Air purification method for household air purification robot

Abstract

This invention provides an air purification method for a home air purifying robot, comprising the following steps: a microprocessor plans the patrol path of the mobile unit; the detection data is transmitted to the microprocessor, and the various detection data are fused using a multimodal data fusion algorithm; combining historical detection data and the fused real-time sensor detection data, a clustering algorithm is used to identify air quality hotspots; using the air quality hotspots, rooms in the home, and obstacles as nodes, a shortest path algorithm is used to plan the purification path of the mobile unit, ensuring it covers the air quality hotspots with the shortest path; the microprocessor controls the mobile unit to move according to the planned purification path; the microprocessor activates the air purification unit in a timely manner based on the real-time sensor detection data. This invention improves purification efficiency and robot utilization, reducing blind movement and repeated purification.

Description

Technical Field

[0001] This invention relates to an air purification method, specifically to an air purification method for a home air purification robot, and belongs to the field of air purification technology. Background Technology

[0002] With increasing air pollution and the spread of infectious diseases, monitoring and purifying indoor air quality in homes has become particularly important. In recent years, with advancements in robotics technology, intelligent and mobile home air purifying robots have become a research hotspot in the field of indoor air purification. However, existing home air purifying robots still have some limitations in home applications. Some home air purifying robots use a fixed-location design, limiting their purification range and only improving air quality in localized areas of the home, making it difficult to meet the purification needs of large areas and multiple rooms. Although some air purifying robots possess autonomous movement capabilities, their movement paths are often preset or highly random, lacking path planning, leading to blind movement and repeated purification, thus reducing purification efficiency. Summary of the Invention

[0003] Based on the above background, the purpose of this invention is to provide an air purification method for a household air purifying robot, thereby solving the problems described in the background art.

[0004] In order to achieve the above-mentioned object of the invention, the present invention provides the following technical solutions:

[0005] An air purification method for a home air purifying robot, the home air purifying robot comprising a robot body and an air quality identification unit, an air purification unit, a moving unit, and a control unit disposed on the robot body; the air quality identification unit includes a multi-parameter air quality sensor, a laser particle counter, and a bioaerosol sensor; the air purification unit includes a pre-filter, an activated carbon filter, a high-efficiency particulate air filter, and an ultraviolet germicidal lamp; the moving unit is used to move the robot body; the control unit includes a microprocessor electrically connected to the air quality identification unit, the air purification unit, and the moving unit; the air purification method includes the following steps:

[0006] The microprocessor plans the patrol path for the mobile unit, controls the mobile unit to move according to the patrol path, and stays at the data acquisition location for a certain period of time.

[0007] At each data acquisition location, the concentration of various air pollutants in different areas of the home is detected by a multi-parameter air quality sensor, the number of particulate matter of different sizes in the air in different areas of the home is measured by a laser particle counter to detect the concentration of particulate matter in the air, and the bioaerosol is detected by a bioaerosol sensor in different areas of the home. The detection data from the multi-parameter air quality sensor, the laser particle counter and the bioaerosol sensor are transmitted to a microprocessor, and the detection data are fused and processed by a multimodal data fusion algorithm.

[0008] By combining historical detection data and fused real-time sensor detection data, a clustering algorithm is used to identify air quality hotspots.

[0009] Using air quality hotspots, rooms in the home, and obstacles as nodes, the shortest path algorithm is used to plan the purification path of the mobile unit, so that it can cover the air quality hotspots with the shortest path. The microprocessor controls the mobile unit to move according to the planned purification path.

[0010] The microprocessor activates the air purification unit in a timely manner based on real-time data detected by the sensors, allowing the air to pass through a pre-filter, activated carbon filter, high-efficiency particulate air filter, and ultraviolet germicidal lamp according to the purification strategy, thereby removing pollutants and pathogenic microorganisms from the air.

