Map partition type determination method and apparatus, and storage medium

Abstract

This invention discloses a method, apparatus, and storage medium for determining map partition types. The method includes: acquiring an environmental map of a target area; determining each map partition of the target area and the connectivity relationships between the map partitions based on the environmental map; determining the predicted probability distribution of the partition type to which each map partition belongs; and determining the partition type of each map partition based on the connectivity relationships and the predicted probability distribution of the partition types to which each map partition belongs. This method significantly improves the accuracy of determining the partition type of each map partition and has strong applicability. Furthermore, this method does not require obtaining the position information of reference objects when determining the map partition type, and can eliminate the need for depth sensors, thus reducing the cost of mobile robots and offering the advantage of low cost.

Description

Technical Field

[0001] This invention relates to the field of semantic map technology, specifically to a method, apparatus, and storage medium for determining map partition types. Background Technology

[0002] Current technologies for determining the type of each region in a map employ object detection to identify reference objects within each zone, and then classify the map zones based on these identified reference objects. However, classifying zones based on these reference objects presents several problems: the relationship between reference objects and zone types requires manually designed mapping rules, which have poor universality in diverse home environments; the location of reference objects is crucial, and identifying these locations requires costly depth sensors; and the performance of the reference object detection algorithm is critical, with false or missed detections significantly impacting room classification performance. Therefore, these technologies cannot meet the current requirements for determining the corresponding zone type for each map zone. Summary of the Invention

[0003] This invention aims to at least partially solve one of the technical problems in related technologies. Therefore, one objective of this invention is to propose a method for determining map partition types, which significantly improves the accuracy of partition type determination for each map partition and has strong applicability.

[0004] The second objective of this invention is to provide a map partition type determination device.

[0005] The third objective of this invention is to provide a storage medium.

[0006] To achieve the above objectives, a first aspect of the present invention provides a method for determining map partition types. The method includes: acquiring an environmental map of a target area; determining each map partition of the target area and the connectivity relationship of each map partition based on the environmental map; determining the predicted probability distribution of the partition type to which each map partition belongs; and determining the partition type of each map partition based on the connectivity relationship of each map partition and the predicted probability distribution of the partition type to which each map partition belongs.

[0007] The map partition type determination method according to embodiments of the present invention acquires an environmental map of the target area, determines each map partition of the target area, the connectivity relationships of each map partition, and the predicted probability distribution of the partition type to which each map partition belongs based on the environmental map, and thus determines the partition type of each map partition. This method significantly improves the accuracy of determining the partition type of each map partition and has strong applicability. Furthermore, this method does not require obtaining the location information of reference objects when determining the map partition type, and does not require the use of depth sensors, reducing the cost of mobile robots and offering the advantage of low cost.

[0008] In addition, the map partition type determination method proposed in the above embodiments of the present invention may also have the following additional technical features:

[0009] According to one embodiment of the present invention, determining the predicted probability distribution of the partition type to which each map partition belongs includes: determining the predicted probability distribution of the partition type to which each map partition belongs based on the area ratio of each map partition to the target area; and / or, collecting object information in each map partition and determining the predicted probability distribution of the partition type to which each map partition belongs based on the object information; and / or, determining the predicted probability distribution of the partition type to which each map partition belongs based on the number of connections between each map partition and other map partitions.

[0010] According to an embodiment of the present invention, determining the map partitions of the target area and the connectivity relationships of the map partitions based on the environmental map includes: collecting spatial information of the target area; determining the map partitions in the target area based on the spatial information and the environmental map of the target area; and determining the connectivity relationships between the map partitions according to the environmental map.

[0011] According to one embodiment of the present invention, determining the partition type of each map partition based on the connectivity of each map partition and the predicted probability distribution of the partition type to which each map partition belongs includes: constructing an undirected graph representing the connectivity between each map partition, using the partition type of each map partition in the environment map as nodes and the connectivity between each map partition as edges; constructing an overall energy function based on each edge, each node and its corresponding predicted probability distribution in the undirected graph, and constructing a Markov random field based on the undirected graph; and determining the partition type of each map partition based on the Markov random field and the overall energy function.

[0012] According to one embodiment of the present invention, the step of constructing an overall energy function based on each edge, each node, and their corresponding predicted probability distribution in the undirected graph includes: determining adjacent nodes in the undirected graph based on the edges in the undirected graph; forming clusters of the adjacent nodes and constructing a first energy function for each cluster based on the partition type corresponding to the nodes in each cluster; constructing a second energy function for each node based on the predicted probability distribution of each node in the undirected graph; and obtaining the overall energy function based on the first energy function and the second energy function.

[0013] According to one embodiment of the present invention, determining the partition type of each map partition based on the Markov random field and the global energy function includes: obtaining a joint probability based on the Markov random field and the global energy function; determining the probability value corresponding to each node when the joint probability is maximized according to the maximum likelihood criterion; and determining the partition type to which each map partition belongs based on the probability value corresponding to the node.

