WO2023109347A1 - 自移动设备的重定位方法、设备及存储介质 - Google Patents

自移动设备的重定位方法、设备及存储介质 Download PDF

Info

Publication number
WO2023109347A1
WO2023109347A1 PCT/CN2022/129455 CN2022129455W WO2023109347A1 WO 2023109347 A1 WO2023109347 A1 WO 2023109347A1 CN 2022129455 W CN2022129455 W CN 2022129455W WO 2023109347 A1 WO2023109347 A1 WO 2023109347A1
Authority
WO
WIPO (PCT)
Prior art keywords
information
environment information
area
map
area identification
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Ceased
Application number
PCT/CN2022/129455
Other languages
English (en)
French (fr)
Inventor
罗绍涵
孙佳佳
曹蒙
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Dreame Innovation Technology Suzhou Co Ltd
Original Assignee
Dreame Innovation Technology Suzhou Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Dreame Innovation Technology Suzhou Co Ltd filed Critical Dreame Innovation Technology Suzhou Co Ltd
Publication of WO2023109347A1 publication Critical patent/WO2023109347A1/zh
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05D—SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
    • G05D1/00—Control of position, course, altitude or attitude of land, water, air or space vehicles, e.g. using automatic pilots
    • G05D1/02—Control of position or course in two dimensions
    • G05D1/021—Control of position or course in two dimensions specially adapted to land vehicles
    • G05D1/0212—Control of position or course in two dimensions specially adapted to land vehicles with means for defining a desired trajectory
    • G05D1/0214—Control of position or course in two dimensions specially adapted to land vehicles with means for defining a desired trajectory in accordance with safety or protection criteria, e.g. avoiding hazardous areas
    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05D—SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
    • G05D1/00—Control of position, course, altitude or attitude of land, water, air or space vehicles, e.g. using automatic pilots
    • G05D1/02—Control of position or course in two dimensions
    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05D—SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
    • G05D1/00—Control of position, course, altitude or attitude of land, water, air or space vehicles, e.g. using automatic pilots
    • G05D1/02—Control of position or course in two dimensions
    • G05D1/021—Control of position or course in two dimensions specially adapted to land vehicles
    • G05D1/0212—Control of position or course in two dimensions specially adapted to land vehicles with means for defining a desired trajectory
    • G05D1/0221—Control of position or course in two dimensions specially adapted to land vehicles with means for defining a desired trajectory involving a learning process

