Event discovery method based on patrol robot and related device

Abstract

This application relates to the field of artificial intelligence technology and provides an event detection method based on patrol robots. The method includes: acquiring historical events detected in a target area, where each historical event includes event type, event detection time, and event detection location; constructing an event detection matrix corresponding to the historical events based on the event type, event detection time, and event detection location; constructing a route optimization model for a multi-target patrol robot, using the patrol robot's dwell time in each statistical area, patrol speed, and accessibility as constraints, and taking the maximization of the event detection probability and efficiency of a predetermined event as the objective function; outputting an optimized patrol route for at least one patrol robot through the multi-target patrol robot route optimization model, and controlling at least one patrol robot to execute the optimized patrol route for event detection.

Description

Technical Field

[0001] This application belongs to the field of artificial intelligence technology, and in particular relates to an event detection method and related equipment based on patrol robots. Background Technology

[0002] With the rapid development of robotics and artificial intelligence, patrol robots are increasingly being used in security, park management, and traffic inspection. Currently, traditional patrol robots typically employ fixed-route patrols, random patrols, or path planning based on simple rules. Fixed routes cannot adapt to the dynamic distribution of events in time and space, resulting in insufficient coverage of high-incidence areas while consuming excessive resources in low-incidence areas, leading to a low probability of event detection. Therefore, there is an urgent need for an event detection method that comprehensively considers the spatiotemporal distribution of events and has high event detection efficiency. This would address the problems of existing patrol robots, which rely on fixed-route patrols, random patrols, or path planning based on simple rules, failing to fully consider the patterns of event occurrence, being unable to adapt to the dynamic distribution of events in time and space, and struggling to achieve dynamic matching between patrol areas and event occurrence times, thus resulting in low event detection efficiency. Summary of the Invention

[0003] This application provides an event detection method based on patrol robots, which addresses the problems of low event detection efficiency caused by existing patrol robots using fixed-route patrols, random patrols, or path planning based on simple rules, which fail to fully consider the patterns of event occurrence, cannot adapt to the dynamic distribution of events in time and space, and struggle to achieve dynamic matching between patrols and event occurrence areas and time periods. By constructing an event detection matrix including time, space, and event feature values, and using the patrol robot's dwell time in each statistical area, patrol speed, and accessibility as constraints, and taking the maximization of the event detection probability and efficiency of predetermined events as the objective function, a multi-objective patrol robot route optimization model is constructed. This model outputs optimized patrol routes for at least one patrol robot, which is then controlled to execute these optimized routes for event detection. This effectively improves the event detection efficiency of patrol robots and solves the problems of low event detection efficiency caused by existing patrol robots using fixed-route patrols, random patrols, or path planning based on simple rules, which fail to fully consider the patterns of event occurrence, cannot adapt to the dynamic distribution of events in time and space, and struggle to achieve dynamic matching between patrols and event occurrence areas and time periods.

[0004] In a first aspect, embodiments of this application provide an event detection method based on a patrol robot, the method comprising the following steps:

[0005] Acquire historical discovery events in the target area, including event type, event discovery time, and event discovery location;

[0006] Based on the event type, the event discovery time, and the event discovery location, an event discovery matrix corresponding to the historical events is constructed. The event discovery matrix includes a time dimension, a spatial dimension, and event feature values. The spatial dimension consists of statistical regions divided by the latitude and longitude of the target area. The event feature values ​​are determined based on the number of historical events discovered and the event type.

[0007] Using the dwell time of the patrol robot in each statistical area, the patrol speed of the patrol robot, and the accessibility of the patrol robot as constraints, and taking the maximization of the event discovery probability and event discovery efficiency of the predetermined event as the objective function, a route optimization model for a multi-objective patrol robot is constructed.

[0008] The route optimization model of the multi-target patrol robot outputs an optimized patrol route for at least one of the patrol robots, and the at least one patrol robot is controlled to execute the optimized patrol route to detect events.

[0009] Optionally, constructing the event discovery matrix corresponding to the historically discovered events based on the event type, the event discovery time, and the event discovery location includes:

[0010] The timeline is divided into multiple time slices according to a preset time slice length, and the time dimension of the event discovery matrix is ​​constructed based on the multiple time slices.

[0011] The target area is divided into multiple statistical regions according to latitude and longitude, and the spatial dimension of the event discovery matrix is ​​constructed based on the multiple statistical regions.

[0012] Based on the number and type of historical discovery events that occurred in each statistical region within each time slice, event feature values ​​are added to each statistical region to obtain the event discovery matrix.

[0013] Optionally, dividing the target area into multiple statistical regions according to latitude and longitude, and constructing the spatial dimension of the event discovery matrix based on the multiple statistical regions, includes:

[0014] The target area is divided into multiple grid units according to a preset grid precision, and each grid unit serves as a statistical area; wherein, the grid precision is set according to the perception range and event localization accuracy of the patrol robot;

[0015] Based on the latitude and longitude information of the target area, the corresponding latitude and longitude range of the statistical area is associated to obtain the spatial dimension of the event discovery matrix.

[0016] Optionally, the step of adding event feature values ​​to each statistical region based on the number and type of historical discovery events occurring in each time slice within each statistical region includes:

[0017] For each statistical region, the number of discoveries corresponding to each event type is calculated for each time slice;

[0018] The event type and the number of discoveries are length-encoded to obtain the event feature corresponding to the event type for each statistical region within each time slice, and the length of the event feature is positively correlated with the number of discoveries.

[0019] Add the event characteristics to the corresponding statistics region.

[0020] Optionally, the process involves constructing a route optimization model for a multi-objective patrol robot, using constraints such as the robot's dwell time in each statistical area, its patrol speed, and its accessibility, with the objective function being to maximize the event detection probability and efficiency of a predetermined event. This model includes:

[0021] The dwell time constraint is defined as the time the patrol robot spends in any statistical area being between a preset minimum dwell time threshold and a preset maximum dwell time threshold.

[0022] The patrol speed constraint is that the distance the patrol robot moves between two adjacent statistical areas does not exceed the maximum distance the patrol robot can travel at a preset maximum patrol speed within the corresponding time interval.

[0023] The accessibility constraint is the path relationship between two adjacent statistical regions.

[0024] The first objective function is to maximize the event discovery probability, which is calculated based on the event feature values ​​of the statistical area traversed by the patrol robot within the corresponding time slice.

[0025] The second objective function is to maximize the event discovery efficiency. The event discovery efficiency is based on the ratio of the event feature value of the statistical area traversed by the patrol robot in the corresponding time slice and the patrol resource consumption. The patrol resource consumption includes at least one of the following: the number of robots participating in the patrol, the total patrol time of all patrol robots, or the total patrol path length of all patrol robots.

[0026] Based on the dwell time constraint, the patrol speed constraint, the reachability constraint, the first objective function, and the second objective function, a route optimization model for the multi-objective patrol robot is constructed.

