A park access device cooperative control method and system based on multi-source authentication

Abstract

The present application relates to the technical field of adaptive control, and more particularly, to a park access equipment cooperative control method and system based on multi-source authentication, comprising: obtaining multi-source authentication original data and environmental quality parameters of each access node; constructing a dynamic trust weight score of any authentication source in each access node at the current time, which is used to quantify the output quality and stability of each authentication source under different environmental quality parameters at the current time; and calculating a multi-source fusion weighted consistency index of each access node at the current time. The present application uses the authentication link degradation risk value as an adaptive adjustment lever, dynamically reconstructs the global reward function of the multi-agent algorithm, realizes the strategy self-switching of prioritizing security bottom line under high-risk situation and maximizing the overall access cooperative efficiency under low-risk situation, and ensures that the park maintains the best balance between strict and smooth management and control under various uncertain environmental disturbances.

Description

Technical Field

[0001] This invention relates to the field of adaptive control technology. More specifically, this invention relates to a collaborative control method and system for park access equipment based on multi-source authentication. Background Technology

[0002] With the rapid development of smart parks, the intelligent collaborative management of park access equipment has become a crucial element. Multi-source authentication technology, as an important security control method, is commonly used in turnstiles, access control systems, and gate barriers. This technology typically integrates data from multiple authentication sources, such as facial recognition, RFID access cards, mobile Bluetooth, or near-field communication, aiming to improve the accuracy of personnel identification. Based on this, combining multi-agent collaborative control algorithms for global scheduling and control of various access devices is currently the mainstream technical solution for achieving efficient and secure passage for massive numbers of people in parks.

[0003] In real-world smart park access control scenarios, access control devices are typically deployed in complex physical environments. The actual output quality of each authentication source is highly susceptible to dynamic influences from the external natural environment and internal hardware status. Due to environmental synergy, this impact is often sudden and can easily lead to synchronized performance degradation of multiple authentication devices at the same access node within the same time period, even posing a risk of systemic collapse of the entire authentication chain.

[0004] In handling multi-source authentication scenarios in industrial parks, existing technologies typically assign fixed decision weights to each authentication source and employ fixed-structure global reward functions. When using this approach to solve access control problems in complex environments, authentication sources that have already experienced performance degradation may still participate in system decision-making with relatively high fixed weights in the event of severe weather or concentrated equipment aging, thus outputting erroneous pseudo-consistency signals to the system. Simultaneously, the fixed control strategy cannot perceive the overall risk level, which can easily lead to the system's inability to adaptively adjust its strategy during actual operation. This can result in either incorrect passage due to misjudgment in severe environments or prolonged congestion due to numerous authentication failures, ultimately causing a severe imbalance between security and access efficiency in the park under changing conditions. Summary of the Invention

[0005] This invention provides a collaborative control method and system for park access equipment based on multi-source authentication, aiming to solve the problem that fixed control strategies in related technologies cannot perceive the overall risk level. This can easily lead to the system being unable to adaptively adjust the strategy in actual operation, either causing incorrect passage due to misjudgment in harsh environments, or causing long-term congestion due to a large number of authentication failures. Ultimately, this results in actual business anomalies in the park where security and access efficiency are severely imbalanced in a variable environment.

[0006] In a first aspect, the present invention provides a collaborative control method for park access equipment based on multi-source authentication, comprising: acquiring multi-source authentication raw data and environmental quality parameters of each access node; constructing a dynamic credibility weight score of any authentication source in each access node at the current moment, used to quantify the output quality and stability of each authentication source under different environmental quality parameters at the current moment; calculating a multi-source fusion weighted consistency index of each access node at the current moment, used to characterize the true reliability of the access node after the fusion of authentication decisions at the current moment; constructing an authentication link degradation risk value of each access node at the current moment, wherein the authentication link degradation risk value is negatively correlated with the multi-source fusion weighted consistency index at the current moment, and also negatively correlated with the minimum dynamic credibility weight score among all currently activated authentication sources; using the authentication link degradation risk value as an adaptive lever to reconstruct the global reward function of the multi-agent near-end policy optimization algorithm, and outputting the optimal control action of each authentication source. This method constructs dynamic trusted weights and combines them with link degradation risk values ​​as the reward function of the adaptive leverage reconstruction algorithm. It can accurately perceive the output quality fluctuations of each authentication source and flexibly adjust the control strategy in complex application scenarios such as sudden changes in severe weather or concentrated aging of park equipment. It effectively avoids the congestion caused by misjudgment and excessive rejection due to environmental interference, and achieves a dynamic balance between safety and traffic efficiency in the park under changing environments.