Definitions

  • the disclosure belongs to the technical field of artificial intelligence, and specifically relates to a relocation method, device and storage medium of a self-mobile device.
  • self-mobile devices can achieve autonomous positioning and navigation with the help of Simultaneous Localization and Mapping (SLAM) technology.
  • SLAM Simultaneous Localization and Mapping
  • it may be hijacked, such as being moved, suspended or dragged in a large area.
  • an uncontrollable drift error will occur in the positioning of the self-mobile device, and repositioning is required.
  • the traditional relocation method includes: searching for the original starting position from the hijacked position of the self-mobile device, thereby completing the relocation of the self-mobile device.
  • the present disclosure provides a relocation method, device and storage medium of a self-mobile device, which can solve the problem of low relocation efficiency of the self-mobile device due to cumbersome relocation methods.
  • the disclosure provides the following technical solutions:
  • a method for relocating a self-mobile device comprising: in response to an instruction for relocating the self-mobile device in a working area, acquiring the current location of the self-mobile device Collected current environmental information;
  • the identifying the current environment information to obtain the area identification information corresponding to the current environment information includes:
  • the area identification model is obtained by training a preset neural network model using training data; the training data includes sample environment information and the The region label corresponding to the sample environment information.
  • the training data further includes the classification label of the first obstacle corresponding to the sample environment information, and the classification label of the first obstacle is used for joint training of the neural network model in combination with the area label , to obtain the area identification model;
  • the inputting the current environment information into the pre-trained area identification model to obtain the area identification information includes: inputting the current environment information into the area identification model to obtain The classification result of the first obstacle corresponding to the current environment information and the area identification information;
  • the training data also includes the first characteristic information of the first obstacle; correspondingly, the inputting the current environment information into the pre-trained area recognition model to obtain the area identification information includes: inputting the current Inputting the environment information and the first feature information into the area identification model to obtain the area identification model;
  • the first obstacle is used to indicate area identification information.
  • the identifying the current environment information to obtain the area identification information corresponding to the current environment information further includes:
  • the current environment information includes information matching the second feature information, determine that the area identification information corresponding to the current environment information is the area identification information indicated by the first obstacle.
  • the second characteristic information includes contour information of the first obstacle; the second characteristic information further includes size information and/or distance information of the first obstacle.
  • the area identification information is the position coordinates of the first obstacle in the area map; determining the local area map indicated by the area identification information in the area map of the working area includes:
  • a local area map of a preset shape and a preset size is determined based on the area identification information
  • the local region map to which the region identification information belongs is determined, and the region map is pre-divided into a plurality of partial region maps.
  • determining the partial area map indicated by the area identification information in the area map of the working area includes:
  • matching the current environment information with the template environment information to determine the location of the mobile device includes:
  • the relocation neural network inputting the current environment information and the template environment information into a pre-trained relocation neural network to obtain the position; the relocation neural network is used to determine whether the current environment information matches the template environment information, and The location corresponding to the matched template environment information is determined as the location.
  • the template environment information includes characteristic information of the second obstacle.
  • the second aspect provides an electronic device, the device includes a processor and a memory; a program is stored in the memory, and the program is loaded and executed by the processor to realize the mobile device described in the first aspect The relocation method.
  • a third aspect provides a computer-readable storage medium, where a program is stored in the storage medium, and when the program is executed by a processor, the program is used to implement the self-mobile device relocation method provided in the first aspect.
  • the beneficial effect of the present disclosure is that: by responding to the instruction of relocating the self-mobile device in the working area, the current environment information collected from the mobile device based on the current location is acquired, the current environment information is identified, and the current environment information is obtained Corresponding area identification information, determine the local area map indicated by the area identification information in the area map of the working area, obtain the template environment information corresponding to at least one map position in the local area map, match the current environment information with the template environment information, and Determine the location from the mobile device in the local area map.
  • the problem of low relocation efficiency caused by cumbersome relocation methods of the self-mobile device can be solved.
  • the self-mobile device By identifying the area identification information when the mobile device needs to be relocated, and using the local area map indicated by the area identification information to relocate the current location of the mobile device.
  • the self-mobile device does not need to search for a certain position in the entire working area, but can realize relocation by using a local area map, which can improve the efficiency of relocation.
  • the self-mobile device does not need to move to the original starting position, but only needs to match the current environment information of the current location with the template environment information of the local area map, which can save resources of the self-mobile device and further improve relocation. s efficiency.
  • the neural network model is jointly trained by using the classification label and the area label of the first obstacle to obtain an area recognition model, and the trained area identification model will be more accurate, which can improve the accuracy of area identification information identification.
  • the area identification model can compare the current environment information with the first feature information to determine whether there is a first obstacle. Since the first obstacle can indicate the area identification information, it can Reduce the computational difficulty of the network model and save computing resources from mobile devices.
  • FIG. 1 is a schematic structural diagram of a mobile device provided by an embodiment of the present disclosure
  • FIG. 2 is a flowchart of a relocation method from a mobile device provided by an embodiment of the present disclosure
  • Fig. 3 is a block diagram of an apparatus for relocating self-mobile equipment provided by an embodiment of the present disclosure
  • Fig. 4 is a block diagram of an electronic device provided by an embodiment of the present disclosure.
  • orientation words such as “up, down, top, bottom” usually refers to the direction shown in the drawings, or refers to the vertical, vertical Or in the direction of gravity; similarly, for the convenience of understanding and description, “inner and outer” refer to the inner and outer relative to the outline of each component itself, but the above-mentioned orientation words are not used to limit the present disclosure.
  • FIG. 1 it is a schematic structural diagram of a self-moving device provided by an embodiment of the present disclosure.
  • the self-moving device can be a self-moving device such as a sweeping robot and a floor washing robot. limited. It can be seen from FIG. 1 that the mobile device at least includes: a driving component 110 , a moving component 120 , a controller 130 and a first sensor 140 .
  • the driving component 110 is connected with the moving component 120 and is used to drive the moving component 120 to run, so as to drive the mobile device to move.
  • the driving component 110 is connected with the controller 130 and is used to drive the moving component 120 to run in response to an instruction issued by the controller 130 .
  • the driving assembly 110 may be implemented as a DC motor, a servo motor, a stepping motor, etc., and this embodiment does not limit the implementation of the driving assembly 110 .
  • the first sensor 140 is used to collect current environment information.
  • the first sensor 140 may be a camera equipped with a color system (Red Green Blue, RGB) detection function, an infrared sensor, or a laser radar sensor, etc.
  • the type of the first sensor 140 is not limited in this embodiment.
  • the first sensor 140 may be installed on the casing of the self-mobile device, and used to collect the environment where the self-mobile device is located.
  • the collection range of the first sensor 140 includes, but is not limited to: the area directly in front of, obliquely above, and/or obliquely below the direction of travel of the self-mobile device; and/or the area on the left side of the direction of travel of the self-mobile device; The area on the right side of the traveling direction of the device; and/or the rear side area of the traveling direction of the self-mobile device, etc., this embodiment does not limit the collection range of the first sensor 140 .
  • the number of the first sensor 140 can be one or at least two. In the case where the number of the first sensor 140 is at least two, the types of different first sensors 140 are the same or different. This embodiment does not apply to the first sensor 140. The number and implementation methods are limited.
  • the first sensor 140 is connected with the controller 130 to send the collected current environment information to the controller 130 .
  • the controller 130 is used to relocate the mobile device.
  • the controller 130 may be implemented as a single-chip microcomputer or a processor, and this embodiment does not limit the implementation manner of the controller 130 .
  • the controller 130 is configured to: respond to an instruction to relocate the mobile device in the working area, obtain the current environment information collected from the mobile device based on the current location; identify the current environment information, and obtain The regional identification information corresponding to the current environmental information; determine the local area map indicated by the regional identification information in the regional map of the working area; obtain the template environmental information corresponding to at least one map position in the local area map; compare the current environmental information with the template environmental information Match to determine the location from the mobile device in the local area map.
  • the instruction for relocating the mobile device within the working area is generated by the mobile device based on the sensing data of the second sensor.
  • the mobile device is also provided with a second sensor 150 , which is connected to the controller 130 and used to send sensing data to the controller 130 .
  • the controller 130 determines whether the self-mobile device is hijacked based on the sensing data; Positioning instructions; if it is determined that the self-mobile device has not been hijacked, the step of whether the self-mobile device is hijacked is performed again until the self-mobile device finishes working.
  • being hijacked refers to an abnormal movement of the mobile device
  • the abnormal movement refers to a movement that does not occur independently of the mobile device, and the abnormal movement cannot be sensed by the mobile device. Therefore, when the self-mobile device is hijacked, it cannot locate or accurately locate its own position. For example, when the self-mobile device is moved, suspended in the air during the movement, or dragged in a large area, etc., it is the case that the self-mobile device is hijacked.
  • disengagement hijacking refers to the end of abnormal movement of the self-mobile device, such as: returning to the ground after being moved, or when the self-mobile device stops being dragged after being dragged, it is self-moving The device is out of the hijacked situation.
  • the sensing data includes, but is not limited to: height data, displacement data, and/or angle data, etc. from the mobile device.
  • the second sensor 150 includes, but is not limited to, a gyroscope, a displacement sensor, or an image collector etc., this embodiment does not limit the implementation manner of the second sensor 150 .
  • determining whether the mobile device is hijacked based on the sensing data includes: comparing the change of the sensing data with the template change in the hijacked state; If the changes match, it is determined that the self-mobile device is hijacked; if the change of the sensing data does not match the template change in the hijacked state, it is determined that the self-mobile device is not hijacked.