[0027] Optionally, the step of outputting at least one optimized patrol route for the patrol robot through the route optimization model of the multi-target patrol robot includes:

[0028] The route optimization model is solved using a multi-objective evolutionary algorithm to generate a Pareto optimal solution set for the predetermined event, which contains multiple non-dominated solutions.

[0029] Obtain the number of currently available patrol robots or the preset resource budget constraints;

[0030] Based on the number of currently available patrol robots or a preset resource budget constraint, a solution is selected as the output from the Pareto optimal solution set, and the solution includes the optimized patrol route of at least one patrol robot.

[0031] Optionally, after controlling the at least one patrol robot to execute an optimized patrol route for event detection, the method further includes:

[0032] During the patrol, information on newly discovered events is acquired in real time, including the type of new event, the time of discovery, and the location of discovery.

[0033] The event discovery matrix is ​​incrementally updated based on the newly discovered event information, updating the event feature values ​​of the corresponding statistical region within the corresponding time slice;

[0034] Based on the updated event discovery matrix, the remaining unexecuted patrol routes are locally replanned to generate dynamically adjusted patrol routes;

[0035] The dynamically adjusted patrol route is then sent to the corresponding patrol robot for execution.

[0036] Secondly, embodiments of this application provide an event detection device based on a patrol robot, the event detection device based on a patrol robot comprising:

[0037] The acquisition module is used to acquire historical discovery events in the target area, including event type, event discovery time, and event discovery location;

[0038] The first construction module is used to construct an event discovery matrix corresponding to the historical events based on the event type, the event discovery time, and the event discovery location. The event discovery matrix includes a time dimension, a spatial dimension, and event feature values. The spatial dimension is composed of statistical regions divided by the latitude and longitude of the target area. The event feature values ​​are determined based on the number of historical events discovered and the event type.

[0039] The second construction module is used to construct a route optimization model for a multi-objective patrol robot, with the patrol robot's dwell time in each statistical area, patrol speed, and accessibility as constraints, and the objective function being to maximize the event detection probability and event detection efficiency of the predetermined event.

[0040] The event detection module is used to output an optimized patrol route for at least one of the patrol robots through the route optimization model of the multi-target patrol robot, and control the at least one patrol robot to execute the optimized patrol route to detect events.

[0041] Thirdly, embodiments of the present invention provide an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps in the event detection method based on a patrol robot provided in embodiments of the present invention.

[0042] Fourthly, embodiments of the present invention provide a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps in the event detection method based on a patrol robot provided in the embodiments of the present invention.

[0043] The above-mentioned solution of this application has the following beneficial effects: It acquires historical discovery events in the target area, including event type, event discovery time, and event discovery location; based on the event type, event discovery time, and event discovery location, it constructs an event discovery matrix corresponding to the historical discovery events; using the patrol robot's dwell time in each statistical area, the patrol robot's patrol speed, and the patrol robot's accessibility as constraints, and taking the maximization of the event discovery probability and event discovery efficiency as the objective function, it constructs a route optimization model for a multi-target patrol robot; through the multi-target patrol robot route optimization model, it outputs an optimized patrol route for at least one patrol robot, and controls at least one patrol robot to execute the optimized patrol route for event discovery. This invention solves the problem that existing patrol robots use fixed-route patrols, random patrols, or path planning methods based on simple rules, which do not fully consider the patterns of event occurrence, cannot adapt to the dynamic distribution of events in time and space, and are difficult to achieve dynamic matching between patrols and event occurrence areas and time periods, resulting in low event discovery efficiency.

[0044] Other beneficial effects of this application will be described in detail in the following detailed description section. Attached Figure Description

[0045] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0046] Figure 1 A flowchart illustrating an event detection method based on a patrol robot, provided as an embodiment of this application;

[0047] Figure 2 This is a schematic diagram of the structure of an event detection device based on a patrol robot according to an embodiment of this application;

[0048] Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0049] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0050] It should be understood that, when used in this application specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or a collection thereof.

[0051] It should also be understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0052] As used in this application specification and the appended claims, the term "if" may be interpreted, depending on the context, as "when," "once," "in response to determination," or "in response to detection." Similarly, the phrase "if determined" or "if detected [the described condition or event]" may be interpreted, depending on the context, as meaning "once determined," "in response to determination," "once detected [the described condition or event]," or "in response to detection [the described condition or event]."

[0053] Furthermore, in the description of this application and the appended claims, the terms "first," "second," "third," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0054] References to "one embodiment" or "some embodiments" as described in this specification mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.

[0055] like Figure 1 As shown, Figure 1 This is a flowchart of an event detection method based on a patrol robot provided by an embodiment of the present invention. The event detection method based on a patrol robot includes the following steps:

[0056] 101. Obtain historical discovery events in the target area.

[0057] In this embodiment of the invention, the above-described event discovery method based on patrol robots can be applied to a server, and a communication connection is established between the server and the patrol robot.

[0058] The target area mentioned above can be the area where the patrol robot needs to perform patrol and monitoring tasks, such as industrial parks, residential areas, city streets, etc.

[0059] The aforementioned historical events can be any events that have occurred in the target area in the past.

[0060] The aforementioned historical events include the event type, the time of discovery, and the location of discovery. The event type is used to distinguish different types of events, such as security-related, environmental, or equipment-related events. The security-related event type could be fighting, vandalism, etc.; the environmental-related event type could be garbage accumulation, illegal dumping, etc.; and the equipment-related event type could be missing manhole covers, malfunctioning streetlights, etc. The event time can be the specific time the event occurred, for example, event A occurred on March 27, 2025, at 14:23:06. The event location can be the specific location where the event occurred, for example, event A occurred at a certain intersection, etc.

[0061] The aforementioned historical events can be obtained from patrol robot history patrol records, monitoring equipment logs, manual event registration forms, etc.

[0062] 102. Based on event type, event discovery time, and event discovery location, construct an event discovery matrix corresponding to historical events.

[0063] In this embodiment of the invention, the above-mentioned construction may be a process of organizing an event discovery matrix corresponding to the historical discovery time according to the event type, event discovery time, and time discovery location.

[0064] The event discovery matrix described above includes time dimensions, spatial dimensions, and event feature values. This matrix is ​​used to quantitatively describe the temporal and spatial distribution patterns of historical events.

[0065] The aforementioned time dimension can be used to reflect the pattern of events occurring over time.

[0066] The aforementioned spatial dimensions consist of statistical regions defined by the latitude and longitude of the target area, used to reflect the spatial clustering characteristics of events. This latitude and longitude division can be achieved by using a latitude and longitude coordinate system as the basis for spatial division, cutting the target area into several grid units. The aforementioned statistical regions can be grid units obtained after dividing the target area using latitude and longitude.

[0067] The aforementioned event characteristic values ​​are determined based on the number of historical discoveries and the event type. These event characteristic values ​​quantify the importance of a specific statistical region within a given time slice in historical discoveries. The number of historical discoveries can be the total number of times a historical discovery was recorded. The number of discoveries reflects the frequency of event occurrence.