[0011] Preferably, the multimodal data fusion algorithm includes the following steps:

[0012] The detection data are standardized, and the mathematical expression for the standardization process is as follows:

[0013]

[0014] In the formula, For the standardized data, x i For the detection data of the i-th sensor, μ i Let σ be the mean value of the i-th sensor. i Let be the standard deviation of the i-th sensor;

[0015] The standardized data from each sensor are then weighted and fused. The mathematical expression for this weighted fusion is as follows:

[0016]

[0017] In the formula, z is the initial fusion feature vector obtained after weighted fusion, and w i This represents the weighting coefficient of the i-th sensor, and its constraint condition is:

[0018] The initial fused feature vectors are input into a deep learning model for further fusion, and the air quality index is output.

[0019] Preferably, the deep learning model is a multilayer perceptron model, and the loss function of the multilayer perceptron model is:

[0020]

[0021] In the formula, θ represents the weight matrix and bias vector of each layer of the multilayer perceptron model, N is the number of training samples, and y i Let be the true air quality index value of the i-th sample. This represents the air quality index value predicted by the multilayer sensor model.

[0022] θ is updated using the backpropagation algorithm and optimizer to minimize the loss function L(θ).

[0023] Preferably, the clustering algorithm is an improved K-means clustering algorithm, which includes the following steps:

[0024] The air quality index value, timestamp, and spatial coordinates are normalized and mapped to the [0,1] interval.

[0025] Calculate the mean and standard deviation of the air quality index values ​​in the historical detection dataset, select samples with air quality index values ​​greater than μ+σ as candidate initial cluster centers, and randomly select multiple samples from the candidate initial cluster centers as initial cluster centers;

[0026] For each sample in the fused real-time sensor detection data, calculate its weighted distance to each cluster center using the following formula: In the formula, d ij x is the weighted distance between the i-th sample and the j-th cluster center. i Let t be the air quality index value of the i-th sample. i Let s be the timestamp of the i-th sample. i Let i be the spatial coordinates of the i-th sample. Let J be the air quality index value of the j-th cluster center. Let j be the timestamp of the j-th cluster center. Let J be the spatial coordinates of the j-th cluster center; assign each sample to the cluster center with the nearest weighted distance to it, forming a new cluster; update the center of each cluster, and the update formula is: In the formula, m j As the cluster center, C j For the j-th cluster;

[0027] Calculate the difference between the current cluster center and the previous cluster center. If the difference is less than the preset threshold or the maximum number of iterations is reached, stop the iteration; otherwise, repeat the previous clustering step.

[0028] For each final cluster, the mean and variance of the air quality index values ​​of its internal samples are calculated as representative indicators of the cluster. Based on the representative indicators and spatial distribution of the clusters, air quality hotspots are identified.

[0029] As a preferred approach, air quality hotspots, rooms within a home, and obstacles are specifically included as nodes:

[0030] The microprocessor stores a floor plan of each room in the house, using the center point of each room as a node and the connecting paths between rooms as edges, to construct a graph that includes a set of nodes and a set of edges.

[0031] The center point of the air quality hotspot area and obstacles in each room of the house are added to the graph as nodes. For nodes representing obstacles, the edge weights between them and their adjacent rooms are set to larger values.

[0032] Preferably, the shortest path algorithm is an improved Dijkstra's algorithm, which includes the following steps:

[0033] Each node in the graph has its corresponding air quality index, and each edge has its corresponding length weight.