[0007] Furthermore, the global reward function of the multi-agent near-end policy optimization algorithm is reconstructed, including: the current global reward function includes the authentication fusion sub-reward and the device collaboration sub-reward; the average of the authentication link degradation risk values ​​of all access nodes at the current time is used as the weight of the authentication fusion sub-reward, and the complement of the average of the authentication link degradation risk values ​​of all access nodes at the current time is used as the weight of the device collaboration sub-reward, so as to achieve adaptive reconstruction of the global reward. This method decouples the global reward function into authentication fusion and device collaboration and performs adaptive dynamic weighting based on the risk average, enabling the access control system to flexibly switch operating modes according to the current overall emergency situation of the park. In high-risk scenarios such as typhoons, rainstorms, or large-scale network jitter, priority is given to ensuring authentication security level, while in low-risk scenarios with stable operation, the passage efficiency of the devices is maximized.

[0008] Furthermore, the calculation method for the authentication fusion sub-reward at the current moment includes: calculating the sum of the products of the multi-source fusion weighted consistency index of all passing nodes at the current moment and the correctness flag value of the fusion decision for processing the current passage request, and multiplying the sum of the products by the reciprocal of the total number of all passing nodes at the current moment to obtain the authentication fusion sub-reward at the current moment. This method can more accurately transmit real reward and punishment signals to the collaborative control model, avoiding the false consistency misleading caused by multiple degraded devices simultaneously outputting erroneous decisions in complex park business scenarios, thereby significantly improving the decision reliability of the underlying control equipment when handling real and false anomalies.

[0009] Furthermore, the method for obtaining the correctness flag value of the fusion decision in processing the current access request includes: automatically generating truth labels in the simulation training environment based on preset personnel access permission rules; when the current fusion decision of the access node is consistent with the truth label, the correctness flag value is 1, otherwise it is 0. Using the preset rules in the simulation training environment to generate truth labels for comparison allows the algorithm model to fully learn compliant access and rejection logic before production deployment. This ensures that in the actual application of pedestrian flow control in the park, the final decisions issued by each gate or barrier strictly comply with the security access control standards of the enterprise park.

[0010] Furthermore, the calculation method for the device collaboration sub-reward at the current moment includes: calculating the mean of the normalized throughput of all passing nodes at the current moment, multiplying the mean by the latency equalization coefficient at the current moment to obtain the device collaboration sub-reward at the current moment. The latency equalization coefficient is negatively correlated with the mean of the normalized standard deviation of the response latency of all passing nodes. Combining the throughput of each node with the latency equalization coefficient, which takes into account the differences in global road network response, enables the entire park's device network to proactively balance resource scheduling and avoid localized congestion and latency spikes when facing massive morning and evening peak traffic pressure, significantly improving the macro-level system smoothness and collaboration capabilities of large parks.

[0011] Furthermore, the throughput of each access node is obtained, including: the throughput of each access node refers to the ratio of the number of people actually allowed to pass through the node within the current time window to the theoretical maximum number of people allowed to pass through the node. Statistically calculating the ratio of the actual number of people allowed to pass through to the theoretical maximum number of people allowed to pass through within a specific time window provides an extremely objective and granular quantitative indicator of system load for macro-level coordinated control, helping the entire access control architecture to intuitively grasp and respond to the actual pressure status of each physical defense zone during peak periods of population flow.