[0068] It should be noted that by constructing an event discovery matrix corresponding to historical events based on event type, event discovery time, and event discovery location, the spatiotemporal distribution patterns of historical events can be quantified into calculable event feature values, enabling patrol robots to accurately identify the area and time period of the event.

[0069] 103. Using the dwell time of the patrol robot in each statistical area, the patrol speed of the patrol robot, and the accessibility of the patrol robot as constraints, and taking the maximization of the event discovery probability and the event discovery efficiency of the predetermined event as the objective function, construct a route optimization model for a multi-objective patrol robot.

[0070] In this embodiment of the invention, the aforementioned dwell time can be the length of time that the patrol robot continuously stays in each statistical area.

[0071] The patrol speed mentioned above can be the actual speed at which the patrol robot moves.

[0072] The accessibility of the patrol robot can be interpreted as the ease with which it can move directly from one statistical area to another. This accessibility reflects the pathways between statistical areas, indicating which areas are connected and passable, and which are disconnected and inaccessible. Understandably, in reality, not all statistical areas can be directly accessed. For example, obstructions such as buildings, walls, rivers, or steep slopes may prevent the patrol robot from crossing in a straight line, requiring it to detour. Alternatively, only roads or designated passageways may be accessible between statistical areas, limiting the robot's movement to where roads exist.

[0073] The above constraints can be mandatory restrictions that must be met, ensuring that the generated route is physically feasible and effective.

[0074] It should be noted that in actual patrol missions, the patrol robots are not constrained in their dwell time. The patrol robots may stay in each area for only 1 second (which will not be able to detect events) or stay for 1 hour (which will waste resources); there is no constraint on patrol speed, and the planned routes are routes that the patrol robots cannot complete on time; there is no constraint on accessibility, and the planned routes are infeasible routes, such as passing through walls or flying over rivers.

[0075] The constraints include the dwell time of the patrol robot in each statistical area, which must be within a preset range; the patrol speed, which must limit the distance the robot can travel between two adjacent statistical areas to the maximum distance it can cover at its maximum patrol speed within the corresponding time interval; and accessibility, which requires a pathway between adjacent statistical areas for the robot to move from one area to another.

[0076] The probability of detecting the aforementioned predetermined event can be defined as the probability that a patrol robot, while patrolling along a pre-planned route, will successfully detect the predetermined event. The predetermined event can be a security incident, an environmental incident, an equipment incident, etc.

[0077] The aforementioned event detection efficiency refers to the ability of patrol robots to promptly and accurately identify the rate and effectiveness of events occurring while patrolling along pre-planned routes.

[0078] The objective function mentioned above can be a core mathematical expression used to measure the merits of a solution. The essence of the objective function is to transform the optimization objective (such as minimizing cost, maximizing profit, minimizing error, etc.) into a function of design variables or model parameters, and to find the optimal solution by maximizing or minimizing this function.

[0079] The objective function can be to maximize the probability of the patrol robot discovering the predetermined event and the event discovery efficiency. Furthermore, the objective function can be to maximize the rate and effectiveness of the patrol robot in timely and accurately identifying the event occurrence by minimizing the number of patrol robots and maximizing the total number of events discovered.

[0080] The aforementioned route optimization model scientifically plans the movement path using mathematical methods to achieve the goals of lowest cost, highest efficiency, or optimal service while satisfying various constraints. It can be implemented using a reinforcement learning algorithm to construct a multi-objective route optimization model for patrol robots, with constraints including the patrol robot's dwell time in each statistical area, patrol speed, and accessibility, and the objective function being maximizing the event discovery probability and efficiency of a predetermined event. Reinforcement learning (RL) is a machine learning method that learns optimal policies through agent-environment interaction. The core objective of this algorithm is to enable the patrol robot to rationally plan its patrol route, given the constraints of dwell time in each statistical area, patrol speed, and accessibility, and the objective function being maximizing the event discovery probability and efficiency of a predetermined event.

[0081] The route optimization model for the aforementioned multi-target patrol robot can be constructed by taking the dwell time of the patrol robot in each statistical area, the patrol speed of the patrol robot, and the accessibility of the patrol robot as constraints, and taking the maximization of the event discovery probability and the event discovery efficiency of the predetermined event as the objective function.

[0082] 104. Output the optimized patrol route of at least one patrol robot through the route optimization model of the multi-target patrol robot, and control at least one patrol robot to execute the optimized patrol route to detect events.

[0083] In this embodiment of the invention, after obtaining the route optimization model of the multi-target patrol robot, the optimized patrol route of at least one patrol robot can be output through the route optimization model of the multi-target patrol robot. The "at least one" can be one or more, that is, the number is greater than or equal to 1.

[0084] The optimized patrol route mentioned above can be calculated by a multi-objective route optimization model. Under the premise of satisfying all constraints, it is the patrol route of the patrol robot that maximizes the event detection probability and event detection efficiency of the predetermined event.

[0085] Furthermore, after obtaining the optimized patrol route of at least one patrol robot, control at least one patrol robot to execute the optimized patrol route for event detection.

[0086] The aforementioned event detection can occur when patrol robots, during patrol missions, use onboard sensors (such as cameras, lidar, and gas sensors) to perceive and analyze target areas to identify and handle predetermined events. These predetermined events can be security incidents, environmental incidents, or equipment-related incidents.

[0087] In this embodiment of the invention, an event discovery matrix comprising time, space, and event feature values ​​is constructed. Constraints include the patrol robot's dwell time in each statistical area, its patrol speed, and its accessibility. The objective function is to maximize the event discovery probability and efficiency of a predetermined event. A route optimization model for a multi-target patrol robot is then constructed. This model outputs optimized patrol routes for at least one patrol robot, which is then controlled to execute these optimized routes for event discovery. This effectively improves the event discovery efficiency of the patrol robot and solves the problem that existing patrol robots, which use fixed-route patrols, random patrols, or path planning based on simple rules, struggle to adapt to the dynamic distribution of events in time and space, resulting in low event discovery efficiency.

[0088] In this embodiment of the invention, historical discovery events in the target area are acquired, including event type, event discovery time, and event discovery location. Based on the event type, event discovery time, and event discovery location, an event discovery matrix corresponding to the historical discovery events is constructed. Using the patrol robot's dwell time in each statistical area, the patrol robot's patrol speed, and the patrol robot's accessibility as constraints, and with the objective function of maximizing the event discovery probability and event discovery efficiency for a predetermined event, a route optimization model for a multi-target patrol robot is constructed. The route optimization model for the multi-target patrol robot outputs an optimized patrol route for at least one patrol robot, and controls at least one patrol robot to execute the optimized patrol route for event discovery. This invention solves the problem that existing patrol robots use fixed-route patrols, random patrols, or path planning methods based on simple rules, which do not fully consider the patterns of event occurrence, cannot adapt to the dynamic distribution of events in time and space, and are difficult to achieve dynamic matching between patrols and event occurrence areas and time periods, resulting in low event discovery efficiency.