[0034] For each node in the graph, initialize the shortest path estimate, which is the shortest path length from the current position of the mobile unit as the starting point to that node; initialize the shortest path estimate of the starting point to zero; initialize the predecessor node of each node to none; initialize the set of unvisited nodes Q to all nodes;

[0035] Select node u with the minimum shortest path estimate from the unvisited node set Q, and remove node u from the unvisited node set Q; for each neighbor node v of node u... j ∈Q, calculate the distance from the starting point through node u to node v. j The path length is calculated using the following formula:

[0036] d new =d[u]+w uj +α·AQI j ;

[0037] In the formula, d[u] is the shortest path estimate for node u, and w uj Let node u and node v be j The length weight of the edge between them, α is the air quality weighting factor, AQI j For node v j The air quality index value; if d new <d[v j If ], then update node vj The shortest path estimate is obtained, and its predecessor node is set as node u;

[0038] Repeat the previous step until the target node, i.e. the center point of the air quality hotspot area, has been visited, or the set of unvisited nodes Q is empty;

[0039] Starting from the target node, backtrack to the starting point based on the predecessor node information, generate the optimal path, and output the optimal path from the starting point to the target node as the purification path.

[0040] Preferably, the robot body is equipped with a lidar sensor, and when the microprocessor controls the moving unit to move according to the planned purification path, the air purification method further includes the following steps:

[0041] During the robot's movement, environmental information of the home room is continuously collected through the LiDAR sensor, and the environmental map is updated in real time using the simultaneous localization and mapping algorithm. When a new obstacle is detected or the position of an existing obstacle changes, path replanning is triggered. Using the updated environmental map and obstacle information, the improved Dijkstra algorithm is re-executed to generate a new purification path.

[0042] Preferably, the purification strategy includes:

[0043] The air quality index is obtained by fusing real-time sensor data.

[0044] If the current air quality index is below the first threshold, the purification unit will be put into primary operation mode, with only the pre-filter working to maintain air cleanliness.

[0045] If the current air quality index is between the first and second thresholds, the purification unit is put into normal operation, with the pre-filter and HEPA filter working simultaneously.

[0046] If the current air quality index is between the second and third thresholds, the purification unit will be put into enhanced operation mode, with the pre-filter, activated carbon filter and HEPA filter working simultaneously.

[0047] If the current air quality index is higher than the third threshold, the purification unit will be in super-powered operation mode, with the pre-filter, activated carbon filter, HEPA filter and UV germicidal lamp working simultaneously.

[0048] Compared with the prior art, the present invention has the following advantages:

[0049] This invention discloses an air purification method for a home air purifying robot. It integrates detection data from various sensors using a multimodal data fusion algorithm, effectively consolidating information from different sensors. A clustering algorithm analyzes historical and real-time detection data to effectively identify hotspots of poor air quality in the indoor environment. Compared to traditional random or full-area purification methods, this method improves purification efficiency and robot utilization. Furthermore, by using air quality hotspots, rooms, and obstacles as nodes, and employing a shortest path algorithm for mobile unit path planning, the robot covers all air quality hotspots with the shortest path, significantly improving purification coverage efficiency and reducing blind movement and repeated purification. Attached Figure Description

[0050] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0051] Figure 1 This is a schematic diagram of the structure of the home air purifying robot in this invention;

[0052] Figure 2 This is a flowchart of the air purification method in this invention;

[0053] Figure 3 This is a flowchart of the multimodal data fusion algorithm in this invention;

[0054] Figure 4 This is a flowchart of the improved K-means clustering algorithm in this invention;

[0055] Figure 5 This is a flowchart of the improved Dijkstra algorithm in this invention. Detailed Implementation

[0056] The technical solution of the present invention will be further described in detail below through specific embodiments and in conjunction with the accompanying drawings. It should be understood that the implementation of the present invention is not limited to the following embodiments, and any modifications and / or alterations made to the present invention will fall within the protection scope of the present invention.

[0057] In this invention, unless otherwise specified, all parts and percentages are by weight, and the equipment and raw materials used are commercially available or commonly used in the art. Unless otherwise specified, the methods in the following embodiments are conventional methods in the art. Unless otherwise specified, the components or equipment in the following embodiments are general standard parts or components known to those skilled in the art, and their structures and principles can be learned by those skilled in the art through technical manuals or conventional experimental methods.