[0012] Furthermore, the calculation method for the dynamic credibility weight score is as follows: Real-time confidence level, historical authentication accuracy, variance of the confidence level sequence, and environmental quality factor are obtained for each authentication source at multiple access nodes within the park. The environmental quality factor reflects the output quality of the authentication source, while the variance of the confidence level sequence reflects the stability of the authentication source. The dynamic credibility weight score is positively correlated with both real-time confidence level and environmental quality factor, and negatively correlated with the variance of the confidence level sequence. By comprehensively incorporating environmental quality changes such as light intensity, temperature, and humidity, as well as the historical sequence fluctuation characteristics of the equipment, a joint evaluation can be conducted. This allows for the precise detection of hardware performance degradation caused by sudden darkness due to external heavy rain or internal antenna aging, preventing security incidents such as accidental door openings caused by low-quality distorted signals continuously interfering with overall decision-making.

[0013] Furthermore, the formula for calculating the multi-source fusion weighted consistency index is as follows: In the formula, Indicates the passage node exist The multi-source fusion weighted consistency index at any given time; For passage nodes The total number of currently active authentication sources; For passage nodes No. Personal authentication rights are based on The decision fusion weight at any given time is the normalized value of the dynamic credibility weight score; For passage nodes No. Personal authentication rights are based on The binary decision value at time t; For passage nodes The binary decision value corresponds to the mean of the binary decision values ​​of each authentication source. When the real-time confidence of each authentication source is greater than or equal to the decision threshold, its binary decision value is 1; otherwise, the binary decision value is 0. In extreme scenarios where multiple access control authentication devices of different types collectively output erroneous commands due to degradation caused by harsh environments, it can strongly suppress the superficial consistency illusion brought about by low-quality degraded devices, fundamentally eliminating meaningless bandwagon voting, and further consolidating the effectiveness of the core security isolation of the park.

[0014] Furthermore, the optimal control actions of each authentication source are output, including: a multi-agent near-end policy optimization algorithm based on the reconstructed global reward function, with the input being the local observation vector of each access node. This local observation vector is composed of the dynamic trust weight score of each authentication source of this access node, the multi-source fusion weighted consistency index, the authentication link degradation risk value, and the current access request feature vector concatenated. The optimal control actions of each authentication source in the current state are output, including allowing access, rejecting access, or initiating secondary verification.

[0015] In a second aspect, a collaborative control system for park access equipment based on multi-source authentication is also provided, including a processor and a memory, wherein the memory stores a computer program, and the processor executes the computer program to implement the collaborative control method for park access equipment based on multi-source authentication as described above.

[0016] Beneficial Effects: Addressing the technical pain points of existing access control systems that employ fixed identification weights and fixed global reward functions, which are prone to false consistency errors due to synchronized equipment degradation when facing sudden severe weather or widespread equipment aging, leading to security vulnerabilities and frequent congestion, this paper constructs a dynamic and reliable weight and consistency index for the authentication source by integrating multi-dimensional dynamic data such as environmental quality factors. This accurately isolates signal interference from degraded equipment. Simultaneously, utilizing the authentication link degradation risk value as an adaptive adjustment lever, the global reward function of the multi-agent algorithm is dynamically reconstructed. This achieves a strategy of automatically switching between prioritizing security in high-risk situations and maximizing overall access control efficiency in low-risk situations, ensuring the optimal balance between strict and smooth management within the park under various uncertain environmental interferences. Attached Figure Description

[0017] Figure 1 This is a schematic flowchart illustrating a control method according to an embodiment of the present invention; Figure 2 This is a schematic diagram illustrating the balance between safety and efficiency under dynamic risks according to an embodiment of the present invention; Figure 3 This is a schematic diagram illustrating the comparison of dynamic trust weight scores of authentication sources according to an embodiment of the present invention. Detailed Implementation

[0018] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0019] like Figure 1 As shown, S101: Multi-source authentication data acquisition and preprocessing.

