An agent navigation map updating method, device and medium

Abstract

The specification discloses an agent navigation map updating method, device and medium, relates to the technical field of artificial intelligence, and the method comprises the following steps: collecting real-time video corresponding to a current scene, and extracting a plurality of current key frames from the real-time video to obtain a current key frame group; based on current position information and current attitude information of an agent, selecting a preset key frame matched with each current key frame from a preset key frame group corresponding to a preset map; calculating a matching degree of the current scene and the preset map according to the current key frame and the corresponding preset key frame; and if the matching degree does not satisfy a preset condition, performing navigation map updating.

Description

Technical Field

[0001] This specification relates to the field of artificial intelligence technology, and in particular to a method, device and medium for updating navigation maps for intelligent agents. Background Technology

[0002] Most current intelligent agents rely on known maps to achieve autonomous indoor navigation. Some intelligent agents (such as robotic vacuum cleaners) use SLAM (Simultaneous Localization and Mapping) technology to achieve autonomous navigation. First, they use sensors (such as LiDAR and vision sensors) to acquire information about the surrounding environment. Then, they construct a temporary local environment map, simultaneously calculating the agent's pose within this map, and finally obtaining a global map for navigation, such as the two-dimensional grid map commonly used for indoor navigation (see attached diagram in the manual). Figure 1 A more detailed 3D semantic map with richer scene information (see attached diagram in the instruction manual). Figure 2 )etc.

[0003] However, the target application scenarios for intelligent agents are mostly complex and ever-changing. For example, some service robots are planned for use in large supermarkets, parking garages, exhibition halls, etc., to perform tasks such as cleaning, delivery, and guidance. After the initial map is built, the operating scenarios of these service robots may change. For instance, the store layout and promotional activities in the shopping mall may change; supermarket shelves may be adjusted; and parking conditions in the garage may change (see...). Figure 1 The orange boxes highlight some comparisons (significant changes in parking availability at different times of the day); the layout of the exhibition hall may also change. If the initial map is still used for navigation, it could lead to inaccurate agent positioning or even large jumps in path, affecting the accuracy of navigation path planning and scene perception.

[0004] Most AI agents on the market currently lack map update capabilities. Their user manuals often specify that changes in the operating environment must not exceed a certain threshold, significantly reducing their usability. Some AI agents allow manual map updates, where the user judges the degree of scene change, manually deletes the existing map, and rebuilds a new one. However, this method is less intelligent, the accuracy of user judgment is poor, and updating requires rebuilding the entire scene map. In large scenes where only some areas have changed, rebuilding the entire map is inefficient. Other AI agents use 2D LiDAR for mapping and positioning. This method is only applicable to grid maps, integrating the map update mechanism into the agent's daily operation and positioning. It merges continuously updated scene changes with the old map to obtain a new one. However, this method can only be used with LiDAR-built maps, and the algorithm needs to be continuously running during agent operation, requiring high computing power. A single-line LiDAR can only capture geometric contour information at its installation height, failing to comprehensively describe the current operating scene.

[0005] Therefore, existing intelligent agents cannot fully and accurately describe the current operating scenario, and cannot achieve precise and efficient positioning and navigation. Summary of the Invention

[0006] This specification provides one or more embodiments of an intelligent agent navigation map update method, device, and medium to solve the following technical problem: existing intelligent agents cannot fully and accurately describe the current operating scenario, and cannot achieve accurate and efficient positioning and navigation.

[0007] This specification provides one or more embodiments of a method for updating a navigation map for an intelligent agent, the method comprising:

[0008] Acquire real-time video corresponding to the current scene, and extract multiple current keyframes from the real-time video to obtain the current keyframe group;

[0009] Based on the agent's current position and current posture information, select preset key frames that match each current key frame from the preset key frame group corresponding to the preset map.

[0010] Based on the current keyframe and the corresponding preset keyframe, calculate the degree of matching between the current scene and the preset map;

[0011] If the matching degree does not meet the preset conditions, the navigation map will be updated.

[0012] Further, the step of calculating the matching degree between the current scene and the preset map based on the current keyframe and the corresponding preset keyframe includes:

[0013] Calculate the image similarity between each current keyframe and each corresponding preset keyframe;

[0014] If the image similarity is higher than the first preset threshold, then the current keyframe is determined to be a similar keyframe that is similar to the corresponding preset keyframe;

[0015] The proportion of similar keyframes in the current keyframe group is statistically analyzed, and this proportion is used as the degree of matching between the current scene and the preset map.

[0016] If the matching degree does not meet the preset conditions, the navigation map will be updated, including:

[0017] If the ratio is not higher than a preset second threshold, then the navigation map is updated.

