A method and system for cable theft prevention based on a drone

Abstract

The application provides a cable anti-theft method and system based on a UAV, wherein the method mainly collects environmental and threat data through sensors such as light, human flow, cameras and communication equipment, constructs a virtual 'digital twin' causal model in a processor, then simulates the effects of different anti-theft methods by using a 'counterfactual reasoning' engine, calculates the possible losses caused by each strategy, automatically selects the scheme with the minimum loss for execution, finally updates the model parameters according to the results of each real theft event, and continuously optimizes the strategy, so that the cable theft caused by repeated trial and error in reality is avoided, and various new theft methods can be flexibly coped with.

Description

Technical Field

[0001] This invention relates to the field of anti-theft technology, and in particular to a method and system for cable anti-theft based on drones. Background Technology

[0002] Cables, being high-value metal products, are easily targeted by thieves. Due to their typically long length, drones are often used for anti-theft monitoring. Traditional drone anti-theft solutions rely on fixed rules like triggering an alarm upon detecting a person, such as always sounding a horn. This approach fails to dynamically combine various countermeasures based on theft methods (cutting / digging) and environmental factors (human traffic / lighting). Furthermore, optimizing anti-theft strategies relies on human experience and lessons learned, resulting in a lengthy and rudimentary process. Using pure machine learning for anti-theft strategy determination requires extensive trial and error in the real world; each failure translates into actual cable theft and power grid damage, a costly prospect. Moreover, pure machine learning is prone to overfitting to historical data patterns, potentially rendering the model completely ineffective in the face of new theft methods or environmental combinations. Summary of the Invention

[0003] One objective of this invention is to provide a method and system for cable anti-theft based on drones. This method and system clearly distinguishes environmental, threat, response, and outcome variables by constructing a causal graph. Furthermore, it utilizes a counterfactual reasoning engine to automatically remove confounding biases during decision-making, thereby accurately identifying truly effective interventions and avoiding misjudgments caused by confounding. In addition, this invention uses a policy network to output the optimal drone action combination that minimizes expected loss based on the environmental and threat context, enabling rapid optimization of the anti-theft strategy. This invention uses theft results, including cable communication information and power outage information, as optimization reference data for the anti-theft strategy, enabling the implementation of the anti-theft strategy that minimizes losses.

[0004] One of the objectives of this invention is to provide a method and system for cable anti-theft based on drones. The method and system construct a causal graph in a digital twin space as a high-fidelity counterfactual simulator. Using a real event, a massive amount of counterfactual experience can be generated through noise inversion, avoiding high-cost online trial and error and achieving reinforcement learning with extremely low sample complexity.

[0005] One objective of this invention is to provide a method and system for cable anti-theft based on drones. This method and system allow for backtracking of each decision to a causal graph path: due to the current environment and maintenance threats, counterfactual comparison reveals that the resulting loss is minimized. This provides operators with clear and transparent decision-making basis, facilitating review and intervention. Furthermore, the causal model and policy network evolve synchronously with each real event, and the counterfactual success rate autonomously converges to an asymptotic optimum, achieving continuous self-reinforcement without human intervention.

[0006] To achieve at least one of the above-mentioned objectives, the present invention provides a method for preventing cable theft based on unmanned aerial vehicles (UAVs), the method comprising: A digital twin space near the cable and the drone is pre-constructed, and variable features are constructed near the cable and the drone based on physical sensors. A causal graph is then constructed in the digital twin space. The variable features are injected into the causal graph, and the counterfactual reasoning engine is used to perform fact acquisition and noise inversion on the causal graph; Construct a drone anti-theft strategy, perform counterfactual intervention deduction on the drone anti-theft strategy based on the counterfactual reasoning engine, and calculate the expected loss under each strategy in the drone anti-theft strategy; The counterfactual reasoning engine selects the response variable of the corresponding drone anti-theft strategy with the smallest expected loss based on the expected loss of each strategy and outputs it to the strategy execution layer to execute the corresponding strategy. Based on each completed real event, the structural equation parameters in the causal graph are updated using maximum likelihood household registration estimation.

[0007] According to a preferred embodiment of the present invention, the variable feature construction method includes: acquiring data from a light sensor, a temperature sensor, and a pedestrian flow sensor respectively, constructing light characteristics, temperature characteristics, and pedestrian flow characteristics of the cable location, and constructing dynamic environmental variables; using an image sensor and an edge computing unit to detect and identify intruders at the cable location in real time, and analyzing the number of thieves, the location of the theft, and the method of theft, constructing threat variables through feature construction; acquiring anti-theft strategy data of drone countermeasures equipment, and constructing response variables based on the anti-theft strategy data.