[0020] In this embodiment, each physical location unit in the smart park where access control equipment is deployed is called an access node, which corresponds to one access node in the multi-agent near-end policy optimization algorithm. At each access node, various types of authentication devices are locally deployed. To achieve comprehensive collection of authentication data, high-definition binocular cameras deployed at turnstiles, access control systems, and barriers are used to acquire facial images, and the confidence score of the facial recognition source is output in real time through the facial recognition model of the edge server. Simultaneously, RFID card readers are used to collect the access control card reading signal strength and legality verification results, and the mobile terminal management platform receives the mobile terminal credential authentication results via Bluetooth or near-field communication protocols. These two types of data constitute the real-time confidence input of the corresponding authentication source. Furthermore, the device management system of the access nodes continuously records the historical authentication accuracy sequence of each access device, the actual number of people granted access, and the response delay from the arrival of the authentication request to the output of the control action at each access node.

[0021] Simultaneously, to assess the impact of the environment on authentication, light sensors, temperature and humidity sensors, and radio frequency signal strength meters need to be deployed at access nodes to collect environmental quality parameters. These parameters specifically include light intensity, temperature and humidity data, and Bluetooth or near-field communication signal strength. Each access control board also integrates an embedded device health monitoring module, which can periodically read the CPU utilization, communication module heartbeat response status, and sensor self-test return codes to output device health status markers. All collected data is uploaded to the park's central computing platform after being tagged with a unified timestamp.

[0022] On the central computing platform, the collected raw data needs to be preprocessed. Specifically, with The system uses milliseconds as the synchronization window to align the time of each data stream and fills in missing timestamps using linear interpolation. Subsequently, it performs min-max normalization on the confidence scores, signal strength, and environmental quality parameters of each authentication source. For missing values ​​caused by brief device downtime, the nearest non-missing value from the same source is used for filling. Finally, the system can extract the real-time confidence of each authentication source at each access node, the historical accuracy within the sliding window, the variance of the confidence sequence, the environmental quality factor of each authentication source, the response latency of each access node, and the throughput of each access node. The throughput of each access node refers to the ratio of the number of people actually allowed to pass through the node within the current time window to the theoretical maximum number of people allowed to pass through the node.

[0023] S102: Construct a dynamic trustworthy weight score for the authentication source.

[0024] In multi-source authentication access control scenarios within parks, the output quality of different types of authentication sources is highly susceptible to environmental factors and their own state, resulting in significant dynamic fluctuations. For example, the confidence level of facial recognition may decline due to a sudden drop in lighting, or the access card's reading success rate may degrade due to antenna aging. Existing collaborative algorithms often assign fixed weights to each authentication source, which allows degraded authentication sources to participate in subsequent calculations with higher weights, thus misleading the update of the global strategy. Therefore, this embodiment constructs a dynamic reliability weight score for authentication sources to quantify the output quality and stability of each access node's authentication source in real time.

[0025] At the current moment As the endpoint, the length of each authentication source at the access node is set to [length value]. A sliding time window is used to calculate the variance of the real-time confidence sequence within that window. Furthermore, the arithmetic mean of different subsets of environmental quality parameters corresponding to the access nodes is taken to obtain the environmental quality factor. This environmental quality factor is the mean of the normalized light intensity of the access node and the equipment health status label. For access card sources, the environmental quality factor is the mean of the normalized temperature suitability, normalized humidity suitability, and the equipment health status label of the access node.

[0026] The dynamic credibility weight score of the authentication source is constructed based on the fact that in complex park scenarios, the confidence level at a single moment is unreliable. Therefore, a comprehensive penalty or incentive should be applied, taking into account both the historical stability of the authentication source and the current physical environment. The calculation formula is as follows: In the formula, Indicates the passage node The Middle Personal authentication rights are based on Dynamic credibility weight score at any given time; For passage nodes The Middle Personal authentication rights are based on Real-time confidence level directly output by the recognition model at any given moment; For passage nodes The Middle The variance of the real-time confidence sequence of each authentication source within the same sliding window; For passage nodes The Middle Personal authentication rights are based on Environmental quality factors at any given time.

[0027] As can be seen from the above formula, when the passage node The Middle When the quality of an authentication source degrades due to a sudden drop in external light or internal equipment failure, its real-time confidence level... It will decrease significantly, and the associated environmental quality factors This fluctuation will decrease the variance of the source confidence sequence. The product of the numerators in the formula decreases sharply, while the variance penalty term in the denominator increases. The combined effect of these three factors causes a significant drop in the dynamic reliability weight score of the authentication source. This indicator successfully achieves a fine-grained perception of the degraded state, avoiding misleading overall judgments from poor-quality signals.