[0014] According to an embodiment of the present invention, the environmental map of the target area includes a first area and a second area. Based on the partition type corresponding to the nodes in each cluster, constructing a first energy function for each cluster includes: if the connected first area and second area have the same partition type, then the first energy function corresponding to the cluster formed by the connected first area and second area is determined as a first preset parameter; if the connected first area and second area have a preset first partition type and a preset second partition type, then the first energy function corresponding to the cluster formed by the connected first area and second area is determined as a second preset parameter; if the connected first area and second area have a preset third partition type and a preset fourth partition type, then the first energy function corresponding to the cluster formed by the connected first area and second area is determined as a third preset parameter; wherein the preset first partition type and the preset second partition type are negatively correlated, the preset third partition type and the preset fourth partition type are positively correlated, the first preset parameter is greater than the second preset parameter, and the second preset parameter is greater than the third preset parameter.

[0015] According to one embodiment of the present invention, the overall energy function is the sum of the preset parameters corresponding to all cliques in the undirected graph and the predicted probabilities corresponding to all nodes.

[0016] To achieve the above objectives, a second aspect of the present invention provides a map partition type determination device, the device comprising: an acquisition module for acquiring an environmental map of a target area; a connectivity determination module for determining each map partition of the target area and the connectivity relationship of each map partition based on the environmental map; a probability distribution determination module for determining a predicted probability distribution of the partition type to which each map partition belongs; and a partition type determination module for determining the partition type of each map partition based on the connectivity relationship of each map partition and the predicted probability distribution of the partition type to which each map partition belongs.

[0017] To achieve the above objectives, a third aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, it implements the map partition type determination method as proposed in the first aspect of the present invention.

[0018] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0019] Figure 1 This is a flowchart of a map partition type determination method according to an embodiment of the present invention;

[0020] Figure 2 This is a flowchart of a map partition type determination method according to a specific embodiment of the present invention;

[0021] Figure 3 This is a schematic diagram of an environmental map according to a specific embodiment of the present invention;

[0022] Figure 4 for Figure 3 A schematic diagram of the undirected graph corresponding to the environment map in the image;

[0023] Figure 5 This is a flowchart of a method for determining map partition types according to another specific embodiment of the present invention;

[0024] Figure 6 This is a schematic diagram of a map partition type determination device according to an embodiment of the present invention. Detailed Implementation

[0025] Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, and should not be construed as limiting the present invention.

[0026] The following is in conjunction with the instruction manual appendix. Figure 1-6 The specific implementation methods, apparatus, and storage media of the present invention will be described in detail below.

[0027] Figure 1 This is a flowchart illustrating a map partition type determination method according to an embodiment of the present invention. Figure 1 As shown, methods for determining map partition types may include:

[0028] S1, Obtain the environmental map of the target area.

[0029] The map partition type determination method provided in this invention can be used for other mobile robots such as robotic vacuum cleaners. These robots acquire an environmental map of a target area, determine the map partition types based on the environmental map, and then perform targeted operations on each map partition according to its type. Specifically, when acquiring the environmental map of the target area, the robotic vacuum cleaner can be controlled to operate within the target area, collect environmental information, and obtain the environmental map. The target area can refer to the entire area or a portion of an area, and this portion may have different types of partitions / rooms. Based on the map partition type determination method provided in this invention, the type of room / room can be determined. Alternatively, the target area can be a single area / room with different sub-areas; in this example, it can be used to experimentally verify the accuracy of the map partition type determination method.

[0030] S2, based on the environment map, determines the map partitions of the target area and the connectivity between the map partitions.

[0031] In one embodiment of the present invention, as shown in Figure 2, determining the map partitions of the target area and the connectivity relationships between the map partitions based on the environmental map may include:

[0032] S21, Collect spatial information of the target area;

[0033] S22, Determine each map partition in the target area based on spatial information and the environmental map of the target area;

[0034] S23, determine the connectivity between map partitions based on the environment map.

[0035] Sensors on a mobile robot can be used to collect spatial information such as the area, width, and height of the target area. Based on the collected spatial information, preset spatial attribute conditions, and the environmental map of the target area, map partitions within the target area are determined. For initial partitions that meet the preset spatial attribute conditions—that is, those that meet the requirements of being a standalone room and also conform to the spatial attributes of a specific area type—then that area can be identified as that specific area. After determining each map partition, the connectivity between them is determined based on the environmental map.

[0036] In embodiments of the present invention, the preset spatial attribute conditions can be flexibly set according to actual conditions, and the area, width and height requirements are generally set to be greater than the corresponding set values.

[0037] In embodiments of the present invention, the connectivity between map partitions is not limited to other connections such as doors and / or the connectivity of isolating map partitions.

[0038] S3, determine the predicted probability distribution of the partition type to which each map partition belongs.