[0018] Furthermore, updating the navigation map includes:

[0019] If the image similarity is lower than the first preset threshold, then the current keyframe is determined to be a distinguishing keyframe;

[0020] The preset map is updated based on the distinguishing keyframes;

[0021] Obtain the difference position information and difference pose information corresponding to the difference keyframes;

[0022] The preset navigation route is updated based on the distinguished location information and the distinguished attitude information.

[0023] Furthermore, before acquiring the real-time video corresponding to the current scene, the method further includes:

[0024] When the preset detection time requirement is met, the first detection command is triggered;

[0025] Based on the first detection command, move along the preset navigation route and execute the steps of claim 1.

[0026] Furthermore, it also includes:

[0027] Obtain the current location information of the intelligent agent, and calculate a location score based on the current location information;

[0028] When the positioning score is lower than the third preset threshold, a second detection command is triggered;

[0029] Based on the second detection command, the current keyframe corresponding to the current scene is collected;

[0030] Calculate the image similarity between the current keyframe and the preset keyframe, and find the matching keyframe in the preset keyframe group when the image similarity is greater than the preset first threshold.

[0031] If the matching keyframe is not found in the preset keyframe group, the navigation map is updated.

[0032] Furthermore, it also includes:

[0033] Based on the selection of a preset keyframe that matches the current keyframe from the preset keyframe group corresponding to the preset map;

[0034] Obtain the preset position information of the agent corresponding to the preset keyframe, and calculate the distance difference between the preset position information and the current position information;

[0035] If the distance difference is greater than a preset fourth threshold, a stop-forward command is triggered and the navigation route is updated.

[0036] Furthermore, the step of extracting multiple current keyframes from the real-time video to obtain a current keyframe group includes:

[0037] Generate video frame groups based on the real-time video;

[0038] The video frame group is clustered using a clustering algorithm to obtain multiple video frame categories;

[0039] Select one video frame from each of the aforementioned video frame categories as the current keyframe.

[0040] Further, calculating the image similarity between each current keyframe and its corresponding preset keyframe includes:

[0041] Extract the visual feature information of the current keyframe and the preset keyframe, and perform clustering operation on the visual feature information to obtain multiple feature words;

[0042] Generate bag-of-words vectors based on the feature words, and calculate the distance between the bag-of-words vectors as the image similarity.

[0043] This specification provides one or more embodiments of an intelligent agent navigation map update method device, including:

[0044] At least one processor; and,

[0045] A memory communicatively connected to the at least one processor; wherein,

[0046] The memory stores instructions that can be executed by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to:

[0047] Acquire real-time video corresponding to the current scene, and extract multiple current keyframes from the real-time video to obtain the current keyframe group;

[0048] Based on the agent's current position and current posture information, select preset key frames that match each current key frame from the preset key frame group corresponding to the preset map.

[0049] Based on the current keyframe and the corresponding preset keyframe, calculate the degree of matching between the current scene and the preset map;

[0050] If the matching degree does not meet the preset conditions, the navigation map will be updated.

[0051] This specification provides one or more embodiments of a non-volatile computer storage medium storing computer-executable instructions, wherein the computer-executable instructions are configured as follows:

[0052] Acquire real-time video corresponding to the current scene, and extract multiple current keyframes from the real-time video to obtain the current keyframe group;

[0053] Based on the agent's current position and current posture information, select preset key frames that match each current key frame from the preset key frame group corresponding to the preset map.

[0054] Based on the current keyframe and the corresponding preset keyframe, calculate the degree of matching between the current scene and the preset map;

[0055] If the matching degree does not meet the preset conditions, the navigation map will be updated.

[0056] The above-described at least one technical solution used in the embodiments of this specification can achieve the following beneficial effects:

[0057] By using visual sensor information and algorithms to determine scene changes, no human intervention is required, resulting in high intelligence and accuracy. It can identify which areas experience significant scene changes and require map updates, thus enabling mapping to be performed only in those areas, ensuring the efficiency of the intelligent agent. It innovatively incorporates visual elements and imposes no restrictions on the mapping and localization algorithms, sensors, or map formats used by the intelligent agent itself, resulting in higher intelligence and flexibility. This technology does not require the intelligent agent to be constantly running, saving computing power. Attached Figure Description

[0058] To more clearly illustrate the technical solutions in the embodiments or prior art of this specification, the drawings used in the description of the embodiments or prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this specification. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. In the drawings:

[0059] Figure 1 A flowchart illustrating an intelligent agent navigation map update method provided in an embodiment of this specification;

[0060] Figure 2 This is a schematic diagram of the structure of an intelligent agent navigation map update device provided in the embodiments of this specification;

[0061] Figure 3 This is a schematic diagram of a conventional detection mode provided in the embodiments of this specification. Detailed Implementation

[0062] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this specification, and not all embodiments. Based on the embodiments of this specification, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this specification.