[0028] S103: Construct a multi-source fusion weighted consistency index for access nodes.

[0029] After obtaining the dynamic quality fluctuation characteristics of a single source, it is further necessary to analyze the interaction characteristics between various authentication sources within the access node. In actual operation, the phenomenon of multi-source pseudo-consistency often exists, that is, under adverse conditions, multiple authentication sources may be simultaneously interfered with and output incorrect rejection decisions. On the surface, the decisions are consistent, but in essence, they are all in a degenerate state. If this situation is simply equated with high-quality majority consensus, it will produce seriously misleading reward signals. In order to accurately reflect the true reliability of the current authentication decision fusion of the access node, this embodiment constructs a multi-source fusion weighted consistency index.

[0030] First, based on the dynamic trust weight score obtained in step S102, the normalized weight of each authentication source's dynamic trust weight score among all authentication sources in the current access node is calculated, thus obtaining the decision fusion weight of each authentication source. Simultaneously, a decision threshold is set for the real-time confidence level of each authentication source. When the real-time confidence level is greater than or equal to this threshold, its binary decision value is 1, and it is marked as allowed; otherwise, the binary decision value is 0, and it is marked as rejected. Then, the mean of the binary decision values ​​of each authentication source is calculated, where the decision threshold is set to 0.7.

[0031] The construction of the multi-source fusion weighted consensus index is based on introducing dynamic credibility weights into the deviation calculation framework of each source's decision results. This ensures that decision disagreements from high-quality sources are given priority, while false consensus results from low-quality sources are strongly suppressed. Its calculation formula is as follows: In the formula, Indicates the passage node exist The multi-source fusion weighted consistency index at any given time; For passage nodes The total number of currently active authentication sources; For passage nodes No. Personal authentication rights are based on Moment-by-moment decision fusion weights; For passage nodes No. Personal authentication rights are based on The binary decision value at time t; For passage nodes The mean of the binary decision values ​​corresponding to each authentication source.

[0032] As can be seen from this formula, when the authentication sources with high confidence weights experience decision disagreements, the binary decision values ​​corresponding to each authentication source will be... with the mean The difference is too large, coupled with its decision fusion weights. A large denominator in the fraction leads to a significant increase in the component calculated from that source, resulting in a substantial decrease in the overall index. In a pseudo-consistency scenario, even if multiple degenerate sources maintain the same erroneous decision, the decision fusion weights of the degenerate sources will still influence the outcome. Extremely low, its corresponding calculated value is strongly suppressed, and the overall consistency index remains at a low level. Compared to traditional voting statistics methods, this index can fundamentally isolate meaningless consistency.

[0033] S104: Risk value of degradation of authentication link for constructing access nodes.

[0034] In complex park operations, multiple authentication sources at the same access node often exhibit environmental linkages. This means that sudden weather changes or large-scale signal obstruction can cause multiple different authentication sources to degrade synchronously within the same time period. This cascading effect exposes the authentication link to a systemic risk of collapse. To comprehensively assess this multi-source linkage degradation risk, this embodiment, based on the degree of consistency deterioration, further integrates the quality level and proportion of extremely degraded sources to construct an authentication link degradation risk value.

[0035] When performing this step, a quality degradation threshold is first set. The system iterates through all current active sources of the passing node and counts the number of degraded sources whose dynamic trust weight scores are lower than the quality degradation threshold. The quality degradation threshold can be determined using the Otsu thresholding method. Simultaneously, the minimum current dynamic trust weight score is extracted from all active sources to characterize the state of the most vulnerable link in the current link.