  • determining whether the self-mobile device is hijacked based on the sensing data includes: determining whether the variation value of the sensing data is within the variation range of the hijacked state; if so, determining that the self-mobile device is hijacked; The mobile device was not hijacked.
  • the sensing data includes but not limited to: contact data, etc.
  • the second sensor 150 includes but not limited to a pressure sensor, or a contact sensor, etc., and this embodiment does not limit the implementation of the second sensor 150 .
  • determining whether the mobile device is hijacked based on the sensing data includes: determining whether the sensing data indicates that there is an object approaching the mobile device; if so, determining that the mobile device is hijacked; if not, determining that the mobile device is not hijacked hijack.
  • the command to relocate the mobile device within the working area is sent from a control device communicatively connected to the mobile device.
  • the control device may be a mobile phone, a remote controller, or a wearable device, etc., and this embodiment does not limit the type of the control device.
  • the instruction to relocate the mobile device within the working area is generated when the mobile device receives a trigger operation acting on the repositioning control.
  • a relocation control is also provided on the self-mobile device, and the relocation control may be a physical button or a virtual control displayed through a touch screen. This embodiment does not limit the implementation of the relocation control.
  • the above manner of obtaining the relocation instruction is only illustrative. In actual implementation, the manner of obtaining the relocation instruction from the mobile device may also be other manners, which will not be listed here in this embodiment.
  • the self-moving device may also include other components, such as a power supply component, a shock absorption component, etc., which will not be listed here in this embodiment.
  • the original starting position is searched from the hijacked position of the mobile device.
  • the search method is usually: the self-mobile device moves randomly in the working area, and receives the infrared signal emitted by the original starting position (such as the charging stand) through the infrared receiving device. When the infrared signal is received by the infrared receiving device, move to the position of the infrared signal to move to the original starting position, and read the map position of the original starting position stored in the area map of the work area to complete the reset position.
  • the traditional relocation method cannot relocate the self-mobile device when it is free from hijacking. It needs to search the original starting position in the entire working area, and the efficiency of relocation is low.
  • the current location of the mobile device is relocated by identifying the area identification information when the mobile device needs to be relocated, and using the local area map indicated by the area identification information.
  • the self-mobile device does not need to search for a certain position in the entire working area, but can realize relocation by using a local area map, which can improve the efficiency of relocation.
  • the self-mobile device does not need to move to the original starting position, but only needs to match the current environment information of the current location with the template environment information of the local area map, which can save resources of the self-mobile device and further improve relocation. s efficiency.
  • FIG. 2 A method for relocating a self-mobile device provided in this embodiment is shown in FIG. 2 .
  • the method is used in the controller 130 shown in FIG. 1 as an example for illustration.
  • the method includes at least the following steps:
  • Step 201 in response to an instruction to relocate the mobile device within the working area, acquire current environment information collected based on the current location of the mobile device.
  • the instruction for relocating the self-mobile device is generated by the self-mobile device based on the sensing data of the second sensor; or, it is sent by a control device connected to the self-mobile device in communication; It is generated when a trigger operation is performed on the positioning control, and this embodiment does not limit the acquisition method of the instruction for relocating from the mobile device.
  • the current environment information is obtained from the current environment where the mobile device is located.
  • the current environment information may be image data and/or point cloud data, and the current environment information may be three-dimensional data or two-dimensional data, and this embodiment does not limit the implementation manner of the current environment information.
  • the current environment information may be collected by the controller controlling the first sensor when the mobile device obtains a relocation instruction; or, it is collected continuously after the first sensor is powered on.
  • the timing of collection is limited.
  • Step 202 identifying the current environment information to obtain area identification information corresponding to the current environment information.
  • the area identification information is used to uniquely indicate a certain local area in the working area.
  • the area identification information is the position coordinates of the first obstacle in the area map.
  • the first obstacle is used to indicate area identification information.
  • the first obstacle refers to an obstacle that can indicate the attribute of the local area.
  • the first obstacle is a dining table, and the attribute of the local area indicated by the dining table is a restaurant; another example: the first obstacle is a toilet, and the attribute of the local area indicated by the toilet is a bathroom; another example: the first obstacle is a bed, the bed The attribute of the indicated local area is a bedroom, and this embodiment does not list all implementations of the first obstacle here.
  • the area identification information is an area identification of the local area, and the area identification may be an attribute or a label of the local area.
  • the area identification information is a restaurant, a bathroom, and/or a bedroom.
  • the label 1, 2, 3, etc. are pre-set for each local area map in the area identification information area map.
  • the attributes of the local area include restaurants, bathrooms, and bedrooms as examples for illustration.
  • the attribute division method of the local area can also be in other ways, for example: divide the attributes of the local area into: office area , coffee break area, etc., this embodiment does not limit the attribute division method of the local area.
  • the way to identify the current environmental information and obtain the area identification information corresponding to the current environmental information includes but is not limited to at least one of the following:
  • the first identification method is to input the current environment data into the pre-trained area identification model to obtain area identification information.
  • the region recognition model is obtained by using training data to train a preset neural network model.
  • the realization of training data includes but not limited to the following situations:
  • the training data only includes the sample environment information and the region labels corresponding to the sample environment information.
  • the training process of the area recognition model includes: inputting the sample environment information into the preset first neural network model to obtain the first training result; inputting the first training result and the area label corresponding to the sample environment information into the first loss function, Obtaining a first loss result; training the first neural network model based on the first loss result to reduce the difference between the first training result and the corresponding region label until the neural network model converges to obtain a region recognition model.
  • the area identification information is the position coordinates of the first obstacle; when the area label is the area attribute of the first obstacle, the area identification information is the The area attribute of an obstacle.
  • Inputting the current environment information into the pre-trained area recognition model to obtain area identification information includes: inputting the current environment information into the pre-trained area identification model to obtain the area identification information corresponding to the current environment information.
  • the first neural network model can be a convolutional neural network (Convolutional Neural Networks, CNN), a recursive neural network (Recursive Neural Network, RNN), a feedforward neural network (Feedforward Neural Network, FNN), and this embodiment does not apply to the first
  • the implementation of the neural network model is limited.
  • the training data not only includes the sample environment information and the area label corresponding to the sample environment information, but also includes the classification label of the first obstacle corresponding to the sample information, wherein the classification label of the first obstacle is used to combine the area label pair
  • the neural network model is jointly trained to obtain a region recognition model.
  • the classification label of the first obstacle is the attribute label of the first obstacle.
  • the training process of the area recognition model includes: inputting the sample environment information into the preset second neural network model to obtain the second training result, the second training result includes the area prediction result and the classification prediction result; the second training result , the region label and the classification label are input into the second loss function to obtain the second loss result; the second neural network model is trained based on the second loss result to reduce the difference between the second training result and the corresponding region label and classification label value until the second neural network model converges to obtain the region recognition model.
  • the second neural network model includes two network branches, one of which is used to calculate the region prediction result, and the other branch is used to calculate the classification result.
  • the trained area identification model will be more accurate, which can improve the accuracy of area identification information identification .
  • inputting the current environment information into the pre-trained area recognition model to obtain area identification information includes: inputting the current environment information into the area identification model to obtain the classification result and area identification information of the first obstacle corresponding to the current environment information.
  • the training data not only includes the sample environment information and the area label corresponding to the sample environment information, but also includes the first characteristic information of the first obstacle corresponding to the sample environment information.
  • the first feature information may be contour information of the first obstacle, or a feature vector of the first obstacle, and this embodiment does not limit the implementation manner of the first feature information.
  • the training process of the region recognition model includes: inputting the sample environment information and the first feature information into the preset third neural network model to obtain the third training result; inputting the third training result and the region label into the third loss function , to obtain the third loss result; based on the third loss result, the third neural network model is trained to reduce the difference between the third training result and the corresponding region label, until the third neural network model converges, and the region recognition model is obtained .
  • the third neural network model is used to compare the sample environment information with the first feature information to determine whether the sample environment information matches the first feature information, thereby determining whether there is a first obstacle in the sample environment information, due to the first Obstacles can indicate area identification information, therefore, the calculation difficulty of the network model can be reduced, and the calculation resources of mobile devices can be saved.
  • inputting the current environment information into the pre-trained area recognition model to obtain the area identification information includes: inputting the current environment information and the first feature information of the first obstacle into the area identification model to obtain the area identification information.
  • the first characteristic information of the first obstacle is pre-stored in the self-mobile device.
  • the second identification method is to obtain the second feature information of the first obstacle; match the acquired current environment information with the second feature information; if the current environment information includes information matching the second feature information, Then it is determined that the area identification information corresponding to the current environment information is the area identification information indicated by the first obstacle.
  • the second feature information is the same as or different from the first feature information.
  • the second feature information may be the outline information of the first obstacle, the size information and/or distance information of the first obstacle, or the information that can be used to describe the characteristics of the first obstacle. This embodiment does not make any contribution to the second feature information. limited.
  • the first obstacle is a dining table