[0063] This specification provides a method for updating navigation maps using an intelligent agent. It should be noted that the executing entity in these embodiments is generally a device with data processing capabilities, such as an intelligent agent or a computing device connected to the intelligent agent via communication means. Figure 1 This is a flowchart illustrating an intelligent agent navigation map update method provided in an embodiment of this specification, as shown below. Figure 1 As shown, the method mainly includes the following steps:

[0064] Step S101: Acquire the real-time video corresponding to the current scene, and extract multiple current keyframes from the real-time video to obtain the current keyframe group.

[0065] In some embodiments, before acquiring the real-time video corresponding to the current scene, the method further includes:

[0066] When the preset detection time requirement is met, the first detection command is triggered;

[0067] Based on the first detection command, move along the preset navigation route and execute the steps of claim 1.

[0068] Specifically, refer to Figure 3 Fixed tasks can be written into the workflow program of the intelligent agent. After starting it every day or at a fixed time every week (based on prior information about the magnitude of scene changes), routine detection on a fixed route will be started to detect scene changes and prompt whether map updates are needed.

[0069] This conventional detection mode is typically applied during relatively quiet periods when the agent is operating. For example, in supermarkets, this could be before and after opening to avoid excessive interference from dynamic obstacles. If static obstacles appear on a fixed data collection route, it indicates a significant change in the scene and will trigger a map update. If the agent is navigating autonomously, it will navigate around the obstacles and continue moving along the route.

[0070] In some embodiments, extracting multiple current keyframes from the real-time video to obtain a current keyframe group includes:

[0071] Generate video frame groups based on the real-time video;

[0072] The video frame group is clustered using a clustering algorithm to obtain multiple video frame categories;

[0073] Select one video frame from each of the aforementioned video frame categories as the current keyframe.

[0074] Step S102: Based on the agent's current position information and current posture information, select preset keyframes that match each of the current keyframes from the preset keyframe group corresponding to the preset map.

[0075] In this invention, position is represented by XYZ coordinates (indoor robots can generally be considered to move only in a fixed XY plane, so their position can be represented by only two-dimensional XY coordinates), and attitude, i.e. orientation, can be represented by quaternions, Euler angles, or rotation matrices.

[0076] Step S103: Calculate the matching degree between the current scene and the preset map based on the current keyframe and the corresponding preset keyframe;

[0077] In some implementations, calculating the matching degree between the current scene and the preset map based on the current keyframe and the corresponding preset keyframe includes the following steps:

[0078] Calculate the image similarity between each current keyframe and each corresponding preset keyframe;

[0079] If the image similarity is higher than the first preset threshold, then the current keyframe is determined to be a similar keyframe that is similar to the corresponding preset keyframe;

[0080] The proportion of similar keyframes in the current keyframe group is statistically analyzed, and this proportion is used as the degree of matching between the current scene and the preset map.

[0081] In some embodiments, updating the navigation map if the matching degree does not meet a preset condition includes:

[0082] If the ratio is not higher than a preset second threshold, then the navigation map is updated.

[0083] Specifically, the image similarity threshold is predefined as a first preset threshold α. s Similarity higher than α s The keyframe percentage threshold is the second preset threshold β. s The values ​​of the aforementioned thresholds are strongly correlated with the scenario and user needs, and those skilled in the art can set them according to actual needs, without being explicitly specified in the technical solution of this invention.

[0084] When the similarity is higher than α s The percentage of keyframes is higher than the threshold β s If the current scene is very similar to the one used when the preset map was built, it means the scene hasn't changed significantly, and the preset map still describes the scene information well, ensuring the agent's accurate localization and scene perception. Therefore, no update is needed. Conversely, if the scene is not as described, feedback should be sent to the agent, suggesting a map update.

[0085] Image similarity can be calculated using existing technologies, such as the DBoW (Bag of Visual Words) technique. In some embodiments, calculating the image similarity between each current keyframe and its corresponding preset keyframe includes:

[0086] Extract the visual feature information of the current keyframe and the preset keyframe, and perform clustering operation on the visual feature information to obtain multiple feature words;

[0087] Generate bag-of-words vectors based on the feature words, and calculate the distance between the bag-of-words vectors as the image similarity.

[0088] Step S104: If the matching degree does not meet the preset conditions, then the navigation map is updated.