[0058] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings. In this detailed description, numerous specific details are set forth to facilitate explanation and provide a thorough understanding of the embodiments of the present invention. However, one or more embodiments may be practiced by those skilled in the art without these specific details.

[0059] An embodiment of the present invention discloses an air purification method for a home air purifying robot.

[0060] like Figure 1 As shown, the home air purifying robot includes a robot body and an air quality recognition unit, an air purification unit, a movement unit, and a control unit mounted on the robot body. The air quality recognition unit includes a multi-parameter air quality sensor, a laser particle counter, and a bioaerosol sensor. The air purification unit includes a pre-filter, an activated carbon filter, a high-efficiency particulate air filter, and an ultraviolet germicidal lamp. The movement unit is used to move the robot body. The control unit includes a microprocessor electrically connected to the air quality recognition unit, the air purification unit, and the movement unit.

[0061] like Figure 2 As shown, the air purification method includes the following steps:

[0062] S1. The microprocessor plans the patrol path for the mobile unit, controls the mobile unit to move according to the patrol path, and stays at the data collection location for a certain period of time.

[0063] S2. At each data acquisition location, the concentration of various air pollutants in different areas of the home is detected by a multi-parameter air quality sensor, the number of particles of different sizes in the air in different areas of the home is measured by a laser particle counter to detect the concentration of particulate matter in the air, and the bioaerosol is detected by a bioaerosol sensor in different areas of the home. The detection data from the multi-parameter air quality sensor, the laser particle counter and the bioaerosol sensor are transmitted to the microprocessor, and the detection data are fused by a multimodal data fusion algorithm.

[0064] S3. Combining historical detection data and fused real-time sensor detection data, a clustering algorithm is used to identify air quality hotspots.

[0065] S4. Using air quality hotspots, rooms in the home, and obstacles as nodes, the shortest path algorithm is used to plan the purification path of the mobile unit so that it covers the air quality hotspots with the shortest path. The microprocessor controls the mobile unit to move according to the planned purification path.

[0066] S5: The microprocessor activates the air purification unit in a timely manner based on real-time sensor data, allowing air to pass through a pre-filter, activated carbon filter, high-efficiency particulate air filter, and ultraviolet germicidal lamp according to the purification strategy, thereby removing pollutants and pathogenic microorganisms from the air.

[0067] like Figure 3 As shown, in step S2, the multimodal data fusion algorithm includes the following steps:

[0068] S201. Standardize all test data. The mathematical expression for standardization is:

[0069]

[0070] In the formula, For the standardized data, x i For the detection data of the i-th sensor, μ i Let σ be the mean value of the i-th sensor. i Let be the standard deviation of the i-th sensor;

[0071] S202. The standardized data from each sensor are weighted and fused. The mathematical expression for weighted fusion is:

[0072]

[0073] In the formula, z is the initial fusion feature vector obtained after weighted fusion, and w i This represents the weighting coefficient of the i-th sensor, and its constraint condition is:

[0074] S203. Input the initially fused feature vectors into the deep learning model for further fusion and output the air quality index.

[0075] The deep learning model is a multilayer perceptron model, and its forward propagation process is represented as follows:

[0076]

[0077] In the formula, h1, h2, ..., h l This represents the hidden layers of the model, where σ represents the activation function, which is ReLU in this embodiment. W1, W2, ..., W l W out This represents the weight matrix of each layer of the model, b1, b2, ..., bl ,b out This represents the bias vector of each layer of the model. The weight matrix and bias vector are learned parameters that are optimized through model training.

[0078] The loss function for the multilayer perceptron model is:

[0079]

[0080] In the formula, θ represents the weight matrix and bias vector of each layer of the multilayer perceptron model, N is the number of training samples, and y i Let be the true air quality index value of the i-th sample. This represents the air quality index value predicted by the multilayer sensor model.

[0081] θ is updated using the backpropagation algorithm and optimizer to minimize the loss function L(θ).