  • the second feature information is the shape and size of the dining table. If the current environment information includes information matching the shape and size of the dining table, it is determined that the area identification information corresponding to the current environment information is the area identification information indicated by the dining table, such as a restaurant.
  • Step 203 determine the partial area map indicated by the area identification information in the area map of the working area.
  • the local area map may change based on changes in area identification information.
  • the area identification information is the position coordinates of the first obstacle in the area map.
  • the manner of determining the partial area map indicated by the area identification information in the area of the working area includes but not limited to at least one of the following:
  • the first type In the area map, a local area map with a preset shape and a preset size is determined based on the area identification information.
  • the preset shape and preset size are pre-stored in the mobile device.
  • the preset shape may be a circle, a rectangle, or an irregular shape, and this embodiment does not limit the implementation of the preset shape.
  • determining a local area map with a preset shape and a preset size based on the area identification information includes: taking the position coordinates of the first obstacle as the centroid of the local area map to generate a local area with a preset shape and a preset size map.
  • the first obstacle is the charging stand
  • the area identification information is the location coordinates of the charging stand.
  • the default shape is a circle
  • the default size is a radius of 2 meters.
  • the local area map is a circular area on the area map with the location coordinates of the charging stand as the center and a radius of 2 meters.
  • the position coordinates of the first obstacle may also be located at the edge of the local area map or other positions. This embodiment does not limit the method of determining the local area map with a preset shape and a preset size based on the area identification information. .
  • the second type in the area map that has been segmented, determine the local area map to which the area identification information belongs, and the area map is pre-divided into a plurality of local area maps.
  • the way of dividing the regional map into multiple local regional maps includes but not limited to: dividing the regional map according to attributes, or dividing the regional map according to the preset segmentation size.
  • the division method of the regional map can also be Other manners are not listed here in this embodiment.
  • the area map For example: Divide the area map into a bedroom area map and a bathroom area map.
  • the area identification information is the location coordinates of the charging stand, the charging stand is located in the bedroom area, and the local area map to which the area identification information belongs is the bedroom area map.
  • determining the local area map indicated by the area identification information in the area map of the working area includes: determining the local area map corresponding to the area identification information based on the correspondence between the area identification information and the local area map.
  • the area identification information is the label of each local area map.
  • the local area map corresponding to the label is found from the correspondence.
  • Step 204 acquiring template environment information corresponding to at least one map position in the local area map.
  • the local area map includes template environment information corresponding to multiple map locations, and the multiple template environment information all have location coordinates corresponding to the map locations.
  • the controller obtains the template environment information corresponding to at least one map position.
  • Step 205 matching the current environment information with the template environment information to determine the location of the ego mobile device in the local area map.
  • the template environment information includes characteristic information of the second obstacle.
  • the manner of matching the current environment information with the template environment information includes but not limited to at least one of the following:
  • the first is to input the current environment information and the template environment information into the pre-trained relocation neural network to obtain the position of the self-mobile device in the local area map.
  • the relocation neural network is used to determine whether the current environment information matches the template environment information, and determine the position corresponding to the matched template environment information as the position of the self-mobile device in the local area map.
  • the template environment information includes characteristic information of the second obstacle.
  • the second obstacle can be any obstacle in the working area, such as: tables, chairs, carpets, walls, etc., and this embodiment does not limit the type of the second obstacle.
  • the feature information of the second obstacle may be the shape, size, or feature vector of the second obstacle, and this embodiment does not limit the implementation manner of the feature information of the second obstacle.
  • the training process of the relocation neural network includes: inputting the sample environment information and each template environment information into the preset neural network model to obtain the similarity result; comparing the similarity result with the real similarity result, and based on the comparison result Train the neural network model to obtain a relocation neural network.
  • Match the current environment information with the template environment information to determine the position of the self-mobile device in the local area map including: input the current environment information and the template environment information into the pre-trained relocation neural network to obtain multiple similarity results,
  • the map position corresponding to the environmental template information ranked first in the similarity result is determined as the position of the self-mobile device in the local area map.
  • the second method is to calculate the similarity between the current environment information and each template environment information; the map position corresponding to the environment template information with the greatest similarity is determined as the position of the self-mobile device in the local area map.
  • the method for relocating the self-mobile device obtains the current environment information collected by the self-mobile device based on the current location by responding to an instruction to relocate the self-mobile device in the work area, Identify the current environmental information, obtain the area identification information corresponding to the current environmental information, determine the local area map indicated by the area identification information in the area map of the working area, obtain the template environment information corresponding to at least one map position in the local area map, and set The current environment information is matched with the template environment information to determine the location of the mobile device in the local area map.
  • the problem of low relocation efficiency caused by cumbersome relocation methods of the self-mobile device can be solved.
  • the self-mobile device By identifying the area identification information when the mobile device needs to be relocated, and using the local area map indicated by the area identification information to relocate the current location of the mobile device.
  • the self-mobile device does not need to search for a certain position in the entire working area, but can realize relocation by using a local area map, which can improve the efficiency of relocation.
  • the self-mobile device does not need to move to the original starting position, but only needs to match the current environment information of the current location with the template environment information of the local area map, which can save resources of the self-mobile device and further improve relocation. s efficiency.
  • the neural network model is jointly trained by using the classification label and the area label of the first obstacle to obtain an area recognition model, and the trained area identification model will be more accurate, which can improve the accuracy of area identification information identification.
  • the area identification model can compare the current environment information with the first feature information to determine whether there is a first obstacle. Since the first obstacle can indicate the area identification information, it can Reduce the computational difficulty of the network model and save computing resources from mobile devices.
  • Fig. 3 is a block diagram of an apparatus for relocating an autonomous mobile device provided by an embodiment of the present disclosure. This embodiment is described by taking the application of the apparatus in the autonomous mobile device shown in Fig. 1 as an example.
  • the device at least includes the following modules: a first acquisition module 310 , an information identification module 320 , a map determination module 330 , a second acquisition module 340 and a relocation module 350 .
  • the first acquiring module 310 is configured to acquire current environment information collected from the mobile device based on the current location in response to an instruction to relocate the mobile device within the working area.
  • the information identification module 320 is configured to identify the current environment information, and obtain area identification information corresponding to the current environment information.
  • the map determination module 330 is configured to determine a partial area map indicated by the area identification information in the area map of the working area.
  • the second acquiring module 340 is configured to acquire template environment information corresponding to at least one map position in the local area map.
  • the relocation module 350 is configured to match the current environment information with the template environment information, so as to determine the location of the ego mobile device in the local area map.
  • the self-mobile device relocation device when the self-mobile device relocation device provided in the above-mentioned embodiments performs relocation, it only uses the division of the above-mentioned functional modules as an example. In practical applications, the above-mentioned functions can be assigned by different The completion of the functional modules means that the internal structure of the self-mobile device relocation device is divided into different functional modules, so as to complete all or part of the functions described above.
  • the apparatus for relocating self-mobile equipment provided by the above-mentioned embodiments and the embodiment of the method for relocating self-mobile equipment belong to the same concept, and its specific implementation process is detailed in the method embodiment, and will not be repeated here.
  • the electronic device may be the self-moving device in FIG. 1 .
  • the electronic device includes at least a processor 401 and a memory 402 .
  • the processor 401 may include one or more processing cores, such as a 4-core processor, an 8-core processor, and the like.
  • Processor 401 can adopt at least one hardware form in DSP (Digital Signal Processing, digital signal processing), FPGA (Field-Programmable Gate Array, field programmable gate array), PLA (Programmable Logic Array, programmable logic array) accomplish.
  • Processor 401 may also include a main processor and a coprocessor, and the main processor is a processor for processing data in a wake-up state, also known as a CPU (Central Processing Unit, central processing unit); the coprocessor is Low-power processor for processing data in standby state.
  • CPU Central Processing Unit, central processing unit
  • the coprocessor is Low-power processor for processing data in standby state.
  • the processor 401 may be integrated with a GPU (Graphics Processing Unit, image processor), and the GPU is used for rendering and drawing the content that needs to be displayed on the display screen.
  • the processor 401 may also include an AI (Artificial Intelligence, artificial intelligence) processor, where the AI processor is used to process computing operations related to machine learning.
  • AI Artificial Intelligence, artificial intelligence
  • Memory 402 may include one or more computer-readable storage media, which may be non-transitory.
  • the memory 402 may also include high-speed random access memory and non-volatile memory, such as one or more magnetic disk storage devices and flash memory storage devices.
  • the non-transitory computer-readable storage medium in the memory 402 is used to store at least one instruction, and the at least one instruction is used to be executed by the processor 401 to realize the motor brake provided by the method embodiment of the present disclosure. method.
  • the electronic device may optionally further include: a peripheral device interface and at least one peripheral device.
  • the processor 401, the memory 402, and the peripheral device interface may be connected through a bus or a signal line.
  • Each peripheral device can be connected with the peripheral device interface through a bus, a signal line or a circuit board.
  • peripheral devices include but are not limited to: radio frequency circuits, touch screens, audio circuits, and power supplies.
  • the electronic device may also include fewer or more components, which is not limited in this embodiment.
  • the present disclosure further provides a computer-readable storage medium, in which a program is stored, and the program is loaded and executed by a processor to implement the method for relocating from a mobile device in the foregoing method embodiments.