[0089] In some embodiments, after the agent reaches the end of the preset navigation route, steps such as calculating the matching degree of the current keyframe and the preset keyframe group can be performed. Since the agent often wants to provide more information to the backend for optimization through loop closure detection, that is, the map end point and the starting point coincide, in order to ensure that a globally consistent trajectory and map are obtained.

[0090] In some embodiments, updating the navigation map includes the following steps:

[0091] If the image similarity is lower than the first preset threshold, then the current keyframe is determined to be a distinguishing keyframe;

[0092] The preset map is updated based on the distinguishing keyframes;

[0093] Obtain the difference position information and difference pose information corresponding to the difference keyframes;

[0094] The preset navigation route is updated based on the distinguished location information and the distinguished attitude information.

[0095] It's important to note that updating the map doesn't necessarily mean completely rebuilding it. Depending on the keyframes, if only a small portion of the scene changes significantly, updating only that part is sufficient. For large-scale scene map updates, this saves computational resources and improves efficiency.

[0096] It is worth noting that, in some embodiments, in addition to performing routine detection along a fixed route, the agent can also trigger detection during daily operation, which may specifically include the following steps:

[0097] Obtain the current location information of the intelligent agent, and calculate a location score based on the current location information;

[0098] When the positioning score is lower than the third preset threshold, a second detection command is triggered;

[0099] Based on the second detection command, the current keyframe corresponding to the current scene is collected;

[0100] Calculate the image similarity between the current keyframe and the preset keyframe, and find the matching keyframe in the preset keyframe group when the image similarity is greater than the preset first threshold.

[0101] If the matching keyframe is not found in the preset keyframe group, the navigation map is updated.

[0102] Intelligent agents typically employ localization evaluation mechanisms, aiming to reflect localization quality through localization scores. Poor localization quality can be caused by significant scene changes, inherently challenging scenes (e.g., repetitive or complex dynamic scenes), and insufficient performance of mapping and localization sensors and algorithms. However, it's important to note that localization scores are often only for reference and do not accurately reflect localization quality. Poor localization quality cannot be solely attributed to scene changes, and the degree of scene change is immeasurable. Therefore, localization scores alone cannot provide an accurate assessment of whether map updates are necessary.

[0103] In this invention, when the agent's localization score is low or its localization changes abruptly, visual keyframes describing the current scene information are collected. A keyframe matching the current keyframe is then searched from the loaded visual keyframe group corresponding to the known map. If a matching keyframe is found, the corresponding agent pose is extracted and compared with the agent's current localization. When the distance difference between the two is less than a fourth preset threshold δ... d A positive result indicates that the current positioning quality is acceptable and the scene has changed little, so a map update is not necessary. Conversely, a negative result suggests that the agent's positioning quality may be problematic, requiring relocalization or termination of the current navigation task. If no keyframe in the saved visual keyframe set matches the current keyframe, it indicates that the scene has indeed changed significantly, prompting a map update.

[0104] Furthermore, it may also include the following steps:

[0105] Based on the selection of a preset keyframe that matches the current keyframe from the preset keyframe group corresponding to the preset map;

[0106] Obtain the preset position information of the agent corresponding to the preset keyframe, and calculate the distance difference between the preset position information and the current position information;

[0107] If the distance difference is greater than a preset fourth threshold, a stop-forward command is triggered and the navigation route is updated.

[0108] This invention's technical solution uses visual sensor information and algorithms to determine scene changes without human intervention, exhibiting high intelligence and accuracy. It can identify areas with significant scene changes and map updates, allowing mapping to be performed only in those areas, thus ensuring the agent's efficiency. It innovatively incorporates visual elements and imposes no restrictions on the agent's mapping and localization algorithms, sensors, or map formats, resulting in higher intelligence and flexibility. Furthermore, this technology does not require the agent to be constantly running, saving computational resources.

[0109] This specification also provides an intelligent agent navigation map update method device, such as... Figure 2 As shown, the device includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to:

[0110] Acquire real-time video corresponding to the current scene, and extract multiple current keyframes from the real-time video to obtain the current keyframe group;

[0111] Based on the agent's current position and current posture information, select preset key frames that match each current key frame from the preset key frame group corresponding to the preset map.

[0112] Based on the current keyframe and the corresponding preset keyframe, calculate the degree of matching between the current scene and the preset map;

[0113] If the matching degree does not meet the preset conditions, the navigation map will be updated.

[0114] This specification also provides a non-volatile computer storage medium storing computer-executable instructions, wherein the computer-executable instructions are configured as follows:

[0115] Acquire real-time video corresponding to the current scene, and extract multiple current keyframes from the real-time video to obtain the current keyframe group;

[0116] Based on the agent's current position and current posture information, select preset key frames that match each current key frame from the preset key frame group corresponding to the preset map.