[0089] It is understood that in the specific implementation of this application, data such as event data, time data, location data, and route data are involved. When the embodiments in this application are applied to specific products or technologies, user permission or consent is required. Furthermore, the collection, use, and processing of related data, as well as the training, deployment, and invocation of algorithm models, must comply with the relevant laws, regulations, and standards of the relevant countries and regions.

[0090] Optionally, in the step of constructing the event discovery matrix corresponding to historically discovered events based on event type, event discovery time, and event discovery location, the timeline can be divided into multiple time slices according to a preset time slice length, and the time dimension of the event discovery matrix can be constructed based on multiple time slices; the target area can be divided into multiple statistical regions according to latitude and longitude, and the spatial dimension of the event discovery matrix can be constructed based on multiple statistical regions; according to the number of historically discovered events and event types that occurred in each statistical region within each time slice, event feature values ​​can be added to each statistical region to obtain the event discovery matrix.

[0091] In this embodiment of the invention, the aforementioned timeline may be a continuous timeline used to describe the chronological order of historical events.

[0092] The preset time slice length can be a pre-set time slice length. The time slice length can be the length of a continuous time axis cut into a series of continuous and non-overlapping time segments. The time slice length can be 10 minutes, 15 minutes, 30 minutes, etc.

[0093] The aforementioned time slices can be obtained by dividing the time axis according to a preset time slice length, resulting in each continuous and non-overlapping time interval.

[0094] The aforementioned time dimension can be a dimension used to represent the time change of an event.

[0095] The latitude and longitude mentioned above can be the geographical coordinates of any point on the Earth's surface, consisting of two values: longitude and latitude, used to accurately describe the location where the event was discovered.

[0096] The aforementioned statistical area can be a grid cell obtained by dividing the target area according to latitude and longitude.

[0097] The aforementioned spatial dimensions can be composed of statistical regions, used to reflect the spatial clustering characteristics of events.

[0098] The number of discoveries mentioned above could be the total number of historical discovery events. The number of discoveries reflects the frequency or intensity of historical events occurring within that spatiotemporal unit.

[0099] The above event types can be tags used to categorize historically discovered events. Event types can include security type, environmental type, equipment type, etc.

[0100] The aforementioned event characteristic values ​​can be determined based on the number and type of historical discoveries. These event characteristic values ​​are used to quantify the importance of a specific statistical region within a given time slice in historical discoveries.

[0101] The event discovery matrix described above includes time dimensions, spatial dimensions, and event feature values. The event discovery matrix is ​​used to quantitatively describe the temporal and spatial distribution patterns of historical events.

[0102] It should be noted that the timeline can be divided into multiple time slices according to a preset time slice length, and the time dimension of the event discovery matrix can be constructed based on multiple time slices. The target area can be divided into multiple statistical regions according to latitude and longitude, and the spatial dimension of the event discovery matrix can be constructed based on multiple statistical regions. According to the number and type of historical events discovered in each statistical region within each time slice, event feature values ​​are added to each statistical region to obtain the event discovery matrix. Through the event discovery matrix, the spatiotemporal distribution pattern of historical events can be quantified into calculable event feature values, enabling patrol robots to accurately identify the location and time period of events.

[0103] Optionally, in the step of dividing the target area into multiple statistical regions according to latitude and longitude, and constructing the spatial dimension of the event discovery matrix based on the multiple statistical regions, the target area can be divided into multiple grid units according to a preset grid precision, with each grid unit serving as a statistical region; the corresponding latitude and longitude range is associated with the statistical regions according to the latitude and longitude information of the target area to obtain the spatial dimension of the event discovery matrix.

[0104] In this embodiment of the invention, the preset grid precision can be a pre-set grid precision.

[0105] The grid accuracy can be set based on the patrol robot's perception range and event localization accuracy. The patrol robot's perception range refers to the maximum spatial area within which its sensors (such as cameras, lidar, thermal imagers, gas sensors, etc.) can identify or detect events during patrol. The perception range can be the effective detection radius radiating outwards from the robot, for example, 20 meters. The event localization accuracy refers to the spatial accuracy achievable by the positioning system (such as GPS, BeiDou, etc.) used to determine the location of the event.

[0106] It should be noted that grid precision can refer to the size of each grid cell. Higher grid precision results in smaller cells, finer divisions, and more statistical regions. Conversely, lower precision results in larger cells, coarser divisions, and fewer statistical regions. For example, if the preset grid precision is 5m × 5m, the target area can be divided into many small 5m × 5m grids, with each grid cell representing a statistical region.

[0107] The latitude and longitude information of the target area mentioned above can be the geographic coordinate boundary data of the target area where the patrol robot needs to perform its tasks, including the longitude range and latitude range of the target area.

[0108] Furthermore, the statistical area can be associated with a corresponding latitude and longitude range based on the latitude and longitude information of the target area.

[0109] The above association can be a process of establishing a correspondence between each statistical region and its corresponding latitude and longitude range based on the latitude and longitude information of the target region, forming a bidirectional queryable mapping relationship.

[0110] The spatial dimension of the event discovery matrix can be obtained by associating the statistical region with the corresponding latitude and longitude range based on the latitude and longitude information of the target region.

[0111] It should be noted that the target area can be divided into multiple grid units according to a preset grid precision. Each grid unit is a statistical region, and the corresponding latitude and longitude range is associated with the statistical region based on the latitude and longitude information of the target area to obtain the spatial dimension of the event discovery matrix. Each statistical region has accurate latitude and longitude information to facilitate event mapping.

[0112] Optionally, in the step of adding event feature values ​​to each statistical region based on the number and type of historical discovery events that occurred in each time slice for each statistical region, the number of discoveries corresponding to each event type can be calculated for each statistical region for each time slice; the event type and the number of discoveries can be length-encoded to obtain the event features for each statistical region in each time slice; and the event features can be added to the corresponding statistical region.

[0113] In this embodiment of the invention, the above-mentioned number of discoveries can be calculated for each statistical region, based on each time slice, to determine the total number of historical discoveries for each event type. The number of discoveries reflects the density of historical events occurring within that time slice.

[0114] The length encoding described above can be a method that integrates event type and discovery quantity into a structured feature vector. Specifically, length encoding can convert the discovery quantity corresponding to each event type into an encoding segment of length equal to the discovery quantity, and then concatenate the encoding segments of all event types in sequence to form a structured feature vector. The core idea of ​​length encoding is to use the length of the encoding sequence to reflect the discovery quantity, and to use the content or segments of the encoding sequence to distinguish event types.