[0039] In one embodiment of the present invention, determining the predicted probability distribution of the partition type to which each map partition belongs may include one or more of the following:

[0040] One method is to determine the predicted probability distribution of the partition type of each map partition based on the area ratio of each map partition to the target area. That is, when determining the predicted probability distribution of the partition type of each map partition, the area of ​​each map partition can be calculated based on the area of ​​each area collected, or the length and width of each area. Based on the area ratio of the map partition to the target area, the predicted probability distribution of the partition type of each map partition is obtained.

[0041] Another method involves collecting object information from each map partition and determining the predicted probability distribution of the partition type to which each map partition belongs based on this object information. Specifically, when determining the predicted probability distribution of the partition type to which each map partition belongs, image acquisition devices on a mobile robot or other sensors can be used to collect object information from each map partition. AI is then used to identify the object information in each map partition to obtain the predicted probability distribution of the partition type to which each map partition belongs. In embodiments of this invention, the object information in each map partition is not limited to furniture such as beds, desks, and sofas, or household appliances such as televisions, refrigerators, and washing machines.

[0042] Another approach involves determining the predicted probability distribution of each map partition's partition type based on the number of connections between each map partition and other map partitions. In other words, when determining the predicted probability distribution of each map partition's partition type, the number of connections between each map partition and other map partitions can be used to determine the predicted probability distribution of each map partition's partition type. For example, in an indoor scene, a living room is generally connected to the balcony, dining room, and various bedrooms. The living room has a relatively large number of connections with other areas. The dining room is generally connected to the kitchen and / or living room. Bedrooms are generally connected to the bathroom, or not connected to other areas. It is evident that different partition types have different numbers of connections with other areas. Therefore, the predicted probability distribution of each map partition's partition type can be determined based on the number of connections between each map partition and other map partitions.

[0043] This invention does not limit the method for obtaining the predicted probability distribution of each map partition's partition type; the predicted probability distribution can be a set of probabilities for each map partition belonging to different partition types. In this embodiment, the predicted probability distribution of each map partition's partition type is as follows:

[0044] P(Xn)=(xn_1,xn_2,…,xn_m,…,xn_M)

[0045] Where Xn represents the partition type of map partition Rn, Rn represents the nth map partition, 1≤n≤N, N represents the total number of map partitions, and xn_m represents the probability that Xn is partition type m, 1≤m≤M, M represents the number of partition types.

[0046] It should be noted that the probability values ​​in the predicted probability distribution P(Xn) of each map partition to which it belongs may include the probability values ​​obtained by any one of the above methods or any two or more of the above methods, and the methods for obtaining probability values ​​are not limited to the above methods.

[0047] As a specific embodiment, when the target area is as follows: Figure 3 When in an indoor area as shown, the mobile robot can be placed in a location such as... Figure 3 The indoor area shown is where the mobile robot operates in the indoor scene, acquiring an environmental map of the indoor area (target area), the map partitions of the target area, the connectivity of each map partition, and the predicted probability distribution of the partition type to which each map partition belongs.

[0048] Each map partition can be represented by Rn, where 1 ≤ n ≤ N, and N represents the total number of map partitions in the target area. Rn represents the nth map partition. The partition type of map partition Rn is represented by the variable Xn, and the probability distribution of the predicted Xn is represented by P(Xn). P(Xn) = (xn_1, xn_2, ..., xn_m, ..., xn_M) is a vector, where xn_m represents the probability that map partition Rn belongs to partition type m, i.e., the probability that variable Xn = m, where 1 ≤ m ≤ M, and M represents the number of partition types. The partition types are living room and non-living room, where non-living room includes, but is not limited to, at least one of the following: bedroom, bathroom, dining room, kitchen, corridor, balcony, and other rooms.

[0049] As an example, when M=6, partition type 1 is the bedroom, partition type 2 is the living room, partition type 3 is the bathroom, partition type 4 is the dining room, partition type 5 is the kitchen, and partition type 6 is other rooms. xn_m represents the probability that map partition Rn belongs to partition type m. For example, x1_4 represents the probability that map partition R1 belongs to the dining room.

[0050] Figure 3The diagram shows the environmental map and map partitioning results for the target area. R1 represents map partition 1, R2 represents map partition 2, R3 represents map partition 3, R4 represents map partition 4, R5 represents map partition 5, R6 represents map partition 6, R7 represents map partition 7, and R8 represents map partition 8. Black lines represent room doors or boundaries connecting map partitions. For example, the predicted probability distributions for the partition types of map partitions R1 and R2 are as follows: map partition R1 has a probability of 0.7 for bedroom, 0.15 for kitchen, 0.1 for toilet, and 0.05 for living room; map partition R2 has a probability of 0.7 for living room, 0.25 for kitchen, and 0.05 for toilet.

[0051] S4. Based on the connectivity of each map partition and the predicted probability distribution of the partition type to which each map partition belongs, determine the partition type of each map partition.