[0036] The risk value for authentication link degradation is constructed based on the following: the risk level is a three-dimensional multiplicative representation of the degree of internal decision-making disagreement, the extent of reaching the worst-case scenario in local performance, and the overall extent of degradation spread. The calculation formula is as follows: In the formula, Indicates the passage node exist Risk value of authentication link degradation at any time; For passage nodes exist The multi-source fusion weighted consistency index at any given time; For passage nodes The minimum value of the dynamic trust weight score among all currently active authentication sources; For passage nodes The number of authentication sources that currently meet the quality degradation threshold; For passage nodes The total number of currently active authentication sources.

[0037] As can be seen from this formula, when multi-source linkage degradation occurs, decision consistency is the first to be disrupted, and the multi-source fusion weighted consistency index... The decrease causes the first subtraction factor to increase. Simultaneously, the vulnerability source's performance bottoms out, leading to a minimum value. The drastic reduction in authentication link size leads to a significant increase in the risk of degradation. The large number of authentication sources marked as degraded causes a substantial rise in the second proportionality factor, strongly driving the final authentication link degradation risk value towards its numerical limit. This multiplicative comprehensive modeling provides an extremely sensitive and accurate risk warning signal for early intervention in subsequent scheduling systems.

[0038] S105: Construction of an improved algorithm based on the risk value of authentication link degradation and implementation of collaborative control.

[0039] After obtaining the authentication link degradation risk value of each access node in the entire park, the global authentication link degradation risk average is obtained by taking the arithmetic mean of all access nodes in the entire park. Addressing the problem that traditional multi-agent near-end policy optimization algorithms cannot adapt to dynamic scenarios due to the fixed global reward function structure, this embodiment utilizes this global risk average to decouple and adaptively reconstruct the fixed reward function.

[0040] Specifically, the global reward function is decomposed into two parts: authentication fusion sub-reward and device collaboration sub-reward. The authentication fusion sub-reward is constructed based on the average of the authentication fusion decision sub-rewards of all access nodes in the entire park, and its specific formula is as follows: In the formula, For a moment Authentication integration sub-rewards; This represents the total number of currently activated access nodes across the entire park. For passage nodes Multi-source fusion weighted consistency index; For passage nodes exist The system continuously processes the correctness flags of the fusion decision for current access requests, automatically generating truth labels in the simulation training environment based on preset personnel access permission rules. When an access node... When the fusion decision matches the truth label, Select 1 if the value is 1, otherwise select 0.

[0041] The equipment collaboration sub-reward is defined as the product of normalized traffic throughput and latency equalization coefficient, used to characterize the overall smooth collaboration capability of the entire park's road network. Its formula is: In the formula, For a moment Device collaboration sub-rewards; For a moment The mean of normalized throughput for all passing nodes; For a moment The latency equalization coefficient is 1 minus the mean of the normalized standard deviation of the response latency of all access nodes in the entire park. The larger the value, the more balanced the response of each node and the higher the coordination efficiency.

[0042] Based on this, using the global risk mean as an adaptive lever, the two sub-rewards are dynamically concatenated, and the reconstructed global reward function is calculated as follows: In the formula, For a moment The global reward for dynamic adaptive reconstruction; For all passing nodes at time The average value of the authentication link degradation risk. For a moment The authentication integration sub-reward, For a moment Device collaboration sub-rewards.

[0043] As can be seen from this global reward formula, when the park encounters extremely severe weather or widespread network jitter, resulting in a higher average global risk for the system, that is... When the value is large, the weight of the right-side authentication fusion sub-reward will be adaptively increased. In this state, the policy network is guided to prioritize ensuring and correcting the authentication reliability of each passable node. Conversely, under stable daily operation and When the value is low, the equipment collaboration sub-reward takes the lead, and the system then focuses on maximizing the efficiency of equipment passage collaboration while ensuring safety.