[0115] Each of the above statistical regions can be obtained by length-encoding the event type and the number of discoveries within each time slice. Each statistical region specifies which statistical region is in which time slice within each time slice. For example, statistical region A is from 0:00 to 0:30 in the morning; statistical region B is from 14:00 to 15:00 in the afternoon; statistical region C is from 20:00 to 20:30 in the evening, etc.

[0116] In this embodiment of the invention, the event features corresponding to the event type have a length that is positively correlated with the number of discoveries. This positive correlation can be a relationship where the length of the event feature increases (decreases) as the number of discoveries increases (decreases), meaning the two variables change in the same direction.

[0117] The aforementioned event features can be structured feature vectors obtained by length encoding the event type and the number of discoveries. The event features reflect the distribution intensity and composition characteristics of events in each statistical region within each time slice.

[0118] The above addition can be a process of adding event features to the corresponding statistical area.

[0119] Optionally, in the step of constructing a route optimization model for a multi-objective patrol robot, with constraints such as the dwell time of the patrol robot in each statistical area, the patrol speed of the patrol robot, and the accessibility of the patrol robot, and with the objective function of maximizing the event discovery probability and event discovery efficiency of a predetermined event, the following constraints can be used: the dwell time of the patrol robot in any statistical area is between a preset minimum dwell time threshold and a preset maximum dwell time threshold; the patrol speed constraint is that the moving distance of the patrol robot between two adjacent statistical areas does not exceed the maximum distance that the patrol robot can travel at a preset maximum patrol speed within the corresponding time interval; the accessibility constraint is the path relationship between two adjacent statistical areas; the first objective function is to maximize the event discovery probability, which is calculated based on the event feature values ​​of the statistical areas traversed by the patrol robot within the corresponding time slice; the second objective function is to maximize the event discovery efficiency, which is based on the ratio of the event feature values ​​of the statistical areas traversed by the patrol robot within the corresponding time slice to the patrol resource consumption; based on the dwell time constraint, patrol speed constraint, accessibility constraint, first objective function, and second objective function, a route optimization model for a multi-objective patrol robot is constructed.

[0120] In this embodiment of the invention, the aforementioned preset minimum dwell time threshold can be a pre-set minimum dwell time threshold for the patrol robot. The aforementioned maximum dwell time threshold can be a pre-set maximum dwell time threshold for the patrol robot.

[0121] By setting the dwell time of the patrol robot in any statistical area to be between the preset minimum dwell time threshold and the maximum dwell time threshold, we can ensure that the patrol robot has enough time to complete effective perception, avoid missing events due to too short a dwell time, and prevent the patrol robot from staying in the same area for too long, thus avoiding the waste of patrol resources.

[0122] The aforementioned preset maximum patrol speed can be the pre-set maximum patrol speed of the patrol robot.

[0123] By setting the patrol speed constraint so that the patrol robot's movement distance between two adjacent statistical areas does not exceed the maximum distance that the patrol robot can travel at the preset maximum patrol speed within the corresponding time interval, it can be ensured that the patrol robot can complete the movement between areas within a specified time, thus guaranteeing the spatiotemporal continuity and feasibility of the patrol route.

[0124] Using the accessibility relationship between two adjacent statistical areas as an accessibility constraint can avoid planning infeasible paths that cross obstacles such as buildings and fences, thus ensuring the physical feasibility of patrol routes.

[0125] The first objective function mentioned above can be optimized by maximizing the event discovery probability, which is calculated based on the event feature values ​​of the statistical areas traversed by the patrol robot within the corresponding time slice.

[0126] The second objective function mentioned above can be to maximize the event detection efficiency, which is the ratio of the event feature value of the statistical area traversed by the patrol robot in the corresponding time slice to the patrol resource consumption.

[0127] The ratio of patrol resource consumption mentioned above includes at least one of the following: the number of robots participating in the patrol, the total patrol time of all patrol robots, or the total patrol path length of all patrol robots.

[0128] The above construction can be a process of integrating dwell time constraints, patrol speed constraints, reachability constraints, the first objective function, and the second objective function into a route optimization model for a multi-objective patrol robot.

[0129] In one possible embodiment, the above-mentioned maximization of event discovery probability (first objective function) is expressed as:

[0130]

[0131] in, For the number of patrol robots, For the first The number of statistical areas traversed by each patrol robot's path. For statistical area In time slice Event feature values ​​within, For the first The patrol robot arrived at the Time for each statistical region The preset time decay coefficient;

[0132] The above-mentioned maximizing event discovery efficiency (second objective function) is expressed as:

[0133]

[0134] in, For the first The total length of the patrol path of each patrol robot , These are the preset weighting coefficients.

[0135] The aforementioned route optimization model can be used to scientifically plan movement paths using mathematical methods, in order to achieve the goals of lowest cost, highest efficiency, or optimal service while satisfying various constraints.

[0136] It should be noted that reinforcement learning algorithms can be used to integrate dwell time constraints, patrol speed constraints, reachability constraints, the first objective function, and the second objective function into a route optimization model for a multi-objective patrol robot. The reinforcement learning (RL) algorithm is a machine learning method that learns optimal policies through interaction between an agent and its environment. The core objective of this algorithm is to enable the patrol robot to rationally plan its patrol route based on dwell time constraints, patrol speed constraints, reachability constraints, the first objective function, and the second objective function.

[0137] Optionally, in the step of outputting the optimized patrol route of at least one patrol robot through the route optimization model of the multi-objective patrol robot, a multi-objective evolutionary algorithm can be used to solve the route optimization model to generate a Pareto optimal solution set containing multiple non-dominated solutions for a predetermined event; obtain the number of currently available patrol robots or a preset resource budget constraint; and select a solution from the Pareto optimal solution set as the output based on the number of currently available patrol robots or the preset resource budget constraint, wherein the solution includes the optimized patrol route of at least one patrol robot.

[0138] In this embodiment of the invention, the aforementioned Multi-Objective Evolutionary Algorithm (MOEA) is a class of global optimization algorithms based on biological evolutionary mechanisms, used to solve complex optimization problems involving multiple conflicting objectives. The core objective of the MOEA is to find a set of compromise solutions known as the Pareto optimal solution set. The MOEA can be NSGA-II, MOEA / D, etc. NSGA-II (Non-dominated Sorting Genetic Algorithm II) is a classic multi-objective optimization algorithm designed to solve problems involving the simultaneous optimization of multiple conflicting objectives. Its goal is not to find a single optimal solution, but rather to obtain a set of Pareto optimal solutions. MOEA / D (Multi-objective Evolutionary Algorithm based on Decomposition) is an evolutionary algorithm framework that transforms a multi-objective optimization problem into multiple single-objective subproblems, which are solved in parallel to approximate the Pareto optimal front.

[0139] The above solution can be obtained by using a multi-objective evolutionary algorithm to solve the route optimization model and find the Pareto optimal solution set that satisfies the constraints and objective function.

[0140] The above non-dominated solution can be understood as a solution that is superior to or equal to it in all optimization objectives, and strictly superior to it in at least one objective.