[0082] The steps described above in the multimodal data fusion algorithm combine weighted averaging and deep learning. First, multi-sensor data is initially fused using weighted averaging. Then, the fused data is input into a deep learning model for feature extraction and further fusion. Finally, the deep learning model outputs the Air Quality Index (AQI), which serves as an important basis for subsequent steps to identify air quality hotspots.

[0083] like Figure 4 As shown, in step S3, the clustering algorithm is an improved K-means clustering algorithm, and its input includes historical air quality datasets. Time weight α, spatial weight β, maximum cluster number K max Its output is the clustering result C = C1, C2, ..., C K Among them, x i t represents the air quality index (AQI) value of the i-th sample. i Let s represent the timestamp of the i-th sample. i Represents the spatial coordinates of the i-th sample.

[0084] The improved K-means clustering algorithm includes the following steps:

[0085] S301, Regarding the air quality index value x i timestamp t i Spatial coordinates s i Perform normalization to map it to the [0,1] interval;

[0086] S302. Calculate the mean and standard deviation of the air quality index values ​​in the historical detection dataset D, select samples with air quality index values ​​greater than μ+σ as candidate initial cluster centers, and randomly select multiple samples from the candidate initial cluster centers as initial cluster centers; where μ is the sample mean of the historical detection dataset D, and σ is the sample standard deviation of the historical detection dataset D.

[0087] S303. For each sample in the fused real-time sensor detection data, calculate its weighted distance to each cluster center. The calculation formula is as follows: In the formula, d ij x is the weighted distance between the i-th sample and the j-th cluster center. i Let t be the air quality index value of the i-th sample. i Let s be the timestamp of the i-th sample. i Let i be the spatial coordinates of the i-th sample. Let J be the air quality index value of the j-th cluster center. Let j be the timestamp of the j-th cluster center. Let J be the spatial coordinates of the j-th cluster center; assign each sample to the cluster center with the nearest weighted distance to it, forming a new cluster; update the center of each cluster, and the update formula is: In the formula, m j As the cluster center, C j For the j-th cluster;

[0088] S304. Calculate the difference between the current cluster center and the previous cluster center. The calculation formula is:

[0089]

[0090] If the difference is less than the preset threshold or the maximum number of iterations is reached, the iteration stops; otherwise, step S303 is repeated.

[0091] S305. For each final cluster, calculate the mean and variance of the air quality index values ​​of its internal samples as representative indicators of the cluster. Based on the representative indicators and spatial distribution of the clusters, identify air quality hotspots.

[0092] Among them, the maximum number of clusters K max The method for determining the cluster quality index (SSE) is as follows: based on the clustering results in step S303, calculate the clustering quality index (SSE) for the current number of clusters k. k The calculation formula is:

[0093]

[0094] Then, calculate the current cluster number k and SSE. k The ratio r k ;

[0095] Next, draw kr k The curve is used to determine the maximum number of clusters K, with the k value corresponding to the inflection point of the curve being taken as the maximum number of clusters. max .

[0096] The improved K-means clustering algorithm combines time weights, spatial weights, and the determination of an adaptive number of clusters. By introducing time and spatial information, the algorithm can better capture the spatiotemporal correlation of air quality data, thereby obtaining better clustering results.

[0097] like Figure 5 As shown, in step S4, the air quality hotspot area, each room in the home, and obstacles are specifically included as nodes:

[0098] S401: The microprocessor stores the floor plan of each room in the house, using the center point of each room as a node and the connecting path between rooms as an edge, to construct a graph including a set of nodes and a set of edges.

[0099] S402. Add the center point of the air quality hotspot area and obstacles in each room of the house as nodes to the graph. For the node representing the obstacle, set the edge weight between it and the adjacent room to a larger value.