[0044] In implementing collaborative control, this embodiment uses an improved multi-agent strategy optimization algorithm to treat turnstiles, access control systems, and gate barriers at various access nodes in the park as agents for collaborative control. Specifically: The system input is the local observation vector of each access node. This local observation vector is composed of the dynamic trust weight score of each authentication source of the access node, the multi-source fusion weighted consistency index, the authentication link degradation risk value, and the current access request feature vector. The current access request feature vector includes the identity code, the security level of the requested area, and the summary of historical access records. The system uses the dynamically adaptively reconstructed global reward R(t) at time t as the guiding target. This reward value can dynamically adjust the weight between security protection and access efficiency according to the real-time global risk average. Finally, based on the above input state, each edge server outputs the optimal control action of each authentication source in the current state in real time under the guidance of the strategy balance of the global reward R(t). The control actions include allowing, rejecting, or initiating secondary verification and are sent to the physical access control interface to ensure that the best access efficiency and regional security isolation level are maintained in various fluctuating environments.

[0045] like Figure 2 As shown in the figure, this demonstrates that the improved MAPPO algorithm successfully overcomes the fundamental flaw of the original MAPPO algorithm, which uses a fixed-structure global reward function. This method uses the authentication link degradation risk value... The global reward function was reconstructed using adaptive weighting coefficients. When the risk index is low, the system prioritizes optimizing traffic efficiency; when the risk index rises, the system automatically switches to a security enhancement mode, increasing the weight of the authentication fusion sub-reward and guiding the policy network to prioritize authentication reliability. This demonstrates that the present invention can perfectly achieve adaptive balanced collaborative control between security and traffic efficiency in complex campus environments with dynamic fluctuations in multi-source authentication.

[0046] like Figure 3 As shown in the figure, this diagram intuitively demonstrates that the dynamic credibility weight scoring mechanism for authentication sources constructed by this method can uniformly incorporate real-time confidence, historical accuracy, confidence variance, and environmental quality factors into a quantitative framework. It can perceive the dynamic quality changes of each authentication source under different environmental conditions in real time and with precision, thereby effectively avoiding the systematic bias caused by the traditional fixed weight allocation method, which still uses distorted equal-weight signals in the global calculation when the quality of a certain authentication source degrades.

[0047] This invention also provides a collaborative control system for park access equipment based on multi-source authentication. The system includes a processor and a memory, the memory storing computer program instructions. When the processor executes the computer program instructions, it implements the collaborative control method for park access equipment based on multi-source authentication according to the first aspect of this invention.

[0048] The system also includes other components well known to those skilled in the art, such as communication buses and communication interfaces, the settings and functions of which are known in the art and therefore will not be described in detail here.

[0049] In this invention, the aforementioned memory can be any tangible medium containing or storing a program that can be used or combined with an instruction execution system, apparatus, or device. For example, a computer-readable storage medium can be any suitable magnetic or magneto-optical storage medium, such as Resistive Random Access Memory (RRAM), Dynamic Random Access Memory (DRAM), Static Random Access Memory (SRAM), Enhanced Dynamic Random Access Memory (EDRAM), High-Bandwidth Memory (HBM), Hybrid Memory Cube (HMC), etc., or any other medium that can be used to store desired information and can be accessed by an application, module, or both. Any such computer storage medium can be part of a device or accessible to or connected to a device. Any application or module described in this invention can be implemented using computer-readable / executable instructions stored or otherwise maintained on such a computer-readable medium.

[0050] The embodiments described above are merely examples of several implementations of the present invention, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention.

Claims

1. A collaborative control method for park access equipment based on multi-source authentication, characterized in that, include: Obtain multi-source authentication raw data and environmental quality parameters for each access node; Construct a dynamic trust weight score for any authentication source in each access node at the current time, which is used to quantify the output quality and stability of each authentication source under different environmental quality parameters at the current time; Calculate the multi-source fusion weighted consistency index of each access node at the current time to characterize the true reliability of the authentication decisions of the access node at the current time after fusion; Construct the authentication link degradation risk value of each access node at the current moment. The authentication link degradation risk value is negatively correlated with the multi-source fusion weighted consistency index at the current moment, and is also negatively correlated with the minimum value of the dynamic trust weight score among all currently activated authentication sources. Using the degradation risk value of the authentication link as an adaptive lever, the global reward function of the multi-agent near-end policy optimization algorithm is reconstructed, and the optimal control action of each authentication source is output.