[0008] According to another preferred embodiment of the present invention, the method for constructing the variable features further includes: real-time detection and identification of cable theft results based on image sensors and edge computing units, wherein the theft results include cable communication information and power outage information, wherein the cable theft results are binarized results, and obtaining the load loss, power outage range and communication interruption range at the corresponding location of the cable, and performing feature transformation on the cable theft results, the load loss, power outage range and communication interruption range at the corresponding location of the cable to obtain the result variables.

[0009] According to another preferred embodiment of the present invention, the context of the causal graph is constructed based on the environmental variables and threat variables. The context of the causal graph is used as the state of reinforcement learning, the response variables based on the anti-theft strategy data are used as the actions of reinforcement learning, and the loss score, which is a weighted sum of the success of theft and the impact on the surrounding power grid, is used as the reward function of reinforcement learning. The probability of selecting the corresponding response variable action in the context state is used as the strategy of reinforcement learning, and the minimization of the loss score is further calculated as the objective function of reinforcement learning.

[0010] According to another preferred embodiment of the present invention, the counterfactual intervention inference method includes: acquiring a real cable theft event and recording the factual trajectory of the real cable theft event; inferring the latent noise driving the event based on the factual trajectory and the causal model structural equation; obtaining the posterior distribution of the latent noise through noise inversion; obtaining the counterfactual result prediction by performing counterfactual query on the latent noise; and calculating the counterfactual reward.

[0011] According to another preferred embodiment of the present invention, after obtaining real event samples, the structural equation of the causal graph in the digital twin space is updated and optimized using the real event samples. The parameters of the structural equation are updated and optimized using maximum likelihood estimation, and new real event sample variable features are injected into the updated and optimized causal graph. The new expected loss under each strategy in the drone anti-theft strategy is calculated using the counterfactual reasoning engine.

[0012] To achieve at least one of the above-mentioned objectives, the present invention further provides a drone-based cable anti-theft system, wherein the system performs the above-mentioned drone-based cable anti-theft method.

[0013] The present invention further provides a computer-readable storage medium storing a computer program, which is executed by a processor to implement the above-described method for preventing cable theft based on a drone. Attached Figure Description

[0014] Figure 1 The diagram shown is a flowchart of a cable anti-theft method based on a drone according to the present invention. Detailed Implementation

[0015] The following description is intended to disclose the present invention and enable those skilled in the art to implement it. The preferred embodiments described below are merely examples, and other obvious variations will occur to those skilled in the art. The basic principles of the invention defined in the following description can be applied to other embodiments, modifications, improvements, equivalents, and other technical solutions that do not depart from the spirit and scope of the invention.

[0016] It is understood that the term "a" should be understood as "at least one" or "one or more," that is, in one embodiment, the number of an element can be one, while in another embodiment, the number of the element can be multiple, and the term "a" should not be understood as a limitation on the number.

[0017] Please refer to Figure 1 This invention discloses a method and system for preventing cable theft based on unmanned aerial vehicles (UAVs), the method comprising: S01. Pre-construct a digital twin space near the cable and the drone, and construct variable features near the cable and the drone based on physical sensors, and construct a causal graph in the digital twin space; S02. Inject the variable features into the causal graph, and use the counterfactual reasoning engine to perform fact acquisition and noise inversion on the causal graph; S03. Construct a drone anti-theft strategy, perform counterfactual intervention deduction on the drone anti-theft strategy based on the counterfactual reasoning engine, and calculate the expected loss under each strategy in the drone anti-theft strategy. S04. The counterfactual reasoning engine selects the response variable of the corresponding drone anti-theft strategy with the smallest expected loss according to the expected loss of each strategy and outputs it to the strategy execution layer to execute the corresponding strategy. S05. Based on each completed real event, update the structural equation parameters in the causal graph using maximum likelihood estimation.

[0018] Specifically, the present invention constructs the following physical space based on the unmanned aerial vehicle (UAV) system: Environmental perception layer: This layer includes light sensors, temperature sensors, and pedestrian flow sensors. The pedestrian flow sensor can be implemented using sensors including, but not limited to, meter-wave radar. These sensors continuously collect light, temperature, and pedestrian flow characteristics at the cable's location to construct dynamic environmental variables such as light, temperature, and pedestrian flow variables. These light, temperature, and pedestrian flow variables characterize the background conditions under which theft occurs, influencing both the thief's behavioral choices and the reliability of sensor identification and the effectiveness of response equipment.