[0100] In step S4, the shortest path algorithm is an improved Dijkstra's algorithm, which includes the following steps:

[0101] S403. Make each node in the graph have the air quality index corresponding to that node, and make each edge have a corresponding length weight.

[0102] S404, For each node v in the graph i ∈V, initialize the shortest path estimate d[v i ] = ∞, which is the shortest path length from the current position of the mobile unit as the starting point to this node; initialize the shortest path estimate of the starting point to zero, d[s] = 0; initialize the predecessor node of each node to none, prev[v i ] = None; Initialize the set of unvisited nodes Q to all nodes, Q = V;

[0103] S405. Select node u with the minimum shortest path estimate from the unvisited node set Q, and remove node u from the unvisited node set Q. The selection formula is:

[0104]

[0105] In the formula, α is the air quality weighting factor;

[0106] For each neighboring node v of node uj ∈Q, calculate the distance from the starting point through node u to node v. j The path length is calculated using the following formula:

[0107] d new =d[u]+w uj +α·AQI j ;

[0108] In the formula, d[u] is the shortest path estimate for node u, and w uj Let node u and node v be j The length weight of the edge between them, α is the air quality weighting factor, AQI j For node v j The air quality index value; if d new <d[v j If ], then update node v j The shortest path estimate is obtained, and its predecessor node is set as node u;

[0109] S406. Repeat step S405 until the target node, i.e. the center point of the air quality hotspot area, has been visited, or the set of unvisited nodes Q is empty.

[0110] S407. Starting from the target node, backtrack to the starting point based on the predecessor node information, generate the optimal path, and output the optimal path from the starting point to the target node as the purification path.

[0111] The improved Dijkstra algorithm introduces the Air Quality Index (AQI) as a selection criterion during the node selection phase. It balances the importance of path length and air quality through an air quality weighting factor. A larger α prioritizes nodes with higher air quality priority, while a smaller α prioritizes path length. When calculating path length, the algorithm considers both edge length weights and node AQIs, thus generating an optimal path that comprehensively considers both path length and air quality. Therefore, the optimal path generated by the improved Dijkstra algorithm is not only the shortest path but also takes into account the air quality conditions along the route.

[0112] Furthermore, the robot itself is also equipped with a lidar sensor. When the microprocessor controls the moving unit to move according to the planned purification path, the air purification method also includes the following steps:

[0113] During the robot's movement, environmental information of the home room is continuously collected through the LiDAR sensor, and the environmental map is updated in real time using the simultaneous localization and mapping algorithm. When a new obstacle is detected or the position of an existing obstacle changes, path replanning is triggered. Using the updated environmental map and obstacle information, the improved Dijkstra algorithm is re-executed to generate a new purification path.

[0114] In step S5, the purification strategy includes:

[0115] The air quality index is obtained by fusing real-time sensor data.

[0116] If the current air quality index is below the first threshold, the purification unit will be put into primary operation mode, with only the pre-filter working to maintain air cleanliness.

[0117] If the current air quality index is between the first and second thresholds, the purification unit is put into normal operation, with the pre-filter and HEPA filter working simultaneously.

[0118] If the current air quality index is between the second and third thresholds, the purification unit will be put into enhanced operation mode, with the pre-filter, activated carbon filter and HEPA filter working simultaneously.

[0119] If the current air quality index is higher than the third threshold, the purification unit will be in super-powered operation mode, with the pre-filter, activated carbon filter, HEPA filter and UV germicidal lamp working simultaneously.

[0120] The air purification method in this embodiment fuses detection data from various sensors using a multimodal data fusion algorithm, effectively integrating information from different sensors. It analyzes historical and real-time detection data using a clustering algorithm to effectively identify hotspots of poor air quality in the indoor environment. Compared to traditional random or full-area purification methods, this improves purification efficiency and robot utilization. By using air quality hotspots, rooms, and obstacles as nodes, and employing a shortest path algorithm for mobile unit path planning, the robot can cover all air quality hotspots with the shortest path, significantly improving purification coverage efficiency and reducing blind movement and repeated purification. This air purification method comprehensively removes various pollutants from the air by employing multi-stage air purification units, including a pre-filter, activated carbon filter, high-efficiency particulate air filter, and ultraviolet germicidal lamp.