[0052] In some examples, when determining the map partition type, by acquiring the environmental map of the target area, determining the map partitions of the target area, the connectivity of each map partition, and the predicted probability distribution of the partition type of each map partition based on the environmental map, the partition type of each map partition can be determined. This can improve the accuracy of determining the map partition type and reduce the losses caused by inaccurate map partition types, which would otherwise lead to other mobile robots, such as robotic vacuum cleaners, performing targeted operations on each map partition based on the map partition type.

[0053] In one embodiment of the present invention, such as Figure 2 As shown, based on the connectivity of each map partition and the predicted probability distribution of the partition type to which each map partition belongs, the partition type of each map partition is determined, which may include:

[0054] S41. Using the partition type of each map partition in the environment map as nodes and the connectivity between each map partition as edges, construct an undirected graph representing the connectivity between each map partition.

[0055] In some examples, an undirected graph is constructed using the partition type of a map partition as nodes and the connectivity relationships between map partitions, such as doors, as edges. An edge connecting node Xk and node Xn indicates that map partition k and map partition n are connected by a door-like relationship, where k ≠ n, 1 ≤ k ≤ N, and 1 ≤ n ≤ N.

[0056] in, Figure 4 It shows Figure 3 The undirected graph corresponding to the environment map in the middle. For example... Figure 3As shown, map partition R5 is connected to map partitions R1, R3, R4, R7, and R8 through connectivity relationships; map partition R1 is connected to map partition R2 through connectivity relationships; and map partition R7 is connected to map partition R6 through connectivity relationships. Figure 4 In the undirected graph shown, X5 is connected to X1, X3, X4, X7 and X8 by edges, X1 is connected to X2 by edges, and X7 is connected to X6 by edges.

[0057] S42. Based on each edge, each node and its corresponding predicted probability distribution in the undirected graph, construct the overall energy function and construct a Markov random field based on the undirected graph.

[0058] S43, based on Markov random fields and the global energy function, determines the partition type of each map partition.

[0059] It's important to clarify that the term "Markov Random Field" has two meanings. First, the Markov property: This refers to the distribution characteristics of a sequence of random variables arranged chronologically, where the distribution at time N+1 is independent of the values ​​of the random variables before time N. Second, a random field: When each position is randomly assigned a value in phase space according to a certain distribution, the entire set of these random variables is called a random field. A random variable in a Markov random field depends only on its neighboring random variables and is independent of non-neighboring random variables.

[0060] In some examples, to improve the classification of map partitions, a Markov random field is constructed based on the undirected graph, and an overall energy function is constructed based on each edge, node, and its corresponding predicted probability distribution in the undirected graph. The partition type of each map partition is then determined based on the Markov random field and the overall energy function.

[0061] In one embodiment of the present invention, such as Figure 5 As shown, based on the edges, nodes, and their corresponding predicted probability distributions in the undirected graph, the overall energy function can be constructed, which may include:

[0062] S421, determine the adjacent nodes in an undirected graph based on the edges in the undirected graph.

[0063] S422, adjacent nodes are grouped into clusters, and the first energy function of each cluster is constructed based on the partition type corresponding to the nodes in each cluster;

[0064] S423, based on the predicted probability distribution of each node in the undirected graph, construct the second energy function for each node;

[0065] S424, based on the first energy function and the second energy function, the overall energy function is obtained.

[0066] In some examples, to improve the accuracy of the partition types of each determined map partition, the characteristics of Markov random fields are utilized to determine adjacent nodes in the undirected graph based on the edges. As shown in the relationship between the environment map and the corresponding undirected graph, two nodes connected by an edge in the undirected graph are two adjacent map partitions in the environment map. All adjacent nodes in the undirected graph are then obtained and grouped into clusters. For example, Figure 4 The adjacent nodes in the undirected graph are X5 and X1, X5 and X3, X5 and X4, X5 and X7, X5 and X8, X1 and X2, X7 and X6. Forming cliques from X5 and X1, X5 and X3, X5 and X4, X5 and X7, X5 and X8, X1 and X2, and X7 and X6 results in cliques {X5, X1}, {X5, X3}, {X5, X4}, {X5, X7}, {X5, X8}, {X1, X2}, and {X7, X6}.

[0067] Furthermore, in some examples, based on the partition type to which the two nodes in the cluster belong, the first energy function of each cluster is constructed, and the prediction probability corresponding to each node is used to construct the second energy function of each node, thus obtaining the overall energy function.

[0068] In one embodiment of the present invention, each map partition includes a first region and a second region. If the connected first region and second region are of the same partition type, the first energy function corresponding to the cluster formed by the connected first region and second region is determined as a first preset parameter. If the connected first region and second region are of a preset first partition type and a preset second partition type, the first energy function corresponding to the cluster formed by the connected first region and second region is determined as a second preset parameter. If the connected first region and second region are of a preset third partition type and a preset fourth partition type, the first energy function corresponding to the cluster formed by the connected first region and second region is determined as a third preset parameter. The preset first partition type and the preset second partition type are negatively correlated, the preset third partition type and the preset fourth partition type are positively correlated, the first preset parameter is greater than the second preset parameter, and the second preset parameter is greater than the third preset parameter.