Landscapes

  • Engineering & Computer Science (AREA)
  • Aviation & Aerospace Engineering (AREA)
  • Radar, Positioning & Navigation (AREA)
  • Remote Sensing (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Automation & Control Theory (AREA)
  • Control Of Position, Course, Altitude, Or Attitude Of Moving Bodies (AREA)
  • Stored Programmes (AREA)
  • Telephone Function (AREA)

Abstract

一种自移动设备的重定位方法、设备和存储介质、其中重定位方法包括：响应于对自移动设备在工作区域内进行重定位的指令，获取子移动设备基于当前所处位置采集的当前环境信息（S201）；对当前环境信息进行识别，得到当前环境信息对应的区域标识信息(S202)；在区域地图中确定区域标识信息指示的局部区域地图(S203)；获取局部区域地图中至少一个地图位置对应的模板环境信息(S204)；将当前环境信息与模板环境信息进行匹配，以确定自移动设备在局部区域地图中位置(S205)。通过在自移动设备需要重定位时识别区域标识信息，并使用区域标识信息指示的局部区域地图进行重定位，可以提高重定位的效率, 解决了自移动设备重定位方式繁琐的问题。

Description

自移动设备的重定位方法、设备及存储介质
本公开要求如下专利申请的优先权：于2021年12月13日提交中国专利局、申请号为202111517584.6、发明名称为“自移动设备的重定位方法、设备及存储介质”的中国专利申请；上述专利申请的全部内容通过引用结合在本公开中。
技术领域
本公开属于人工智能技术领域，具体涉及一种自移动设备的重定位方法、设备及存储介质。
背景技术
目前，自移动设备可以借助同步定位与建图(Simultaneous Localization and Mapping，SLAM)技术实现自主定位和导航。然而，在进行SLAM过程中，可能会被劫持等情况，例如：被搬动、悬空或者被大范围拖动等情况。此时，自移动设备的定位就会出现不可控的漂移误差，需要进行重定位。
传统的重定位方法，包括：自移动设备从被劫持后的位置寻找原始的出发位置，从而完成对自移动设备的重定位。
然而，传统的重定位方式过于繁琐，这就会导致重定位效率较低的问题。
发明内容
本公开提供了自移动设备的重定位方法、设备及存储介质，可以解决自移动设备由于重定位方式繁琐导致的重定位效率低的问题。本公开提供如下技术方案：
第一方面，提供了一种自移动设备的重定位方法，所述方法包括：响应于对所述自移动设备在工作区域内进行重定位的指令，获取所述自移动设备基于当前所处位置采集的当前环境信息；
对所述当前环境信息进行识别，得到所述当前环境信息对应的区域标识信息；
在所述工作区域的区域地图中确定所述区域标识信息指示的局部区域地图；
获取所述局部区域地图中至少一个地图位置对应的模板环境信息；
将所述当前环境信息与所述模板环境信息进行匹配，以确定所述自移动设备在所述局部区域地图中的位置。
可选地，所述对所述当前环境信息进行识别，得到所述当前环境信息对应的区域标识信 息，包括：
将所述当前环境信息输入预先训练的区域识别模型，得到所述区域标识信息；所述区域识别模型使用训练数据对预设的神经网络模型训练得到；所述训练数据包括样本环境信息和所述样本环境信息对应的区域标签。
可选地，所述训练数据还包括所述样本环境信息对应的第一障碍物的分类标签，所述第一障碍物的分类标签用于结合所述区域标签对所述神经网络模型进行联合训练，得到所述区域识别模型；相应地，所述将所述当前环境信息输入预先训练的区域识别模型，得到所述区域标识信息，包括：将所述当前环境信息输入所述区域识别模型，得到所述当前环境信息对应的第一障碍物的分类结果和所述区域标识信息；
或者，
所述训练数据还包括所述第一障碍物的第一特征信息；相应地，所述将所述当前环境信息输入预先训练的区域识别模型，得到所述区域标识信息，包括：将所述当前环境信息和所述第一特征信息输入所述区域识别模型，得到所述区域识别模型；
其中，所述第一障碍物用于指示区域标识信息。
可选地，所述对所述当前环境信息进行识别，得到所述当前环境信息对应的区域标识信息，还包括：
获取第一障碍物的第二特征信息，所述第一障碍物用于指示区域标识信息；
将所述当前环境信息与所述第二特征信息进行匹配；
在所述当前环境信息包括与所述第二特征信息相匹配的信息的情况下，确定所述当前环境信息对应的区域标识信息为所述第一障碍物指示的区域标识信息。
可选地，所述第二特征信息包括所述第一障碍物的轮廓信息；所述第二特征信息还包括所述第一障碍物的尺寸信息和/或距离信息。
可选地，所述区域标识信息为第一障碍物在所述区域地图中的位置坐标；所述在所述工作区域的区域地图中确定所述区域标识信息指示的局部区域地图，包括：
在所述区域地图中，基于所述区域标识信息确定预设形状和预设尺寸的局部区域地图；
或者，
在已进行区域分割的区域地图中，确定所述区域标识信息所属的局部区域地图，所述区域地图预先分割为多个局部区域地图。
可选地，在所述工作区域的区域地图中确定所述区域标识信息指示的局部区域地图，包括：
基于区域标识信息与局部区域地图之间的对应关系，确定所述区域标识信息对应的局部 区域地图。
可选地，将所述当前环境信息与所述模板环境信息进行匹配，以确定所述自移动设备的位置，包括：
将所述当前环境信息和所述模板环境信息输入预先训练的重定位神经网络，得到所述位置；所述重定位神经网络用于确定所述当前环境信息和所述模板环境信息是否匹配，并将匹配的模板环境信息所对应的位置确定为所述位置。
可选地，所述模板环境信息包括第二障碍物的特征信息。
第二方面，提供了一种电子设备，所述设备包括处理器和存储器；所述存储器中存储有程序，所述程序由所述处理器加载并执行以实现第一方面所述的自移动设备的重定位方法。
第三方面，提供一种计算机可读存储介质，所述存储介质中存储有程序，所述程序被处理器执行时用于实现第一方面提供的自移动设备重定位方法。
本公开的有益效果在于：通过响应于对自移动设备在工作区域内进行重定位的指令，获取自移动设备基于当前所处位置采集的当前环境信息，对当前环境信息进行识别，得到当前环境信息对应的区域标识信息，在工作区域的区域地图中确定区域标识信息指示的局部区域地图，获取局部区域地图中至少一个地图位置对应的模板环境信息，将当前环境信息与模板环境信息进行匹配，以确定自移动设备在局部区域地图中的位置。可以解决自移动设备由于重定位方式繁琐导致的重定位效率低的问题。通过在自移动设备需要进行重定位时识别区域标识信息，并使用区域标识信息指示的局部区域地图对自移动设备当前所在的位置进行重定位。此时，自移动设备无需在整个工作区域内寻找某个位置，而是使用局部区域地图即可实现重定位，可以提高重定位的效率。同时，自移动设备无需移动至原始的出发位置，只需要将当前所处的位置的当前环境信息与局部区域地图的模板环境信息进行匹配，可以节省自移动设备的资源，且可以进一步提高重定位的效率。
另外，通过使用第一障碍物的分类标签和区域标签对神经网络模型进行联合训练，得到区域识别模型，训练得到的区域识别模型会更加精确，可以提高区域标识信息识别的准确性。
另外，在对当前环境信息进行识别时，区域识别模型可以将当前环境信息与第一特征信息进行比较，以确定是否存在第一障碍物，由于第一障碍物可以指示区域标识信息，因此，可以降低网络模型的计算难度，节省自移动设备的计算资源。
附图说明
此处所说明的附图用来提供对本公开的进一步理解，构成本公开的一部分，本公开的示意性实施例及其说明用于解释本公开，并不构成对本公开的不当限定。在附图中：
图1是本公开一个实施例提供的自移动设备的结构示意图；
图2是本公开一个实施例提供的自移动设备的重定位方法的流程图；
图3是本公开一个实施例提供的自移动设备重定位装置的框图；
图4是本公开一个实施例提供的电子设备的框图。
具体实施方式
下面将结合附图对本公开的技术方案进行清楚、完整地描述，显然，所描述的实施例是本公开一部分实施例，而不是全部的实施例。下文中将参考附图并结合实施例来详细说明本公开。需要说明的是，在不冲突的情况下，本公开中的实施例及实施例中的特征可以相互组合。
需要说明的是，本公开的说明书和权利要求书及上述附图中的术语“第一”、“第二”等是用于区别类似的对象，而不必用于描述特定的顺序或先后次序。
在公开中，在未作相反说明的情况下，使用的方位词如“上、下、顶、底”通常是针对附图所示的方向而言的，或者是针对部件本身在竖直、垂直或重力方向上而言的；同样地，为便于理解和描述，“内、外”是指相对于各部件本身的轮廓的内、外，但上述方位词并不用于限制本公开。
如图1所示为本公开一个实施例提供的自移动设备的结构示意图，该自移动设备可以为扫地机器人，洗地机器人等可自行移动的设备，本实施例不对自移动设备的设备类型作限定。根据图1可知，自移动设备至少包括：驱动组件110、移动组件120、控制器130和第一传感器140。
驱动组件110与移动组件120相连，并用于驱动移动组件120运行，以带动自移动设备移动。
驱动组件110与控制器130相连，并用于响应控制器130发出的指令，以驱动移动组件120运行。
可选地，驱动组件110可以实现为直流电机、伺服电机、步进电机等，本实施例不对驱动组件110的实现方式作限定。
第一传感器140用于采集当前环境信息。可选地，第一传感器140可以为搭载有颜色系统(Red Green Blue，RGB)检测功能的相机、红外传感器、或者激光雷达传感器等，本实施例不对第一传感器140的类型作限定。
可选地，第一传感器140可以安装于自移动设备的壳体上，且用于对自移动设备所处的环境进行采集。第一传感器140的采集范围包括但不限于：在自移动设备的行进方向正前方、 斜上方和/或斜下方区域；和/或自移动设备的行进方向的左侧区域；和/或自移动设备的行进方向的右侧区域；和/或自移动设备的行进方向的后侧区域等，本实施例不对第一传感器140的采集范围作限定。
另外，第一传感器140的数量可以为一个或者至少两个，在第一传感器140的数量为至少两个的情况下，不同第一传感器140的类型相同或不同，本实施例不对第一传感器140的数量和实现方式作限定。