[0121] This article uses specific examples to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. It should be noted that those skilled in the art can make several improvements and modifications to the present invention without departing from the principles of the present invention, and these improvements and modifications also fall within the protection scope of the claims of the present invention.

Claims

1. An air purification method for a home air purifying robot, characterized in that: The home air purifying robot includes a robot body and an air quality recognition unit, an air purification unit, a moving unit, and a control unit disposed on the robot body. The air quality recognition unit includes a multi-parameter air quality sensor, a laser particle counter, and a bioaerosol sensor. The air purification unit includes a pre-filter, an activated carbon filter, a high-efficiency particulate air filter, and an ultraviolet germicidal lamp. The moving unit is used to move the robot body. The control unit includes a microprocessor electrically connected to the air quality recognition unit, the air purification unit, and the moving unit. The air purification method includes the following steps: The microprocessor plans the patrol path for the mobile unit, controls the mobile unit to move according to the patrol path, and stays at the data acquisition location for a certain period of time. At each data acquisition location, the concentration of various air pollutants in different areas of the home is detected by a multi-parameter air quality sensor, the number of particulate matter of different sizes in the air in different areas of the home is measured by a laser particle counter to detect the concentration of particulate matter in the air, and the bioaerosol is detected by a bioaerosol sensor in different areas of the home. The detection data from the multi-parameter air quality sensor, the laser particle counter and the bioaerosol sensor are transmitted to a microprocessor, and the detection data are fused and processed by a multimodal data fusion algorithm. By combining historical detection data and fused real-time sensor detection data, a clustering algorithm is used to identify air quality hotspots. Using air quality hotspots, rooms in the home, and obstacles as nodes, the shortest path algorithm is used to plan the purification path of the mobile unit, so that it can cover the air quality hotspots with the shortest path. The microprocessor controls the mobile unit to move according to the planned purification path. The microprocessor activates the air purification unit in a timely manner based on real-time data detected by the sensors, so that the air passes through a pre-filter, activated carbon filter, high-efficiency particulate air filter and ultraviolet germicidal lamp according to the purification strategy, thereby removing pollutants and pathogenic microorganisms from the air; The multimodal data fusion algorithm includes the following steps: The detection data are standardized, and the mathematical expression for the standardization process is as follows: ； In the formula, For standardized data, For the first The detection data from each sensor For the first The average value of each sensor, For the first The standard deviation of each sensor; The standardized data from each sensor are then weighted and fused. The mathematical expression for this weighted fusion is as follows: ； In the formula, This is the initial fusion feature vector obtained after weighted fusion. Indicates the first The weighting coefficients of each sensor are subject to the following constraints: ; The initially fused feature vectors are input into a deep learning model for further fusion, and the air quality index is output. The deep learning model is a multilayer perceptron model, and the loss function of the multilayer perceptron model is: ； In the formula, Here are the weight matrices and bias vectors for each layer of the multilayer perceptron model. The number of training samples, For the first The true air quality index value for each sample. This represents the air quality index value predicted by the multilayer sensor model. Updated via backpropagation algorithm and optimizer To minimize the loss function ; The clustering algorithm is an improved K-means clustering algorithm, which includes the following steps: The air quality index value, timestamp, and spatial coordinates are normalized and mapped to... Within the range; Calculate the mean and standard deviation of the air quality index values ​​in the historical monitoring dataset, and select air quality index values ​​greater than 100%. The samples are used as candidate initial cluster centers, and multiple samples are randomly selected from the candidate initial cluster centers as initial cluster centers; For each sample in the fused real-time sensor detection data, calculate its weighted distance to each cluster center using the following formula: In the formula, For the first The sample and the first The weighted distance between the cluster centers For the first The air quality index value of each sample For the first The timestamp of each sample For the first The spatial coordinates of each sample For the first Air quality index values ​​for each cluster center For the first The timestamp of each cluster center For the first The spatial coordinates of each cluster center are calculated; each sample is assigned to the cluster center with the nearest weighted distance, forming a new cluster; the center of each cluster is updated using the following formula: In the formula, As cluster center, For the first One cluster; Calculate the difference between the current cluster center and the previous cluster center. If the difference is less than the preset threshold or the maximum number of iterations is reached, stop the iteration; otherwise, repeat the previous clustering step. For each final cluster, the mean and variance of the air quality index values ​​of its internal samples are calculated as representative indicators of the cluster. Based on the representative indicators and spatial distribution of the clusters, air quality hotspots are identified. Specifically, the nodes include air quality hotspots, rooms in the home, and obstacles: The microprocessor stores a floor plan of each room in the house, using the center point of each room as a node and the connecting paths between rooms as edges, to construct a graph that includes a set of nodes and a set of edges. The center point of the air quality hotspot area and obstacles in each room of the house are added to the graph as nodes. For nodes representing obstacles, the edge weights between them and their adjacent rooms are set to larger values.