[0069] In some examples, considering the typical layout of different partition types in a typical apartment, the probability of two map areas connected by a door having the same partition type is low or impossible. Therefore, if the connected first and second areas have the same partition type, the first energy function corresponding to the cluster formed by the connected first and second areas is determined as the first preset parameter.

[0070] The preset first partition type and the preset second partition type are negatively correlated, meaning that the probability of the preset first partition type and the preset second partition type being connected by a door is low. For example, in a normal setup, related area types such as the dining room and bathroom, or the kitchen and bathroom, would not be connected by a door. Therefore, when the connected first area and the second area are of the preset first partition type and the preset second partition type, the first energy function corresponding to the cluster formed by the connected first area and the second area is determined as the second preset parameter.

[0071] The preset third and fourth partition types are positively correlated, meaning they are more likely to be connected by doors. For example, in a typical setup, related areas such as the living room and non-living room, bedroom and bathroom, dining room and kitchen, dining room and living room, and kitchen and living room are connected by doors. Therefore, if the connected first and second areas are of the preset third and fourth partition types, the first energy function corresponding to the cluster formed by the connected first and second areas is determined as the third preset parameter.

[0072] In this embodiment, the first energy function is negatively correlated with the correlation between two adjacent nodes in the corresponding cluster. Therefore, the first preset parameter > the second preset parameter > the third preset parameter.

[0073] In embodiments of the present invention, the preset first partition type, the preset second partition type, the preset third partition type, and the preset fourth partition type can be set according to specific circumstances.

[0074] In one embodiment of the present invention, the partition type includes living room and non-living room, wherein the non-living room includes at least two of dining room, kitchen, bedroom and bathroom. Based on the partition type corresponding to the nodes in each cluster, the first energy function corresponding to each cluster is:

[0075] For each cluster, when the partition types corresponding to two nodes Xk and Xn in the cluster are the same, the first energy function H1_1(Xk,Xn)=a is obtained;

[0076] When the partition types corresponding to two nodes Xk and Xn in the cluster are restaurant and bathroom respectively, the first energy function of the cluster is H1_2(Xk,Xn)=b; when the partition types corresponding to two nodes Xk and Xn in the cluster are kitchen and bathroom respectively, the first energy function of the cluster is H1_3(Xk,Xn)=c.

[0077] When the partition types corresponding to two nodes Xk and Xn in the cluster are living room and non-living room respectively, the first energy function of the cluster is H1_4(Xk,Xn) = -d, where non-living room includes dining room, kitchen, bedroom or bathroom; when the partition types corresponding to two nodes Xk and Xn in the cluster are bedroom and bathroom respectively, the first energy function of the cluster is H1_5(Xk,Xn) = -e; when the partition types corresponding to two nodes Xk and Xn in the cluster are dining room and kitchen respectively, the first energy function of the cluster is H1_6(Xk,Xn) = -f; when the partition types corresponding to two nodes Xk and Xn in the cluster are dining room and living room respectively, the first energy function of the cluster is H1_7(Xk,Xn) = -g; when the partition types corresponding to two nodes Xk and Xn in the cluster are kitchen and living room respectively, the first energy function of the cluster is H1_8(Xk,Xn) = -h; where a, b, c, d, e, f, g and h are all positive numbers.

[0078] In some examples, for a clique {Xk, Xn}, the first energy function H1(Xk, Xn) has a smaller function value for the preferred combination of partition types and a larger function value for the unpreferred combination of partition types. In other words, for two map partitions Rk and Rn connected by gates, the first energy function H1(Xk, Xn) takes different function values ​​when their partition types Xk and Xn take different combinations of partition types, and the first energy function H1(Xk, Xn) has a smaller function value for the preferred combination of partition types.

[0079] As a specific embodiment, when the partition type includes a living room and a non-living room, and the non-living room includes at least two of the dining room, kitchen, bedroom, and bathroom, the value of the first energy function H1(Xk, Xn) includes, but is not limited to, the following cases. In some examples, m1 can be let represent the bathroom, m2 the dining room, m3 the kitchen, and m4 the living room.

[0080] Considering the conventional layout of various partition types in a typical apartment, it is rare for two partition types to be connected by a door. Therefore, the first energy function H1_1(Xk, Xn) = a is defined, if Xk = Xn, where a is a positive number.

[0081] Considering the typical layout of various zones in a typical apartment, it's rare for a door to connect the dining room and bathroom. Therefore, a first energy function H1_2(Xk, Xn) = b is defined, if Xk ! = Xn ("!=" indicates not equal to) and Xk, Xn in {m1, m2}, where m1 represents the bathroom, m2 represents the dining room, and b is a positive number. Optionally, a > b. Considering the typical layout of various zones in a typical apartment, it's rare for a door to connect the kitchen and bathroom. Therefore, a first energy function H1_3(Xk, Xn) = c is defined, if Xk ! = Xn and Xk, Xn in {m1, m3}, where m1 represents the bathroom, m3 represents the kitchen, and c is a positive number. Optionally, a > c.