第一传感器140与控制器130相连，以将采集到的当前环境信息发送至控制器130。
控制器130用于对自移动设备进行重定位。可选地，控制器130可以实现为单片机，或者处理器，本实施例不对控制器130的实现方式作限定。
本实施例中，控制器130用于：响应于对自移动设备在工作区域内进行重定位的指令，获取自移动设备基于当前所处位置采集的当前环境信息；对当前环境信息进行识别，得到当前环境信息对应的区域标识信息；在工作区域的区域地图中确定区域标识信息指示的局部区域地图；获取局部区域地图中至少一个地图位置对应的模板环境信息；将当前环境信息与模板环境信息进行匹配，以确定自移动设备在局部区域地图中的位置。
可选地，对自移动设备在工作区域内进行重定位的指令是自移动设备基于第二传感器的传感数据生成的。此时，自移动设备上还设置有第二传感器150，第二传感器150与控制器130相连，并用于向控制器130发送传感数据。相应地，控制器130接收到传感数据后，基于传感数据确定自移动设备是否被劫持；若是，则在传感数据指示自移动设备脱离劫持的情况，确定进行重定位，并生成进行重定位的指令；若确定自移动设备未被劫持，则再次执行自移动设备是否被劫持的步骤，直至自移动设备结束工作时停止。
其中，被劫持是指自移动设备发生非正常移动，非正常移动是指不是自移动设备自主发生的移动，非正常移动无法被自移动设备感应到。因此，自移动设备在被劫持的情况下，无法定位出或者无法准确定位出自身的位置。比如：自移动设备被搬动、移动过程中悬空、或被大范围拖动等情况，均为自移动设备被劫持的情况。
相应地，脱离劫持是指自移动设备结束非正常移动，比如：自移动设备被搬动后重新回到地面，或者自移动设备被拖动后当其停止被拖动的情况，均为自移动设备脱离劫持的情况。
示意性地，传感数据包括但不限于：自移动设备的高度数据、位移数据、和/或角度数据等，相应地，第二传感器150包括但不限陀螺仪、位移传感器、或者图像采集器等，本实施例不对第二传感器150的实现方式作限定。此时，基于传感数据确定自移动设备是否被劫持，包括：将传感数据的变化情况与被劫持状态下的模板变化情况进行比较；若传感数据的变化情况与被劫持状态下的模板变化情况相匹配，则确定自移动设备被劫持；若传感数据的变化 情况与被劫持状态下的模板变化情况不匹配，则确定自移动设备未被劫持。或者，基于传感数据确定自移动设备是否被劫持，包括：确定传感数据的变化值是否在被劫持状态下的变化范围内；若是，则确定自移动设备被劫持；若不是，则确定自移动设备未被劫持。
示意性地，传感数据包括但不限于：接触数据等，相应地，第二传感器150包括但不限压力传感器、或者接触传感器等，本实施例不对第二传感器150的实现方式作限定。此时，基于传感数据确定自移动设备是否被劫持，包括：确定传感数据是否指示存在物体接近自移动设备；若是，则确定自移动设备被劫持；若不是，则确定自移动设备未被劫持。
或者，对自移动设备在工作区域内进行重定位的指令是与自移动设备通信相连的控制设备发送的。其中，控制设备可以为手机、遥控器、或者可穿戴式设备等，本实施例不对控制设备的类型作限定。
或者，对自移动设备在工作区域内进行重定位的指令是自移动设备接收到作用于重定位控件的触发操作的情况下生成。此时，自移动设备上还设置有重定位控件，该重定位控件可以为实体按键或者为通过触摸显示屏显示的虚拟控件，本实施例不对重定位控件的实现方式作限定。
上述重定位的指令的获取方式仅是示意性的，在实际实现时，自移动设备获取重定位的指令的方式也可以是其它方式，本实施例在此不再一一列举。
需要补充说明的是，在实际实现时，自移动设备还可以包括其它元器件，如：供电组件、减震组件等，本实施例在此不再一一列举。
传统的重定位方法中，自移动设备从被劫持后的位置寻找原始的出发位置。寻找方式通常是：自移动设备在工作区域中随机移动，并通过红外接收装置接收原始的出发位置(如充电座)发射的红外信号。在通过红外接收装置接收到红外信号的情况下，向该红外信号的位置移动，以移动至原始的出发位置，并读取工作区域的区域地图中存储的原始的出发位置的地图位置，完成重定位。然而，传统的重定位方式并不能在脱离劫持时对自移动设备进行重定位，需在整个工作区域内搜索原始的出发位置，重定位的效率较低。而本实施例中，通过在自移动设备需要进行重定位时识别区域标识信息，并使用区域标识信息指示的局部区域地图对自移动设备当前所在的位置进行重定位。此时，自移动设备无需在整个工作区域内寻找某个位置，而是使用局部区域地图即可实现重定位，可以提高重定位的效率。同时，自移动设备无需移动至原始的出发位置，只需要将当前所处的位置的当前环境信息与局部区域地图的模板环境信息进行匹配，可以节省自移动设备的资源，且可以进一步提高重定位的效率。
下面对本公开提供的自移动设备的重定位方法进行详细介绍。
本实施例提供的一种自移动设备的重定位方法，如图2所示。本实施例以该方法用于图1所示的控制器130为例进行说明。该方法至少包括以下几个步骤：
步骤201，响应于对自移动设备在工作区域内进行重定位的指令，获取自移动设备基于当前所处位置采集的当前环境信息。
对自移动设备进行重定位的指令是自移动设备基于第二传感器的传感数据生成的；或者，是与自移动设备通信相连的控制设备发送的；或者，是自移动设备接收到作用于重定位控件的触发操作的情况下生成，本实施例不对自移动设备进行重定位的指令的获取方式作限定。
在一个示例中，当前环境信息是对自移动设备当前所处的环境进行采集得到的。当前环境信息可以为图像数据、和/或点云数据，且当前环境信息可以为三维数据或者二维数据，本实施例不对当前环境信息的实现方式作限定。
可选地，当前环境信息可以是自移动设备获取到进行重定位的指令时，控制器控制第一传感器采集的；或者，是第一传感器上电后持续采集的，本实施例不对当前环境信息的采集时机作限定。
步骤202，对当前环境信息进行识别，得到当前环境信息对应的区域标识信息。
区域标识信息用于唯一地指示工作区域中的某个局部区域。
可选地，区域标识信息为第一障碍物在区域地图中的位置坐标。其中，第一障碍物用于指示区域标识信息。具体地，第一障碍物是指能够指示局部区域的属性的障碍物。比如：第一障碍物为餐桌，餐桌指示的局部区域的属性为餐厅；又比如：第一障碍物为马桶，马桶指示的局部区域的属性为卫生间；再比如：第一障碍物为床，床指示的局部区域的属性为卧室，本实施例在此不对第一障碍物的实现方式一一列举。
或者，区域标识信息为局部区域的区域标识，该区域标识可以为局部区域的属性或标号。比如：在区域标识为属性的情况下，区域标识信息为餐厅、卫生间和/或卧室等。比如：在区域标识为标号的情况下，区域标识信息区域地图中预先为各个局部区域地图设置的标号1、2、3等。
本实施例中，以局部区域的属性包括餐厅、卫生间和卧室为例进行说明，在实际实现时，局部区域的属性划分方式也可以是其它方式，比如：将局部区域的属性划分为：办公区域、茶歇区域等，本实施例不对局部区域的属性划分方式作限定。
对当前环境信息进行识别，得到当前环境信息对应的区域标识信息的方式包括但不限于以下中的至少一种：
第一种识别方式，将当前环境数据输入预先训练的区域识别模型，得到区域标识信息。其中，区域识别模型是使用训练数据对预设的神经网络模型训练得到的。
可选地，训练数据的实现情况包括但不限于以下几种情况：
第一种情况，训练数据仅包括样本环境信息和样本环境信息对应的区域标签。
相应地，区域识别模型的训练过程包括：将样本环境信息输入预设的第一神经网络模型，得到第一训练结果；将第一训练结果和样本环境信息对应的区域标签输入第一损失函数，得到第一损失结果；基于第一损失结果对第一神经网络模型进行训练，以缩小第一训练结果和对应的区域标签之间的差异值，直至神经网络模型收敛，得到区域识别模型。
其中，在区域标签为第一障碍物的位置坐标标签的情况下，区域标识信息为第一障碍物的位置坐标；在区域标签为第一障碍物的区域属性的情况下，区域标识信息为第一障碍物的区域属性。
当前环境信息输入预先训练的区域识别模型，得到区域标识信息，包括：将当前环境信息输入预先训练的区域识别模型，得到当前环境信息对应的区域标识信息。
其中，第一神经网络模型可以为卷积神经网络(Convolutional Neural Networks，CNN)、递归神经网络(Recursive Neural Network，RNN)、前馈神经网络(Feedforward Neural Network，FNN)，本实施例不对第一神经网络模型的实现方式作限定。
第二种情况，训练数据不仅包括样本环境信息和样本环境信息对应的区域标签，还包括样本信息对应的第一障碍物的分类标签，其中，第一障碍物的分类标签用于结合区域标签对神经网络模型进行联合训练，得到区域识别模型。
其中，第一障碍物的分类标签为第一障碍物的属性标签。
相应地，区域识别模型的训练过程包括：将样本环境信息输入预设的第二神经网络模型，得到第二训练结果，该第二训练结果包括区域预测结果和分类预测结果；将第二训练结果、区域标签和分类标签输入第二损失函数，得到第二损失结果；基于第二损失结果对第二神经网络模型进行训练，以缩小第二训练结果和对应的区域标签和分类标签之间的差异值，直至第二神经网络模型收敛，得到区域识别模型。
其中，第二神经网络模型包括两个网络分支，其中一个网络分支用于计算区域预测结果，另一个分支用于计算分类结果。
由于在训练过程中，两个网络分支的结果均接近真实值时，第二神经网络模型的训练才会结束，因此，训练得到的区域识别模型会更加精确，可以提高区域标识信息识别的准确性。
相应地，当前环境信息输入预先训练的区域识别模型，得到区域标识信息，包括：将当前环境信息输入区域识别模型，得到当前环境信息对应的第一障碍物的分类结果和区域标识信息。
第三种情况，训练数据不仅包括样本环境信息和样本环境信息对应的区域标签，还包括 样本环境信息对应的第一障碍物的第一特征信息。
可选地，第一特征信息可以为第一障碍物的轮廓信息、或者为第一障碍物的特征向量，本实施例不对第一特征信息的实现方式作限定。
相应地，区域识别模型的训练过程包括：将样本环境信息和第一特征信息输入至预设的第三神经网络模型，得到第三训练结果；将第三训练结果和区域标签输入第三损失函数，得到第三损失结果；基于第三损失结果对第三神经网络模型进行训练，以缩小第三训练结果和对应的区域标签之间的差异值，直至第三神经网络模型收敛，得到区域识别模型。
第三神经网络模型用于将样本环境信息与第一特征信息进行比较，以确定样本环境信息是否与第一特征信息相匹配的信息，从而确定样本环境信息是否存在第一障碍物，由于第一障碍物可以指示区域标识信息，因此，可以降低网络模型的计算难度，节省自移动设备的计算资源。
