleaning control method for sanitation equipment according to claim 7, characterized in that: The training steps of the support vector regression model include: Obtain training data based on historical operation records, including road condition characteristics, equipment operating parameters and corresponding cleaning parameters; Select the support vector regression model of RBF kernel function, and select the best model parameters including regularization parameter, RBF kernel function parameter and insensitive loss function parameter through cross validation and grid search method; The support vector regression model is trained using the training data so that it can predict the cleaning parameters based on the feature vector of the input current road condition data and the equipment operating parameters.

9. The road surface cleaning control method for sanitation equipment according to claim 7, characterized in that: In the step of optimizing the equipment operating parameters, a method of dynamically adjusting weight coefficients is adopted to dynamically adjust the weights of cleaning effect, energy consumption and operation time in the multi-objective optimization function according to the current road condition characteristics, so as to find a balance point between multiple objectives.

10. The road surface cleaning control method for sanitation equipment according to claim 7, characterized in that: The steps of Pareto front exploration include: Initial equipment operating parameter population generation: Taking the equipment operating parameters to be optimized as the basis, a set of initial parameter settings are randomly generated; Cleaning parameter evaluation: Use support vector regression model to predict the cleaning parameters under each set of parameter settings; Select non-dominated solutions: Select non-dominated solutions according to the Pareto principle; Mutation: Perform mutation operations on the selected non-dominated solutions to generate new parameter settings; Iteration: repeating the steps of cleaning parameter evaluation, selecting non-dominated solutions and mutation until the first convergence condition is met or the first maximum number of iterations is reached, and the optimization is completed; That is, a set of non-dominated solutions is obtained through Pareto front exploration; The natural evolution strategy is used to select a non-dominated solution from the primary optimization result for secondary optimization by calculating the natural gradient of the objective function to the parameter and updating the parameter with an adaptive learning rate until a second convergence condition is met or a second maximum number of iterations is reached.

Images