2. The air purification method of a household air purifying robot according to claim 1, characterized in that: The shortest path algorithm is an improved Dijkstra's algorithm, which includes the following steps: Each node in the graph has its corresponding air quality index, and each edge has its corresponding length weight. For each node in the graph, initialize the shortest path estimate, which is the shortest path length from the current position of the mobile unit as the starting point to that node; initialize the shortest path estimate of the starting point to zero; initialize each node to have no predecessor nodes; initialize the set of unvisited nodes. For all nodes; Unvisited node set Select the node with the minimum shortest path estimate. , will node Unvisited node set Remove from; for nodes Each neighbor node Calculate the number of nodes from the starting point. Reaching the node The path length is calculated using the following formula: ； In the formula, For nodes Shortest path estimate, For nodes With nodes The length weight of the edges between them As an air quality weighting factor, For nodes The air quality index value; if Then update the node. The shortest path estimate is obtained, and its predecessor node is set as the node. ; Repeat the previous step until the target node, i.e., the center point of the air quality hotspot area, has been visited, or the set of unvisited nodes has been determined. Empty; Starting from the target node, backtrack to the starting point based on the predecessor node information, generate the optimal path, and output the optimal path from the starting point to the target node.

3. The air purification method of a household air purifying robot according to claim 2, characterized in that: The robot body is equipped with a lidar sensor. When the microprocessor controls the moving unit to move according to the planned purification path, the air purification method further includes the following steps: During the robot's movement, environmental information of the home room is continuously collected through the LiDAR sensor, and the environmental map is updated in real time using the simultaneous localization and mapping algorithm. When a new obstacle is detected or the position of an existing obstacle changes, path replanning is triggered. Using the updated environmental map and obstacle information, the improved Dijkstra algorithm is re-executed to generate a new purification path.

4. The air purification method of a household air purifying robot according to claim 1, characterized in that: The purification strategy includes: The air quality index is obtained by fusing real-time sensor data. If the current air quality index is below the first threshold, the purification unit will be put into primary operation mode, with only the pre-filter working to maintain air cleanliness. If the current air quality index is between the first and second thresholds, the purification unit is put into normal operation, with the pre-filter and HEPA filter working simultaneously. If the current air quality index is between the second and third thresholds, the purification unit will be put into enhanced operation mode, with the pre-filter, activated carbon filter and HEPA filter working simultaneously. If the current air quality index is higher than the third threshold, the purification unit will be in super-powered operation mode, with the pre-filter, activated carbon filter, HEPA filter and UV germicidal lamp working simultaneously.

Images