[0082] Considering the common layout of various zoning types in typical apartment layouts, many living rooms are connected to non-living rooms. Therefore, the first energy function H1_4(Xk, Xn) = -d, if Xk! = Xn and Xk = m4, where m4 represents the living room and d is a positive number. Considering the common layout of various zoning types in typical apartment layouts, bedrooms are often connected to bathrooms. Therefore, the first energy function H1_5(Xk, Xn) = -e, if Xk! = Xn and Xk, Xn in {m1, m5}, where m1 represents the bathroom, m5 represents the bedroom, and e is a positive number. Optionally, d > e. Considering the common layout of various zoning types in typical apartment layouts, dining rooms are often connected to kitchens. Therefore, the first energy function H1_6(Xk, Xn) = -f, if Xk! = Xn and Xk, Xn in {m1, m5} ... =Xn and Xk, Xn in{m2, m3}, where m2 represents the dining room, m3 represents the kitchen, and f is a positive number. Considering the common layout of various zoning types in general apartment layouts, the dining room and living room are often connected. Therefore, the first energy function H1_7(Xk, Xn) = -g, if Xk! = Xn and Xk, Xn in{m2, m4}, where m2 represents the dining room, m4 represents the living room, and g is a positive number. Considering the common layout of various zoning types in general apartment layouts, the kitchen and living room are often connected. Therefore, the first energy function H1_8(Xk, Xn) = -h, if Xk! = Xn and Xk, Xn in{m3, m4}, where m3 represents the kitchen, m4 represents the living room, and h is a positive number.

[0083] It should be noted that the first energy function can be redefined based on the actual zoning type in the indoor unit layout, and the value of the first energy function can be adjusted according to the actual situation.

[0084] In one embodiment of the present invention, the second energy function is negatively correlated with XK_m, where XK_m represents the probability that the nth map partition Rn belongs to partition type m.

[0085] In some examples, the second energy function H2(Xk) is negatively correlated with XK_m. In other words, the higher the probability (prediction probability) that the predicted map partition Rn belongs to partition type m, the smaller the function value of energy function H2(Xk) when Xk = m.

[0086] As a specific implementation, H2(Xk) = u*xk_m, if Xk = m, where u is a negative coefficient.

[0087] In one embodiment of the present invention, the overall energy function is the sum of the preset parameters corresponding to all cliques in the undirected graph and the predicted probabilities corresponding to all nodes.

[0088] In some examples, the expression for the global energy function H(X) can be:

[0089] H(X)=∑ k,n∈E H1(Xk,Xn)+∑ k H2(Xk)

[0090] Where H(Xk) represents the overall energy function, H1(Xk, Xn) represents the first energy function, H2(Xk) represents the second energy function, the k-th map partition Xk is connected to the n-th map partition Xn, and E represents the edge in the undirected graph.

[0091] In some examples, the overall energy function represents the sum of the first energy functions of each cluster plus the second energy function corresponding to each node. For example... Figure 4 The undirected graph shown has 8 nodes and 7 cliques. The overall energy function H(X) needs to be the sum of the first energy functions of the 7 cliques and the second energy function of the 8 nodes.

[0092] In one embodiment of the present invention, such as Figure 4 As shown, based on Markov random fields and the global energy function, the partition type of each map partition is determined, which may include:

[0093] S431, based on Markov random fields and the global energy function, obtains the joint probability;

[0094] S432, Based on the maximum likelihood criterion, determine the probability values ​​corresponding to each node when the joint probability is maximized;

[0095] S433: Determine the partition type of each map partition based on the probability value corresponding to the node.

[0096] In one embodiment of the present invention, the expression for the joint probability can be:

[0097]

[0098] H(X)=∑ k,n∈E H1(Xk,Xn)+∑ k H2(Xk)

[0099] Where P(X) represents the joint probability, Z′ represents the normalization factor, H(Xk) represents the global energy function, H1(Xk,Xn) represents the first energy function, H2(Xk) represents the second energy function, the k-th map partition Xk is connected to the n-th map partition Xn, and E represents the edge in the undirected graph.

[0100] Based on Markov random field theory, the joint probability of the partition types of all nodes corresponding to the probability values ​​X = (X1, X2, ..., Xk, XN) is: Where Z′ is the normalization factor, ensuring that P(X) satisfies the definition of probability. This embodiment of the invention models the classification problem of indoor scene partitions as a maximum a posteriori probability problem on the Markov random field, that is, finding the value of X when P(X) is maximized. Further, this means finding the value m of the partition type corresponding to each node X when the overall energy function H(Xk) is minimized, thus obtaining the partition type of each map partition.

[0101] According to the maximum likelihood criterion, the optimization objective that maximizes the joint probability is to minimize the overall energy function H(Xk), i.e.