相应地，将当前环境信息输入预先训练的区域识别模型，得到区域标识信息，包括：将当前环境信息和第一障碍物的第一特征信息输入区域识别模型，得到区域标识信息。
其中，第一障碍物的第一特征信息预存在自移动设备中。
第二种识别方式，获取第一障碍物的第二特征信息；将获取的当前环境信息与第二特征信息进行匹配；若在当前环境信息包括与第二特征信息相匹配的信息的情况下，则确定当前环境信息对应的区域标识信息为第一障碍物指示的区域标识信息。
可选地，第二特征信息与第一特征信息相同或不同。第二特征信息可以为第一障碍物的轮廓信息、第一障碍物的尺寸信息和/或距离信息或者可以用来对第一障碍物进行特征描述的信息，本实施例不对第二特征信息作限定。
比如：第一障碍物为餐桌，第二特征信息为餐桌的形状与尺寸。在当前环境信息包括与餐桌的形状和尺寸相匹配的信息的情况下，则确定当前环境信息对应的区域标识信息为餐桌所指示的区域标识信息，如餐厅。
步骤203，在工作区域的区域地图中确定区域标识信息指示的局部区域地图。
本实施例中，局部区域地图可以基于区域标识信息的改变而变化。
在一个示例中，区域标识信息为第一障碍物在区域地图中的位置坐标。相应地，在工作区域的区域地中确定区域标识信息指示的局部区域地图的方式包括但不限于以下的至少一种：
第一种：在区域地图中，基于区域标识信息确定预设形状和预设尺寸的局部区域地图。
其中，预设形状和预设尺寸预存在自移动设备中。预设形状可以为圆形、矩形、或者不规则形状，本实施例不对预设形状的实现方式作限定。
示意性地，基于区域标识信息确定预设形状和预设尺寸的局部区域地图，包括：以第一障碍物的位置坐标为局部区域地图的形心，生成预设形状和预设尺寸的局部区域地图。
比如：第一障碍物为充电座，区域标识信息为充电座的位置坐标。预设形状为圆形、且预设尺寸为半径为2米。此时，确定该局部区域地图是区域地图上以充电座的位置坐标为圆心，半径为2米的圆形区域。
在其它实施例中，第一障碍物的位置坐标也可以为位于局部区域地图的边缘或其它位置，本实施例不对基于区域标识信息确定预设形状和预设尺寸的局部区域地图的方式作限定。
第二种：在已进行区域分割的区域地图中，确定区域标识信息所属的局部区域地图，该区域地图预先分割为多个局部区域地图。
其中，将区域地图分割为多个局部区域地图的方式包括但不限于：按照属性对区域地图划分，或者按照预设分割尺寸对区域地图划分，在实际实现时，区域地图的分割方式也可以为其它方式，本实施例在此不再一一列举。
比如：将区域地图分割为一个卧室区域地图，卫生间区域地。而区域标识信息为充电座的位置坐标时，该充电座位于卧室区域内，则而区域标识信息所属的局部区域地图为卧室区域地图。
在另一个示例中，自移动设备中存储有区域标识信息与局部区域地图之间的对应关系。此时，在工作区域的区域地图中确定区域标识信息指示的局部区域地图，包括：基于区域标识信息与局部区域地图之间的对应关系，确定区域标识信息对应的局部区域地图。
比如：区域标识信息为各个局部区域地图的标号，此时，自移动设备获取到该标号后，从对应关系中查找到该标号对应的局部区域地图。
步骤204，获取局部区域地图中至少一个地图位置对应的模板环境信息。
本实施例中，局部区域地图中包括多个地图位置对应的模板环境信息，多个模板环境信息均具有对应地图位置的位置坐标。控制器获取其中至少一个地图位置对应的模板环境信息。
步骤205，将当前环境信息与模板环境信息进行匹配，以确定自移动设备在局部区域地图中的位置。
其中，模板环境信息包括第二障碍物的特征信息。
将当前环境信息与模板环境信息进行匹配的方式包括但不限于以下中的至少一种：
第一种，将当前环境信息和模板环境信息输入预先训练的重定位神经网络，得到自移动设备在局部区域地图中的位置。重定位神经网络用于确定当前环境信息和模板环境信息是否匹配，并将匹配的模板环境信息所对应的位置确定为自移动设备在局部区域地图中的位置。
可选地，模板环境信息包括第二障碍物的特征信息。第二障碍物可以为工作区域内的任 意障碍物，如：桌子、椅子、地毯、墙壁等，本实施例不对第二障碍物的类型作限定。
第二障碍物的特征信息可以为第二障碍物的形状、尺寸或者特征向量等，本实施例不对第二障碍物的特征信息的实现方式作限定。
相应地，重定位神经网络的训练过程包括：将样本环境信息与各个模板环境信息输入预设的神经网络模型，得到相似度结果；将该相似度结果与真实相似度结果进行比较，基于比较结果训练该神经网络模型，得到重定位神经网络。
将当前环境信息与模板环境信息进行匹配，以确定自移动设备在局部区域地图中的位置，包括：将当前环境信息与模板环境信息输入预先训练的重定位神经网络，得到多个相似度结果，将相似度结果排在第一位的环境模板信息所对应的地图位置确定为自移动设备在局部区域地图中的位置。
第二种，计算当前环境信息与各个模板环境信息之间的相似度；相似度最大的环境模板信息所对应的地图位置确定为自移动设备在局部区域地图中的位置。
综上所述，本实施例提供的自移动设备的重定位方法，通过响应于对自移动设备在工作区域内进行重定位的指令，获取自移动设备基于当前所处位置采集的当前环境信息，对当前环境信息进行识别，得到当前环境信息对应的区域标识信息，在工作区域的区域地图中确定区域标识信息指示的局部区域地图，获取局部区域地图中至少一个地图位置对应的模板环境信息，将当前环境信息与模板环境信息进行匹配，以确定自移动设备在局部区域地图中的位置。可以解决自移动设备由于重定位方式繁琐导致的重定位效率低的问题。通过在自移动设备需要进行重定位时识别区域标识信息，并使用区域标识信息指示的局部区域地图对自移动设备当前所在的位置进行重定位。此时，自移动设备无需在整个工作区域内寻找某个位置，而是使用局部区域地图即可实现重定位，可以提高重定位的效率。同时，自移动设备无需移动至原始的出发位置，只需要将当前所处的位置的当前环境信息与局部区域地图的模板环境信息进行匹配，可以节省自移动设备的资源，且可以进一步提高重定位的效率。
另外，通过使用第一障碍物的分类标签和区域标签对神经网络模型进行联合训练，得到区域识别模型，训练得到的区域识别模型会更加精确，可以提高区域标识信息识别的准确性。
另外，在对当前环境信息进行识别时，区域识别模型可以将当前环境信息与第一特征信息进行比较，以确定是否存在第一障碍物，由于第一障碍物可以指示区域标识信息，因此，可以降低网络模型的计算难度，节省自移动设备的计算资源。
图3是本公开一个实施例提供的自移动设备重定位装置的框图，本实施例以该装置应用于图1所示的自移动设备中为例进行说明。该装置至少包括以下几个模块：第一获取模块310、 信息识别模块320、地图确定模块330、第二获取模块340和重定位模块350。
第一获取模块310，用于响应于对自移动设备在工作区域内进行重定位的指令，获取自移动设备基于当前所处位置采集的当前环境信息。
信息识别模块320，用于对当前环境信息进行识别，得到当前环境信息对应的区域标识信息。
地图确定模块330，用于在工作区域的区域地图中确定区域标识信息指示的局部区域地图。
第二获取模块340，用于获取局部区域地图中至少一个地图位置对应的模板环境信息。
重定位模块350，用于将当前环境信息与模板环境信息进行匹配，以确定自移动设备在局部区域地图中的位置。
相关细节参考上述实施例。
需要说明的是：上述实施例中提供的自移动设备重定位装置在进行重定位时，仅以上述各功能模块的划分进行举例说明，实际应用中，可以根据需要而将上述功能分配由不同的功能模块完成，即将自移动设备重定位装置的内部结构划分成不同的功能模块，以完成以上描述的全部或者部分功能。另外，上述实施例提供的自移动设备重定位装置与自移动设备重定位方法实施例属于同一构思，其具体实现过程详见方法实施例，这里不再赘述。
本实施例提供一种电子设备，如图4所示，该电子设备可以为图1中的自移动设备。该电子设备至少包括处理器401和存储器402。
处理器401可以包括一个或多个处理核心，比如：4核心处理器、8核心处理器等。处理器401可以采用DSP(Digital Signal Processing，数字信号处理)、FPGA(Field－Programmable Gate Array，现场可编程门阵列)、PLA(Programmable Logic Array，可编程逻辑阵列)中的至少一种硬件形式来实现。处理器401也可以包括主处理器和协处理器，主处理器是用于对在唤醒状态下的数据进行处理的处理器，也称CPU(Central Processing Unit，中央处理器)；协处理器是用于对在待机状态下的数据进行处理的低功耗处理器。在一些实施例中，处理器401可以在集成有GPU(Graphics Processing Unit，图像处理器)，GPU用于负责显示屏所需要显示的内容的渲染和绘制。一些实施例中，处理器401还可以包括AI(Artificial Intelligence，人工智能)处理器，该AI处理器用于处理有关机器学习的计算操作。
存储器402可以包括一个或多个计算机可读存储介质，该计算机可读存储介质可以是非暂态的。存储器402还可包括高速随机存取存储器，以及非易失性存储器，比如一个或多个磁盘存储设备、闪存存储设备。在一些实施例中，存储器402中的非暂态的计算机可读存储 介质用于存储至少一个指令，该至少一个指令用于被处理器401所执行以实现本公开中方法实施例提供的电机刹车方法。
在一些实施例中，电子设备还可选包括有：外围设备接口和至少一个外围设备。处理器401、存储器402和外围设备接口之间可以通过总线或信号线相连。各个外围设备可以通过总线、信号线或电路板与外围设备接口相连。示意性地，外围设备包括但不限于：射频电路、触摸显示屏、音频电路、和电源等。
当然，电子设备还可以包括更少或更多的组件，本实施例对此不作限定。
可选地，本公开还提供有一种计算机可读存储介质，计算机可读存储介质中存储有程序，程序由处理器加载并执行以实现上述方法实施例的自移动设备重定位方法。
以上所述实施例的各技术特征可以进行任意的组合，为使描述简洁，未对上述实施例中的各个技术特征所有可能的组合都进行描述，然而，只要这些技术特征的组合不存在矛盾，都应当认为是本说明书记载的范围。
以上所述实施例仅表达了本公开的几种实施方式，其描述较为具体和详细，但并不能因此而理解为对发明专利范围的限制。应当指出的是，对于本领域的普通技术人员来说，在不脱离本公开构思的前提下，还可以做出若干变形和改进，这些都属于本公开的保护范围。因此，本公开专利的保护范围应以所附权利要求为准。