[0019] Threat Identification Layer: This layer primarily consists of high-resolution image sensors and edge computing units. It can detect intruders in real time and, using target detection and behavior analysis models, output instance values ​​of threat variables. These instance values ​​include: the number of thieves, the location of the stolen cable, and the method of cable theft. The theft method can include, but is not limited to, cutting, digging, and climbing towers. Response variables are constructed based on the identification results of the threat identification layer and anti-theft strategy data. The threat identification layer can also acquire cable communication information and cable power outage information at the corresponding location. This information is used to identify the execution result of the current anti-theft strategy and to optimize the strategy.

[0020] Response Execution Layer: Based on the UAV system, the response execution layer integrates various countermeasure devices, including automatically taking off and landing UAVs, high-brightness strobe lights, directional voice broadcasting devices, and projectile jamming emitters. All devices can communicate with the upper-layer system in real time via a low-latency IoT network, receiving commands and transmitting execution status.

[0021] Furthermore, this invention defines an environmental variable as E, a threat variable as T, a response variable as R, and an outcome variable as Y. The outcome variable includes a binary function representing whether the theft was successful (1 for successful theft, 0 for unsuccessful theft) and the impact of the cable theft on the surrounding power system. The success of the theft and its impact on the surrounding power system are quantified as a loss score as the outcome variable Y. The cable theft outcome includes cable communication information and cable power outage information, which can identify whether there is a cable communication interruption or power outage. The impact on the surrounding power system can be quantified, including but not limited to load loss, power outage range level, and communication interruption range, etc., and the corresponding outcome variable Y is obtained by weighted summation after normalization calculation.

[0022] Furthermore, the reinforcement learning implementation method based on causal graphs of the present invention includes: when the corresponding sensor in the physical space inputs event evidence E=e and T=t of environmental and threat variables, the system then selects a response variable R, and performs a corresponding UAV decision based on the response variable R to obtain the observed result variable Y and the loss score L, wherein the present invention configures the reinforcement learning context state s composed of the environmental and threat variables: s=(E,T); the response variable R is used as the reinforcement learning action a=R; the probability of selecting the corresponding action a in the context state s is used as the policy network for reinforcement learning, and finally reinforcement learning is performed with minimizing the expected loss as the objective function, wherein the objective function is: ; in It is the structural equation of the outcome variable in a causal graph, i.e. ;in This is exogenous random noise. It represents negative expected loss; only E in the formula represents the expectation operator, which means taking the average value over the random variable. This indicates that state s follows a distribution in the real environment. , This indicates that action 1 is sampled from the current policy. This indicates that the same loss function was used previously.

[0023] It is worth mentioning that this invention records the following factual trajectory based on real events that have already occurred: ; in This represents a record of a real event, where e is a specific observed value of the environmental variable and t is a specific observed value of the threat variable. For the actual response action performed, This indicates the actual action value executed in the corresponding response action. This represents the actual observed values ​​of the result variable. This represents the actual loss value. The posterior sampling formula for the corresponding latent noise is: ;in Indicates exogenous random noise A specific sample value, where - in the formula indicates that it follows a distribution of ... Let represent the posterior probability distribution of the noise variable given observed evidence. The deterministic inversion function is: ,in This is the inverse function of the causal graph structure equation. Under a deterministic causal mechanism, the noise can be directly deduced from the input and output. The counterfactual result prediction formula is: ,in This represents any candidate action that differs from the actual action. This indicates that in a counterfactual world, if action is taken... The resulting value. Furthermore, the counterfactual reward is: ;in This corresponds to the counterfactual reward value.

[0024] It should be noted that, since this invention employs reinforcement learning under a causal model, the total interaction sample complexity of this reinforcement learning framework is far lower than that of model-free reinforcement learning. High-fidelity counterfactual simulations can be completed based on a small number of real samples. The structural equation parameters of the causal model are updated and optimized using maximum likelihood estimation with the use of real samples. The specific formula is as follows: ;in The trainable parameters represent the structural equations of the causal model. This indicates finding the parameter value that maximizes the expression. This represents the summation over all real historical times i; Indicates that under given parameters The following observation results The log-likelihood; , , and Let represent the observed values ​​of the outcome, environment, threat, and response of the i-th real event, respectively.