[0102]

[0103] For the optimization problem of minimizing the overall energy function H(Xk), since each node's Xk can take M values ​​{1, 2, 3, 4, ..., M}, and the number of nodes is N, there are M**N (** represents exponentiation) possible value strategies. When the predicted probability distribution for each map partition includes probabilities of various partition types, to reduce the computational load when calculating the overall energy function, we can either use only the prediction probability with the highest probability in the predicted probability distribution for each map partition, or use a prediction probability greater than a preset threshold for calculation.

[0104] The strategies for obtaining the optimal values ​​include, but are not limited to: brute-force search, ICM (iteration condition model) algorithm, simulated annealing algorithm, or graph cuts (image segmentation) algorithm. If N is large, the search space for parameters will be relatively large, and ICM, simulated annealing, or graph cuts algorithms can be used to reduce the algorithm complexity.

[0105] Based on the value selection strategy of X, the partition type of each node can be determined, thus identifying the partition type of each map partition. A semantic map can then be constructed based on the environment map and the partition types of each map partition. During semantic map construction, each map partition is named according to its corresponding partition type.

[0106] The map partition type determination method of this invention acquires an environmental map of the target area, determines each map partition of the target area, the connectivity relationships of each map partition, and the predicted probability distribution of the partition type of each map partition based on the environmental map, and determines the partition type of each map partition by constructing an undirected map of the target area's environmental map, and constructing a Markov random field and overall energy function based on the undirected map. This significantly improves the accuracy of determining the partition type of each map partition and has strong applicability. Furthermore, this invention does not require obtaining the position information of reference objects when determining the map partition type, thus eliminating the need for depth sensors, reducing the cost of mobile robots, and offering the advantage of low cost.

[0107] This invention provides a device for determining map partition types.

[0108] Figure 6 This is a schematic diagram of a map partition type determination device according to an embodiment of the present invention. Figure 6 As shown, the map partition type determination device 100 includes an acquisition module 10, a connectivity determination module 20, a probability distribution determination module 30, and a partition type determination module 40.

[0109] The acquisition module 10 is used to acquire an environmental map of the target area; the connectivity determination module 20 is used to determine each map partition of the target area and the connectivity of each map partition based on the environmental map; the probability distribution determination module 30 is used to determine the predicted probability distribution of the partition type to which each map partition belongs; and the partition type determination module 40 is used to determine the partition type of each map partition based on the connectivity of each map partition and the predicted probability distribution of the partition type to which each map partition belongs.

[0110] It should be noted that other specific embodiments of the map partition type determination device provided in the embodiments of the present invention can be found in other specific embodiments of the map partition type determination method of the above embodiments of the present invention.

[0111] The map partition type determination device of this invention obtains an environmental map of a target area, determines each map partition of the target area, the connectivity relationship of each map partition, and the predicted probability distribution of the partition type of each map partition based on the environmental map, and determines the partition type of each map partition by constructing an undirected map of the environmental map of the target area, and constructing a Markov random field and an overall energy function based on the undirected map. This greatly improves the accuracy of determining the partition type of each map partition and has strong applicability.

[0112] This invention provides a computer-readable storage medium.

[0113] In this embodiment, a computer-readable storage medium stores a computer program thereon, which, when executed by a processor, implements the map partition type determination method as described above.

[0114] The readable storage medium of this invention, utilizing the map partition type determination method described above, greatly improves the accuracy of partition type determination for each map partition and has strong applicability.

[0115] It should be noted that the logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable media include: an electrical connection having one or more wires (electronic device), a portable computer disk drive (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Alternatively, the computer-readable medium may be paper or other suitable media on which the program can be printed, since the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in a computer memory.

[0116] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0117] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0118] In the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," "clockwise," "counterclockwise," "axial," "radial," and "circumferential" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing this invention and simplifying the description, and are not intended to indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0119] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "a plurality of" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0120] In this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," "linking," and "fixing," etc., should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components, unless otherwise explicitly limited. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0121] In this invention, unless otherwise explicitly specified and limited, "above" or "below" the second feature can mean that the first feature is in direct contact with the second feature, or that the first feature is in indirect contact with the second feature through an intermediate medium. Furthermore, "above," "over," and "on top" of the second feature can mean that the first feature is directly above or diagonally above the second feature, or simply that the first feature is at a higher horizontal level than the second feature. "Below," "below," and "under" the second feature can mean that the first feature is directly below or diagonally below the second feature, or simply that the first feature is at a lower horizontal level than the second feature.