Claims (11)

  1. 一种自移动设备的重定位方法，其特征在于，所述方法包括：
    响应于对所述自移动设备在工作区域内进行重定位的指令，获取所述自移动设备基于当前所处位置采集的当前环境信息；
    对所述当前环境信息进行识别，得到所述当前环境信息对应的区域标识信息；
    在所述工作区域的区域地图中确定所述区域标识信息指示的局部区域地图；
    获取所述局部区域地图中至少一个地图位置对应的模板环境信息；
    将所述当前环境信息与所述模板环境信息进行匹配，以确定所述自移动设备在所述局部区域地图中的位置。
  2. 根据权利要求1所述的方法，其中，所述对所述当前环境信息进行识别，得到所述当前环境信息对应的区域标识信息，包括：
    将所述当前环境信息输入预先训练的区域识别模型，得到所述区域标识信息；所述区域识别模型使用训练数据对预设的神经网络模型训练得到；所述训练数据包括样本环境信息和所述样本环境信息对应的区域标签。
  3. 根据权利要求2所述的方法，其中，
    所述训练数据还包括所述样本环境信息对应的第一障碍物的分类标签，所述第一障碍物的分类标签用于结合所述区域标签对所述神经网络模型进行联合训练，得到所述区域识别模型；相应地，所述将所述当前环境信息输入预先训练的区域识别模型，得到所述区域标识信息，包括：将所述当前环境信息输入所述区域识别模型，得到所述当前环境信息对应的第一障碍物的分类结果和所述区域标识信息；
    或者，
    所述训练数据还包括所述第一障碍物的第一特征信息；相应地，所述将所述当前环境信息输入预先训练的区域识别模型，得到所述区域标识信息，包括：将所述当前环境信息和所述第一特征信息输入所述区域识别模型，得到所述区域识别模型；
    其中，所述第一障碍物用于指示区域标识信息。
  4. 根据权利要求1所述的方法，其中，所述对所述当前环境信息进行识别，得到所述当前环境信息对应的区域标识信息，还包括：
    获取第一障碍物的第二特征信息，所述第一障碍物用于指示区域标识信息；
    将所述当前环境信息与所述第二特征信息进行匹配；
    在所述当前环境信息包括与所述第二特征信息相匹配的信息的情况下，确定所述当前环境信息对应的区域标识信息为所述第一障碍物指示的区域标识信息。
  5. 根据权利要求3所述的方法，其中，所述第二特征信息包括所述第一障碍物的轮廓信 息；所述第二特征信息还包括所述第一障碍物的尺寸信息和/或距离信息。
  6. 根据权利要求1所述的方法，其中，所述区域标识信息为第一障碍物在所述区域地图中的位置坐标；所述在所述工作区域的区域地图中确定所述区域标识信息指示的局部区域地图，包括：
    在所述区域地图中，基于所述区域标识信息确定预设形状和预设尺寸的局部区域地图；
    或者，
    在已进行区域分割的区域地图中，确定所述区域标识信息所属的局部区域地图，所述区域地图预先分割为多个局部区域地图。
  7. 根据权利要求1所述的方法，其中，在所述工作区域的区域地图中确定所述区域标识信息指示的局部区域地图，包括：
    基于区域标识信息与局部区域地图之间的对应关系，确定所述区域标识信息对应的局部区域地图。
  8. 根据权利要求1所述的方法，其中，将所述当前环境信息与所述模板环境信息进行匹配，以确定所述自移动设备的位置，包括：
    将所述当前环境信息和所述模板环境信息输入预先训练的重定位神经网络，得到所述位置；所述重定位神经网络用于确定所述当前环境信息和所述模板环境信息是否匹配，并将匹配的模板环境信息所对应的位置确定为所述位置。
  9. 根据权利要求8所述的方法，其中，所述模板环境信息包括第二障碍物的特征信息。
  10. 一种电子设备，其中，所述设备包括处理器和存储器；所述存储器中存储有程序，所述程序由所述处理器加载并执行以实现如权利要求1至9任一所述的自移动设备的重定位方法。
  11. 一种计算机可读存储介质，其中，所述存储介质中存储有程序，所述程序被处理器执行时用于实现如权利要求1至9任一所述的自移动设备的重定位方法。
PCT/CN2022/129455 2021-12-13 2022-11-03 自移动设备的重定位方法、设备及存储介质 Ceased WO2023109347A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202111517584.6 2021-12-13
CN202111517584.6A CN116263598A (zh) 2021-12-13 2021-12-13 自移动设备的重定位方法、设备及存储介质

Publications (1)

Publication Number Publication Date
WO2023109347A1 true WO2023109347A1 (zh) 2023-06-22

Family

ID=86721797

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2022/129455 Ceased WO2023109347A1 (zh) 2021-12-13 2022-11-03 自移动设备的重定位方法、设备及存储介质

Country Status (2)

Country Link
CN (1) CN116263598A (zh)
WO (1) WO2023109347A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2026037087A1 (zh) * 2024-08-13 2026-02-19 深圳市普渡科技股份有限公司 重定位方法、装置、机器人和存储介质

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2012107892A2 (en) * 2011-02-09 2012-08-16 Primesense Ltd. Gaze detection in a 3d mapping environment
CN107037806A (zh) * 2016-02-04 2017-08-11 科沃斯机器人股份有限公司 自移动机器人重新定位方法及采用该方法的自移动机器人
CN111158374A (zh) * 2020-01-10 2020-05-15 惠州拓邦电气技术有限公司 重定位方法、系统、移动机器人及存储介质
CN111539398A (zh) * 2020-07-13 2020-08-14 追创科技(苏州)有限公司 自移动设备的控制方法、装置及存储介质
CN111539400A (zh) * 2020-07-13 2020-08-14 追创科技(苏州)有限公司 自移动设备的控制方法、装置、存储介质及自移动设备

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110605713B (zh) * 2018-06-15 2023-06-27 科沃斯机器人股份有限公司 机器人定位方法、机器人及存储介质
CN113343830B (zh) * 2021-06-01 2024-05-24 上海追势科技有限公司 一种地下停车场车辆快速重定位的方法

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2012107892A2 (en) * 2011-02-09 2012-08-16 Primesense Ltd. Gaze detection in a 3d mapping environment
CN107037806A (zh) * 2016-02-04 2017-08-11 科沃斯机器人股份有限公司 自移动机器人重新定位方法及采用该方法的自移动机器人
CN111158374A (zh) * 2020-01-10 2020-05-15 惠州拓邦电气技术有限公司 重定位方法、系统、移动机器人及存储介质
CN111539398A (zh) * 2020-07-13 2020-08-14 追创科技(苏州)有限公司 自移动设备的控制方法、装置及存储介质
CN111539400A (zh) * 2020-07-13 2020-08-14 追创科技(苏州)有限公司 自移动设备的控制方法、装置、存储介质及自移动设备

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2026037087A1 (zh) * 2024-08-13 2026-02-19 深圳市普渡科技股份有限公司 重定位方法、装置、机器人和存储介质

Also Published As

Publication number Publication date
CN116263598A (zh) 2023-06-16

Similar Documents

Publication Publication Date Title
CN109643127B (zh) 构建地图、定位、导航、控制方法及系统、移动机器人
US11227434B2 (en) Map constructing apparatus and map constructing method
KR102235270B1 (ko) 이동 로봇 시스템 및 그 제어방법
CN109863003B (zh) 移动机器人及其控制方法
KR20240063820A (ko) 청소 로봇 및 그의 태스크 수행 방법
CN113116224B (zh) 机器人及其控制方法
WO2019232806A1 (zh) 导航方法、导航系统、移动控制系统及移动机器人
KR20250069501A (ko) 청소 로봇 및 그의 태스크 수행 방법
US11027425B1 (en) Space extrapolation for robot task performance
CN114612786B (zh) 一种障碍物检测方法、移动机器人及机器可读存储介质
CN114245091B (zh) 投影位置修正方法、投影定位方法及控制装置、机器人
WO2022078467A1 (zh) 机器人自动回充方法、装置、机器人和存储介质
WO2019091310A1 (zh) 区域属性确定
WO2020223975A1 (zh) 在地图上定位设备的方法、服务端及移动机器人
KR20210029586A (ko) 이미지 내의 특징적 객체에 기반하여 슬램을 수행하는 방법 및 이를 구현하는 로봇과 클라우드 서버
WO2022017341A1 (zh) 自动回充方法、装置、存储介质、充电基座及系统
CN109933061A (zh) 基于人工智能的机器人及控制方法
CN115164906A (zh) 定位方法、机器人和计算机可读存储介质
US20230297120A1 (en) Method, apparatus, and device for creating map for self-moving device with improved map generation efficiency
CN111679661A (zh) 基于深度相机的语义地图构建方法及扫地机器人
US12504758B1 (en) Shared floor plan system for autonomous mobile devices
CN114739408A (zh) 一种语义导航方法、语义导航装置及机器人
WO2023109347A1 (zh) 自移动设备的重定位方法、设备及存储介质
KR102780703B1 (ko) 청소 로봇 및 그의 태스크 수행 방법
CN116412824A (zh) 自移动设备的重定位方法、设备及存储介质

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 22906099

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 22906099

Country of ref document: EP

Kind code of ref document: A1