[0141] The Pareto optimal solution set mentioned above can be the set of all non-dominated solutions. The Pareto optimal solution set describes the set of all possible optimal trade-off states among multiple conflicting objectives. A Pareto optimal solution satisfies the condition that no objective can be further improved without worsening other objectives.

[0142] The number of patrol robots currently available can be the number of patrol robots that can be used to perform patrol operations during the current time period of the patrol mission.

[0143] The aforementioned preset resource budget constraint can be a pre-set constraint on the upper limit of patrol resource consumption, which can be used to constrain the output scheme of the route optimization model from exceeding the upper limit of the resource budget.

[0144] The above solution includes an optimized patrol route for at least one patrol robot. The optimized patrol route can be an executable patrol path scheme that achieves an optimal trade-off between the predetermined event detection probability and the event detection efficiency, under the premise of satisfying the constraints and objective function, by solving a multi-objective optimization model using a multi-objective evolutionary algorithm, and based on the number of currently available patrol robots or a preset resource budget constraint.

[0145] It should be noted that this invention can select a solution from the Pareto optimal solution set as the output based on the real-time resource status, ensuring that the output route is the optimal solution under the current available resource conditions.

[0146] Optionally, after controlling at least one patrol robot to perform the step of optimizing patrol routes for event discovery, new event information can be acquired in real time during the patrol process; the event discovery matrix can be incrementally updated based on the new event information, and the event feature values ​​of the corresponding statistical regions within the corresponding time slices can be updated; based on the updated event discovery matrix, the remaining unexecuted patrol routes can be locally replanned to generate dynamically adjusted patrol routes; and the dynamically adjusted patrol routes can be sent to the corresponding patrol robots for execution.

[0147] In this embodiment of the invention, during the patrol process, the patrol robot can acquire new event information in real time through its sensors, cameras, and other means. This newly discovered event information includes the new event type, the time of discovery, and the location of the discovery.

[0148] The aforementioned incremental update can be a process of incrementally updating the event discovery matrix based on newly discovered event information to update the event feature values ​​of the corresponding statistical region within the corresponding time slice. Incremental update is a strategy that only updates the data or content that has changed, rather than retransmitting or reprocessing all data. The core of incremental update is to only process the newly added, modified, or deleted parts, thereby significantly improving efficiency and saving bandwidth and storage resources.

[0149] The updated event discovery matrix can be obtained by incrementally updating the event discovery matrix based on newly discovered event information, thereby updating the event feature values ​​of the corresponding statistical region within the corresponding time slice.

[0150] The aforementioned unexecuted remaining patrol routes can be patrol paths that the patrol robot has not traversed or completed at the current moment. Unexecuted remaining patrol routes can be patrol paths that were not executed in the optimized patrol routes executed by the patrol robot from the current moment to the end of the task, and can be untraversed paths, unvisited statistical areas, uncompleted stops, etc.

[0151] The aforementioned local replanning can be a process of replanning the remaining unexecuted patrol routes based on the updated event discovery matrix, starting from the patrol robot's current position. It can be understood that local replanning can be performed from the current moment to the task's end time; the scope of replanning can include statistical areas that the robot has not yet visited.

[0152] The dynamically adjusted patrol routes mentioned above can be new patrol paths generated by locally replanning the remaining unexecuted patrol routes based on the updated event discovery matrix, and used to replace the unexecuted patrols in the optimized patrol routes.

[0153] Furthermore, the dynamically adjusted patrol routes can be distributed to the corresponding patrol robots for execution.

[0154] In this embodiment of the invention, the invention can incrementally update the event discovery matrix based on new event information, update the event feature values ​​of the corresponding statistical region within the corresponding time slice, and perform local replanning of the remaining unexecuted patrol routes through the updated event discovery matrix. Local replanning can be for replanning from the current time to the task end time, and the scope of replanning can be statistical regions that the robot has not yet visited, etc., which effectively reduces the consumption of computing resources and improves the efficiency of event discovery.

[0155] like Figure 2 As shown, this embodiment of the invention provides an event detection device based on a patrol robot, which includes:

[0156] The acquisition module 201 is used to acquire historical discovery events in the target area, wherein the historical events include event type, event discovery time and event discovery location;

[0157] The first construction module 202 is used to construct an event discovery matrix corresponding to the historical events based on the event type, the event discovery time, and the event discovery location. The event discovery matrix includes a time dimension, a spatial dimension, and event feature values. The spatial dimension is composed of statistical regions divided by the latitude and longitude of the target area. The event feature values ​​are determined based on the number of historical events discovered and the event type.

[0158] The second construction module 203 is used to construct a route optimization model for a multi-objective patrol robot, with the patrol robot's dwell time in each statistical area, patrol speed, and accessibility as constraints, and the objective function being to maximize the event discovery probability and event discovery efficiency of the predetermined event.

[0159] The event detection module 204 is used to output an optimized patrol route for at least one of the patrol robots through the route optimization model of the multi-target patrol robot, and control the at least one patrol robot to execute the optimized patrol route to detect events.

[0160] Optionally, the first construction module 202 is further configured to divide the time axis into multiple time slices according to a preset time slice length, and construct the time dimension of the event discovery matrix based on the multiple time slices; divide the target area into multiple statistical regions according to latitude and longitude, and construct the spatial dimension of the event discovery matrix based on the multiple statistical regions; add event feature values ​​to each statistical region according to the number of historical discovery events and event types that occurred in each time slice in each statistical region, to obtain the event discovery matrix.

[0161] Optionally, the first construction module 202 is further configured to divide the target area into multiple grid units according to a preset grid precision, with each grid unit serving as a statistical area; wherein the grid precision is set according to the perception range and event positioning accuracy of the patrol robot; and to associate the corresponding latitude and longitude range of the statistical area with the latitude and longitude information of the target area to obtain the spatial dimension of the event discovery matrix.

[0162] Optionally, the first construction module 202 is further configured to calculate the number of discoveries corresponding to each event type for each statistical region and each time slice; encode the event type and the number of discoveries by length to obtain the event feature corresponding to the event type in each statistical region within each time slice, wherein the length of the event feature is positively correlated with the number of discoveries; and add the event feature to the corresponding statistical region.

[0163] Optionally, the second construction module 203 is further configured to: constrain the patrol robot's dwell time in any statistical area to be between a preset minimum dwell time threshold and a preset maximum dwell time threshold; constrain the patrol robot's movement distance between two adjacent statistical areas to not exceed the maximum distance the patrol robot can travel at a preset maximum patrol speed within the corresponding time interval; constrain the accessibility relationship between two adjacent statistical areas; use maximizing the event discovery probability as the first objective function, the event discovery probability being calculated based on the event feature values ​​of the statistical areas traversed by the patrol robot within the corresponding time slice; and use maximizing the event discovery efficiency as the second objective function, the event discovery efficiency being based on the ratio of the event feature values ​​of the statistical areas traversed by the patrol robot within the corresponding time slice to the patrol resource consumption, the patrol resource consumption including at least one of the number of robots participating in the patrol, the total patrol time of all patrol robots, or the total patrol path length of all patrol robots; and construct a route optimization model for a multi-objective patrol robot based on the dwell time constraint, the patrol speed constraint, the accessibility constraint, the first objective function, and the second objective function.