[0122] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A method for determining map partition type, characterized in that, The method includes: Obtain an environmental map of the target area; Based on the environmental map, determine the map partitions of the target area and the connectivity relationships of the map partitions; Determine the predicted probability distribution of the partition type to which each of the map partitions belongs; Based on the connectivity of each map partition and the predicted probability distribution of the partition type to which each map partition belongs, the partition type of each map partition is determined. This determination of the partition type of each map partition, based on the connectivity of each map partition and the predicted probability distribution of the partition type to which each map partition belongs, includes: Using the partition type of each map partition in the environment map as nodes and the connectivity between each map partition as edges, an undirected graph representing the connectivity between each map partition is constructed. Based on the edges, nodes, and their corresponding predicted probability distributions in the undirected graph, an overall energy function is constructed, and a Markov random field is constructed based on the undirected graph. The construction of the overall energy function includes: determining adjacent nodes in the undirected graph based on the edges, and grouping these adjacent nodes into clusters; constructing a first energy function for each cluster based on the partition type corresponding to the nodes in each cluster; constructing a second energy function for each node based on the predicted probability distribution of each node in the undirected graph; and obtaining the overall energy function based on the first and second energy functions. The overall energy function is the sum of preset parameters corresponding to all clusters and the predicted probabilities corresponding to all nodes in the undirected graph. Based on the Markov random field and the overall energy function, the partition type of each map partition is determined. This determination includes: obtaining a joint probability based on the Markov random field and the overall energy function; determining the probability value corresponding to each node when the joint probability is maximized according to the maximum likelihood criterion; and determining the partition type to which each map partition belongs based on the probability value corresponding to the node. The determination of the predicted probability distribution of the partition type to which each of the map partitions belongs includes: The predicted probability distribution of the partition type to which each map partition belongs is determined based on the area proportion of each map partition to the target area; and / or Collect object information from each of the map partitions and determine the predicted probability distribution of the partition type to which each of the map partitions belongs based on the object information; and / or, Based on the number of connections between each map partition and other map partitions, the predicted probability distribution of the partition type to which each map partition belongs is determined.

2. The method for determining map partition type according to claim 1, characterized in that, The step of determining the map partitions of the target area and the connectivity relationships of the map partitions based on the environmental map includes: Collect spatial information of the target area; Based on the spatial information and the environmental map of the target area, each map partition in the target area is determined; The connectivity between map partitions is determined based on the environmental map.

3. The method for determining map partition types according to claim 1, characterized in that, The environmental map of the target area includes a first region and a second region. Therefore, constructing the first energy function for each cluster based on the partition type corresponding to the nodes in each cluster includes: If the connected first region and the second region are of the same partition type, then the first energy function corresponding to the cluster formed by the connected first region and the second region is determined as the first preset parameter; If the connected first region and the second region are of a preset first partition type and a preset second partition type, the first energy function corresponding to the cluster formed by the connected first region and the second region is determined as the second preset parameter; If the connected first region and the second region are of the preset third partition type and the preset fourth partition type, the first energy function corresponding to the cluster formed by the connected first region and the second region is determined as the third preset parameter; Wherein, the preset first partition type and the preset second partition type are negatively correlated, the preset third partition type and the preset fourth partition type are positively correlated, the first preset parameter is greater than the second preset parameter, and the second preset parameter is greater than the third preset parameter.

4. A map partition type determination device, characterized in that, The device includes: The acquisition module is used to acquire an environmental map of the target area; The connectivity determination module is used to determine each map partition of the target area and the connectivity relationship of each map partition based on the environment map; A probability distribution determination module is used to determine the predicted probability distribution of the partition type to which each of the map partitions belongs, wherein the probability distribution determination module is used to determine the predicted probability distribution of the partition type to which each of the map partitions belongs based on the area ratio of each of the map partitions to the target area; and / or, to collect object information in each of the map partitions and determine the predicted probability distribution of the partition type to which each of the map partitions belongs based on the object information; and / or, to determine the predicted probability distribution of the partition type to which each of the map partitions belongs based on the number of connections between each of the map partitions and other map partitions; A partition type determination module is used to determine the partition type of each map partition based on the connectivity of each map partition and the predicted probability distribution of the partition type to which each map partition belongs. This module is used to construct an undirected graph representing the connectivity between map partitions, using the partition type of each map partition in the environment map as nodes and the connectivity between map partitions as edges; construct an overall energy function based on each edge, each node, and its corresponding predicted probability distribution in the undirected graph; and construct a Markov random field based on the undirected graph. The construction of the overall energy function based on each edge, each node, and its corresponding predicted probability distribution in the undirected graph includes: determining adjacent nodes in the undirected graph based on the edges in the undirected graph and grouping the adjacent nodes into clusters; and determining the partition type of each node in each cluster based on its corresponding partition type. The process involves: constructing a first energy function for each cluster; constructing a second energy function for each node based on the predicted probability distribution of each node in the undirected graph; obtaining the overall energy function based on the first and second energy functions, where the overall energy function is the sum of preset parameters corresponding to all clusters and predicted probabilities corresponding to all nodes in the undirected graph; determining the partition type of each map partition based on the Markov random field and the overall energy function, wherein determining the partition type of each map partition based on the Markov random field and the overall energy function includes: obtaining a joint probability based on the Markov random field and the overall energy function; determining the probability value corresponding to each node when the joint probability is maximized according to the maximum likelihood criterion; and determining the partition type to which each map partition belongs based on the probability value corresponding to the node.

5. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the map partition type determination method as described in any one of claims 1-3.

Images