[0025] The processes described in the flowcharts above, as disclosed in the embodiments of this invention, can be implemented as computer software programs. The embodiments disclosed in this invention include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication component, and / or installed from a removable medium. When the computer program is executed by a central processing unit (CPU), the methods of this application are not limited to the aforementioned functions. It should be noted that the computer-readable medium described above in this application can be a computer-readable signal medium or a computer-readable storage medium, or any combination of the two. Computer-readable storage media can be, for example, but not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatuses, or devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wire segments, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this application, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in connection with an instruction execution system, apparatus, or device. In this application, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium can also be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on a computer-readable medium may be transmitted using any suitable medium, including but not limited to: wireless segments, wire segments, optical fibers, RF, etc., or any suitable combination thereof.

[0026] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0027] Those skilled in the art should understand that the embodiments of the present invention described above and shown in the accompanying drawings are merely examples and do not limit the present invention. The purpose of the present invention has been fully and effectively achieved. The functions and structural principles of the present invention have been shown and explained in the embodiments. Without departing from the stated principles, the implementation of the present invention may have any variations or modifications.

Claims

1. A method for preventing cable theft based on unmanned aerial vehicles (UAVs), characterized in that, The method includes: A digital twin space near the cable and the drone is pre-constructed, and variable features are constructed near the cable and the drone based on physical sensors. A causal graph is then constructed in the digital twin space. The variable features are injected into the causal graph, and the counterfactual reasoning engine is used to perform fact acquisition and noise inversion on the causal graph; Construct a drone anti-theft strategy, perform counterfactual intervention deduction on the drone anti-theft strategy based on the counterfactual reasoning engine, and calculate the expected loss under each strategy in the drone anti-theft strategy; The counterfactual reasoning engine selects the response variable of the corresponding drone anti-theft strategy with the smallest expected loss based on the expected loss of each strategy and outputs it to the strategy execution layer to execute the corresponding strategy. Based on each completed real event, the structural equation parameters in the causal graph are updated using maximum likelihood estimation.

2. The cable anti-theft method based on unmanned aerial vehicles according to claim 1, characterized in that, The variable feature construction method includes: acquiring data from light sensors, temperature sensors, and pedestrian flow sensors respectively, constructing light characteristics, temperature characteristics, and pedestrian flow characteristics of the cable location, and constructing dynamic environmental variables; using image sensors and edge computing units to detect and identify intruders at the cable location in real time, and analyzing the number of thieves, the location of the theft, and the theft method, constructing threat variables through feature construction; acquiring anti-theft strategy data of drone countermeasure equipment, and constructing response variables based on the anti-theft strategy data.

3. The cable anti-theft method based on unmanned aerial vehicles according to claim 1, characterized in that, The method for constructing the variable features further includes: real-time detection and identification of cable theft results based on image sensors and edge computing units, wherein the theft results include cable communication information and power outage information, wherein the cable theft results are binarized results, and the load loss, power outage range and communication interruption range at the corresponding location of the cable are obtained, and the cable theft results, the load loss, power outage range and communication interruption range at the corresponding location of the cable are transformed to obtain the result variables.

4. A method for preventing cable theft based on unmanned aerial vehicles (UAVs) according to claim 2, characterized in that, The context of the causal graph is constructed based on the environmental and threat variables. The context of the causal graph is used as the state of reinforcement learning. The response variables based on the anti-theft strategy data are used as the actions of reinforcement learning. The reward function of reinforcement learning is a weighted sum of the loss score based on whether the theft is successful and the impact on the surrounding power grid. The probability of selecting the corresponding response variable action in the context state is used as the strategy of reinforcement learning. The objective function of reinforcement learning is to further calculate and minimize the loss score.

5. A method for preventing cable theft based on unmanned aerial vehicles (UAVs) according to claim 1, characterized in that, The counterfactual intervention inference method includes: acquiring real cable theft events and recording the factual trajectory of the real cable theft events; inferring the latent noise driving the events based on the factual trajectory and the causal model structural equation; obtaining the posterior distribution of the latent noise through noise inversion; obtaining the counterfactual result prediction by counterfactual querying the latent noise; and calculating the counterfactual reward.

6. A method for preventing cable theft based on unmanned aerial vehicles (UAVs) according to claim 1, characterized in that, After obtaining real event samples, the parameters of the structural equation of the causal graph in the digital twin space are updated and optimized using the real event samples. The parameters of the structural equation are updated and optimized using maximum likelihood estimation. New real event sample variable features are injected into the updated and optimized causal graph. The new expected loss under each strategy in the drone anti-theft strategy is calculated using the counterfactual inference engine.

7. A cable anti-theft system based on unmanned aerial vehicles (UAVs), characterized in that, The system implements a cable anti-theft method based on a drone as described in any one of claims 1-8.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that is executed by a processor to implement a cable anti-theft method based on any one of claims 1-8.

Images