[0117] Based on the current keyframe and the corresponding preset keyframe, calculate the degree of matching between the current scene and the preset map;

[0118] If the matching degree does not meet the preset conditions, the navigation map will be updated.

[0119] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the embodiments of apparatus, devices, and non-volatile computer storage media are basically similar to the method embodiments, so the descriptions are relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0120] The foregoing has described specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0121] The devices, media, and methods provided in the embodiments of this specification are one-to-one correspondences. Therefore, the devices and media also have similar beneficial technical effects as their corresponding methods. Since the beneficial technical effects of the methods have been described in detail above, the beneficial technical effects of the devices and media will not be repeated here.

[0122] Those skilled in the art will understand that embodiments of this specification can be provided as methods, systems, or computer program products. Therefore, this specification may take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this specification may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0123] This specification is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this specification. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create a machine for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0124] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0125] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0126] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0127] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0128] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0129] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0130] The above description is merely one or more embodiments of this specification and is not intended to limit this specification. Various modifications and variations can be made to the one or more embodiments of this specification by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle of one or more embodiments of this specification should be included within the scope of the claims of this specification.

Claims

1. An agent navigation map updating method, characterized by, The method comprises: collecting real-time video corresponding to a current scene, and extracting a plurality of current key frames from the real-time video to obtain a current key frame group; selecting, based on current position information and current attitude information of an agent, preset key frames matched with each of the current key frames from a preset key frame group corresponding to a preset map; calculating a matching degree of the current scene and the preset map according to the current key frames and the corresponding preset key frames; if the matching degree does not satisfy a preset condition, updating a navigation map; the calculation of the matching degree of the current scene and the preset map according to the current key frames and the corresponding preset key frames comprises: calculating image similarity between each of the current key frames and the corresponding preset key frames; if the image similarity is higher than a first preset threshold, determining that the current key frame is a similar key frame similar to the corresponding preset key frame; statistically obtaining a proportion of the similar key frame in the current key frame group, and taking the proportion as the matching degree of the current scene and the preset map; if the proportion is not higher than a second preset threshold, updating the navigation map; before the collecting of the real-time video corresponding to the current scene, the method further comprises: triggering a first detection instruction when a preset detection time requirement is met; moving along a preset navigation route based on the first detection instruction; the method further comprises: obtaining current position information of the agent, and performing positioning score calculation based on the current position information; triggering a second detection instruction when the positioning score is lower than a third preset threshold; collecting a current key frame corresponding to a current scene based on the second detection instruction; calculating image similarity between the current key frame and the preset key frame, and searching for a matching key frame corresponding to the image similarity greater than the first preset threshold in the preset key frame group; if the matching key frame does not exist in the preset key frame group, updating the navigation map; the method further comprises: selecting, based on the preset key frame matched with the current key frame from the preset key frame group corresponding to the preset map; obtaining preset position information of the agent corresponding to the preset key frame, and calculating a distance difference between the preset position information and the current position information; if the distance difference is greater than a fourth preset threshold, triggering a stop advancing instruction and triggering an update of a navigation route. the updating of the navigation map comprises:

2. The method of claim 1, wherein, if the image similarity is lower than the first preset threshold, determining that the current key frame is a distinguished key frame; updating the preset map based on the distinguished key frame; obtaining distinguished position information and distinguished attitude information corresponding to the distinguished key frame; updating a preset navigation route according to the distinguished position information and the distinguished attitude information. the extraction of the plurality of current key frames from the real-time video to obtain the current key frame group comprises:

3. The method of claim 1, wherein, generating a video frame group based on the real-time video; performing clustering operation on the video frame group by using a clustering algorithm to obtain a plurality of video frame categories; ​ selecting one video frame as the current key frame from each of the video frame categories.

4. The method of claim 1, wherein, The calculation of the image similarity between the current key frame and the corresponding preset key frame comprises: extracting visual feature information of the current key frame and the preset key frame, and performing clustering operation on the visual feature information to obtain a plurality of feature words; generating a bag-of-words vector according to the feature words, and calculating the distance between the bag-of-words vectors as the image similarity.

5. An agent navigation map updating method device, characterized by, The device comprises: at least one processor; and a memory connected in communication with the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the steps of the method according to any one of claims 1-4.

6. A non-transitory computer storage medium storing computer-executable instructions that, when executed, cause a computer to perform: The computer executable instructions are configured to perform the steps of the method according to any one of claims 1-4.

Images