[0164] Optionally, the event detection module 204 is further configured to use a multi-objective evolutionary algorithm to solve the route optimization model, generate a Pareto optimal solution set containing multiple non-dominated solutions for the predetermined event; obtain the number of currently available patrol robots or a preset resource budget constraint; and select a solution from the Pareto optimal solution set as output based on the number of currently available patrol robots or the preset resource budget constraint, wherein the solution includes an optimized patrol route for at least one patrol robot.

[0165] Optionally, the device is further configured to acquire newly discovered event information in real time during the patrol process, the newly discovered event information including new event type, new event discovery time, and new event discovery location; incrementally update the event discovery matrix based on the newly discovered event information, updating the event feature values ​​of the corresponding statistical area within the corresponding time slice; perform local replanning on the remaining unexecuted patrol routes based on the updated event discovery matrix, generating dynamically adjusted patrol routes; and distribute the dynamically adjusted patrol routes to the corresponding patrol robots for execution.

[0166] like Figure 3 As shown, this embodiment of the invention also provides an electronic device, including a processor, which can execute any of the above-described event detection methods based on patrol robots.

[0167] Specifically, it includes a processor 301 and a memory 302, as well as a computer program stored in the memory 302 and capable of running on the processor 301, which executes an event detection method based on a patrol robot, wherein:

[0168] The processor 301 executes the calculator program based on the event detection method of the patrol robot stored in the memory 302, and performs the following steps:

[0169] Acquire historical discovery events in the target area, including event type, event discovery time, and event discovery location;

[0170] Based on the event type, the event discovery time, and the event discovery location, an event discovery matrix corresponding to the historical events is constructed. The event discovery matrix includes a time dimension, a spatial dimension, and event feature values. The spatial dimension consists of statistical regions divided by the latitude and longitude of the target area. The event feature values ​​are determined based on the number of historical events discovered and the event type.

[0171] Using the dwell time of the patrol robot in each statistical area, the patrol speed of the patrol robot, and the accessibility of the patrol robot as constraints, and taking the maximization of the event discovery probability and event discovery efficiency of the predetermined event as the objective function, a route optimization model for a multi-objective patrol robot is constructed.

[0172] The route optimization model of the multi-target patrol robot outputs an optimized patrol route for at least one of the patrol robots, and the at least one patrol robot is controlled to execute the optimized patrol route to detect events.

[0173] Optionally, the process executed by processor 301 to construct the event discovery matrix corresponding to the historically discovered events based on the event type, the event discovery time, and the event discovery location includes:

[0174] The timeline is divided into multiple time slices according to a preset time slice length, and the time dimension of the event discovery matrix is ​​constructed based on the multiple time slices.

[0175] The target area is divided into multiple statistical regions according to latitude and longitude, and the spatial dimension of the event discovery matrix is ​​constructed based on the multiple statistical regions.

[0176] Based on the number and type of historical discovery events that occurred in each statistical region within each time slice, event feature values ​​are added to each statistical region to obtain the event discovery matrix.

[0177] Optionally, the step of processor 301 executing the division of the target area into multiple statistical regions according to latitude and longitude, and constructing the spatial dimension of the event discovery matrix based on the multiple statistical regions, includes:

[0178] The target area is divided into multiple grid units according to a preset grid precision, and each grid unit serves as a statistical area; wherein, the grid precision is set according to the perception range and event localization accuracy of the patrol robot;

[0179] Based on the latitude and longitude information of the target area, the corresponding latitude and longitude range of the statistical area is associated to obtain the spatial dimension of the event discovery matrix.

[0180] Optionally, the process executed by processor 301 to add event feature values ​​to each statistical region based on the number and type of historical discovery events occurring in each time slice, including:

[0181] For each statistical region, the number of discoveries corresponding to each event type is calculated for each time slice;

[0182] The event type and the number of discoveries are length-encoded to obtain the event feature corresponding to the event type for each statistical region within each time slice, and the length of the event feature is positively correlated with the number of discoveries.

[0183] Add the event characteristics to the corresponding statistics region.

[0184] Optionally, the processor 301 executes a route optimization model for a multi-objective patrol robot, using the patrol robot's dwell time in each statistical area, patrol speed, and reachability as constraints, and taking the maximization of the event detection probability and event detection efficiency as the objective function. This model includes:

[0185] The dwell time constraint is defined as the time the patrol robot spends in any statistical area being between a preset minimum dwell time threshold and a preset maximum dwell time threshold.

[0186] The patrol speed constraint is that the distance the patrol robot moves between two adjacent statistical areas does not exceed the maximum distance the patrol robot can travel at a preset maximum patrol speed within the corresponding time interval.

[0187] The accessibility constraint is the path relationship between two adjacent statistical regions.

[0188] The first objective function is to maximize the event discovery probability, which is calculated based on the event feature values ​​of the statistical area traversed by the patrol robot within the corresponding time slice.

[0189] The second objective function is to maximize the event discovery efficiency. The event discovery efficiency is based on the ratio of the event feature value of the statistical area traversed by the patrol robot in the corresponding time slice and the patrol resource consumption. The patrol resource consumption includes at least one of the following: the number of robots participating in the patrol, the total patrol time of all patrol robots, or the total patrol path length of all patrol robots.

[0190] Based on the dwell time constraint, the patrol speed constraint, the reachability constraint, the first objective function, and the second objective function, a route optimization model for the multi-objective patrol robot is constructed.

[0191] Optionally, the processor 301 executes the route optimization model of the multi-target patrol robot, outputting at least one optimized patrol route for the patrol robot, including:

[0192] The route optimization model is solved using a multi-objective evolutionary algorithm to generate a Pareto optimal solution set for the predetermined event, which contains multiple non-dominated solutions.

[0193] Obtain the number of currently available patrol robots or the preset resource budget constraints;

[0194] Based on the number of currently available patrol robots or a preset resource budget constraint, a solution is selected as the output from the Pareto optimal solution set, and the solution includes the optimized patrol route of at least one patrol robot.

[0195] Optionally, after controlling the at least one patrol robot to perform optimized patrol routes for event detection, the method executed by the processor 301 further includes:

[0196] During the patrol, information on newly discovered events is acquired in real time, including the type of new event, the time of discovery, and the location of discovery.

[0197] The event discovery matrix is ​​incrementally updated based on the newly discovered event information, updating the event feature values ​​of the corresponding statistical region within the corresponding time slice;

[0198] Based on the updated event discovery matrix, the remaining unexecuted patrol routes are locally replanned to generate dynamically adjusted patrol routes;

[0199] The dynamically adjusted patrol route is then sent to the corresponding patrol robot for execution.