2. The collaborative control method for park access equipment based on multi-source authentication according to claim 1, characterized in that, The global reward function of the multi-agent proximal policy optimization algorithm is reconstructed, including: The global reward function at the current moment includes the authentication fusion sub-reward at the current moment and the device collaboration sub-reward at the current moment; The average of the authentication link degradation risk values ​​of all access nodes at the current moment is used as the weight of the authentication fusion sub-reward at the current moment, and the complement of the average of the authentication link degradation risk values ​​of all access nodes at the current moment is used as the weight of the device collaboration sub-reward at the current moment, so as to achieve adaptive reconstruction of the global reward.

3. The collaborative control method for park access equipment based on multi-source authentication according to claim 2, characterized in that, The calculation method for the authentication fusion sub-reward at the current moment includes: Calculate the sum of the products of the multi-source fusion weighted consistency index of all passing nodes at the current time and the correctness flag value of the fusion decision for processing the current passage request, and multiply the sum of the products by the reciprocal of the total number of all passing nodes at the current time to obtain the authentication fusion sub-reward at the current time.

4. The collaborative control method for park access equipment based on multi-source authentication according to claim 3, characterized in that, The method for obtaining the correctness flag value of the fusion decision in processing the current passage request includes: In the simulation training environment, truth labels are automatically generated based on preset personnel access permission rules. When the current fusion decision of the access node is consistent with the truth label, the correctness flag of the fusion decision is set to 1; otherwise, it is set to 0.

5. The collaborative control method for park access equipment based on multi-source authentication according to claim 2, characterized in that, The calculation method for the device collaboration sub-reward at the current moment includes: The mean of the normalized throughput of all passing nodes at the current moment is calculated. The mean is multiplied by the delay balance coefficient at the current moment to obtain the device coordination sub-reward at the current moment. The delay balance coefficient is negatively correlated with the mean of the normalized standard deviation of the response delay of all passing nodes.

6. The collaborative control method for park access equipment based on multi-source authentication according to claim 5, characterized in that, Obtain the throughput of each access node, including: The throughput of each passage node refers to the ratio of the number of people actually allowed to pass through the node within the current time window to the theoretical maximum number of people allowed to pass through the node.

7. The collaborative control method for park access equipment based on multi-source authentication according to claim 1, characterized in that, The calculation method for dynamic credibility weight score is as follows: The real-time confidence level, variance of the confidence level sequence, and environmental quality factor of each authentication source for multiple access nodes within the park are obtained. The environmental quality factor reflects the output quality of the authentication source, while the variance of the confidence level sequence reflects the stability of the authentication source. The dynamic confidence weight score is positively correlated with the real-time confidence level and the environmental quality factor, and negatively correlated with the variance of the confidence level sequence.

8. The collaborative control method for park access equipment based on multi-source authentication according to claim 1, characterized in that, The formula for calculating the multi-source fusion weighted consensus index is: ； In the formula, Indicates the passage node exist The multi-source fusion weighted consistency index at any given time; For passage nodes The total number of currently active authentication sources; For passage nodes No. Personal authentication rights are based on The decision fusion weight at any given time is the normalized value of the dynamic credibility weight score; For passage nodes No. Personal authentication rights are based on The binary decision value at time t; For passage nodes The binary decision value of each authentication source is 1 when the real-time confidence level of each authentication source is greater than or equal to the decision threshold; otherwise, the binary decision value is 0.

9. The collaborative control method for park access equipment based on multi-source authentication according to claim 1, characterized in that, Output the optimal control actions for each authentication source, including: The multi-agent near-end policy optimization algorithm based on the reconstructed global reward function takes as input the local observation vector of each access node. This local observation vector is composed of the dynamic trust weight score of each authentication source of the access node, the multi-source fusion weighted consistency index, the authentication link degradation risk value, and the current access request feature vector. The output is the optimal control action of each authentication source in the current state. The control action includes allowing access, rejecting access, or initiating secondary verification.

10. A collaborative control system for park access equipment based on multi-source authentication, comprising a processor and a memory, characterized in that, The memory stores a computer program, and the processor executes the computer program to implement the collaborative control method for park access equipment based on multi-source authentication as described in any one of claims 1-9.

Images