[0200] This invention also provides a computer-readable storage medium storing a computer program. When executed by a processor, the computer program implements the various processes of the event detection method based on a patrol robot provided in this invention and achieves the same technical effect. To avoid repetition, it will not be described again here.

[0201] The above description is the preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principles described in this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. An event discovery method based on a patrol robot, characterized by, The method includes the following steps: Acquire historical discovery events in the target area, including event type, event discovery time, and event discovery location; Based on the event type, the event discovery time, and the event discovery location, an event discovery matrix corresponding to the historical events is constructed. The event discovery matrix includes a time dimension, a spatial dimension, and event feature values. The spatial dimension consists of statistical regions divided by the latitude and longitude of the target area. The event feature values ​​are determined based on the number of historical events discovered and the event type. The event feature values ​​are used to quantify the importance of a certain statistical region in a historical event within a certain time slice. Using the dwell time of the patrol robot in each statistical area, the patrol speed of the patrol robot, and the accessibility of the patrol robot as constraints, and taking the maximization of the event discovery probability and event discovery efficiency of the predetermined event as the objective function, a route optimization model for a multi-objective patrol robot is constructed; including: the dwell time constraint is that the dwell time of the patrol robot in any statistical area is between a preset minimum dwell time threshold and a maximum dwell time threshold. The patrol speed constraint is that the distance the patrol robot moves between two adjacent statistical areas does not exceed the maximum distance the patrol robot can travel at a preset maximum patrol speed within the corresponding time interval. The accessibility constraint is the path relationship between two adjacent statistical regions. The first objective function is to maximize the event detection probability, which is calculated based on the event feature values ​​of the statistical areas traversed by the patrol robot within the corresponding time slices. The first objective function is expressed as: in, For the number of patrol robots, For the first The number of statistical areas traversed by each patrol robot's path. For statistical area In time slice Event feature values ​​within, For the first The patrol robot arrived at the Time for each statistical region The preset time decay coefficient; The second objective function is to maximize event detection efficiency. This efficiency is based on the ratio of event characteristic values ​​within the corresponding time slice of the statistical area traversed by the patrol robot to the patrol resource consumption. The patrol resource consumption includes at least one of the following: the number of robots participating in the patrol, the total patrol time of all patrol robots, or the total patrol path length of all patrol robots. The second objective function is expressed as: in, For the first The total length of the patrol path of each patrol robot , These are preset weighting coefficients; Based on the dwell time constraint, the patrol speed constraint, the reachability constraint, the first objective function, and the second objective function, a route optimization model for the multi-objective patrol robot is constructed. The route optimization model of the multi-target patrol robot outputs an optimized patrol route for at least one of the patrol robots, and the at least one patrol robot is controlled to execute the optimized patrol route to detect events.

2. The patrol robot-based event discovery method of claim 1, wherein, The process of constructing an event discovery matrix corresponding to the historically discovered events based on the event type, the event discovery time, and the event discovery location includes: The timeline is divided into multiple time slices according to a preset time slice length, and the time dimension of the event discovery matrix is ​​constructed based on the multiple time slices. The target area is divided into multiple statistical regions according to latitude and longitude, and the spatial dimension of the event discovery matrix is ​​constructed based on the multiple statistical regions. Based on the number and type of historical discovery events that occurred in each statistical region within each time slice, event feature values ​​are added to each statistical region to obtain the event discovery matrix.

3. The patrol robot-based event discovery method of claim 2, wherein, The step of dividing the target area into multiple statistical regions according to latitude and longitude, and constructing the spatial dimension of the event discovery matrix based on the multiple statistical regions, includes: The target area is divided into multiple grid units according to a preset grid precision, and each grid unit serves as a statistical area; wherein, the grid precision is set according to the perception range and event localization accuracy of the patrol robot; Based on the latitude and longitude information of the target area, the corresponding latitude and longitude range of the statistical area is associated to obtain the spatial dimension of the event discovery matrix.

4. The patrol robot-based event discovery method of claim 2, wherein, The step of adding event feature values ​​to each statistical region based on the number and type of historical discovery events occurring in each time slice, including: For each statistical region, the number of discoveries corresponding to each event type is calculated for each time slice; The event type and the number of discoveries are length-encoded to obtain the event feature corresponding to the event type for each statistical region within each time slice, and the length of the event feature is positively correlated with the number of discoveries. Add the event characteristics to the corresponding statistics region.

5. The event detection method based on a patrol robot as described in any one of claims 1 to 4, characterized in that, The method of outputting at least one optimized patrol route for the patrol robot through the route optimization model of the multi-target patrol robot includes: The route optimization model is solved using a multi-objective evolutionary algorithm to generate a Pareto optimal solution set for the predetermined event, which contains multiple non-dominated solutions. Obtain the number of currently available patrol robots or the preset resource budget constraints; Based on the number of currently available patrol robots or a preset resource budget constraint, a solution is selected as the output from the Pareto optimal solution set, and the solution includes the optimized patrol route of at least one patrol robot.

6. The event detection method based on a patrol robot as described in any one of claims 1 to 4, characterized in that, After controlling the at least one patrol robot to execute an optimized patrol route for event detection, the method further includes: During the patrol, information on newly discovered events is acquired in real time, including the type of new event, the time of discovery, and the location of discovery. The event discovery matrix is ​​incrementally updated based on the newly discovered event information, updating the event feature values ​​of the corresponding statistical region within the corresponding time slice; Based on the updated event discovery matrix, the remaining unexecuted patrol routes are locally replanned to generate dynamically adjusted patrol routes; The dynamically adjusted patrol route is then sent to the corresponding patrol robot for execution.

7. An event detection device based on a patrol robot, characterized in that, The event detection method based on a patrol robot according to any one of claims 1 to 6, wherein the event detection device based on the patrol robot comprises: The acquisition module is used to acquire historical discovery events in the target area, including event type, event discovery time, and event discovery location. The first construction module is used to construct an event discovery matrix corresponding to the historical events based on the event type, the event discovery time, and the event discovery location. The event discovery matrix includes a time dimension, a spatial dimension, and event feature values. The spatial dimension is composed of statistical regions divided by the latitude and longitude of the target area. The event feature values ​​are determined based on the number of historical events discovered and the event type. The second construction module is used to construct a route optimization model for a multi-objective patrol robot, with the patrol robot's dwell time in each statistical area, patrol speed, and accessibility as constraints, and the objective function being to maximize the event detection probability and event detection efficiency of the predetermined event. The event detection module is used to output an optimized patrol route for at least one of the patrol robots through the route optimization model of the multi-target patrol robot, and control the at least one patrol robot to execute the optimized patrol route to detect events.

8. An electronic device, comprising: include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the steps of the event detection method based on a patrol robot as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of the event detection method based on a patrol robot as described in any one of claims 1 to 6.

Images