Parking lot entrance and exit control system based on internet of things

Abstract

The application relates to the technical field of intelligent control of parking lot entrances and exits, in particular to a parking lot entrance and exit control system based on the Internet of Things, which comprises a data acquisition unit for generating multi-dimensional dynamic data; a risk prediction unit for generating a system collapse risk index; a strategy adjustment unit for generating a risk level signal; the strategy adjustment unit is also used for generating a control instruction based on the risk level signal and the system collapse risk index; and an adaptive optimization unit for collecting an entrance average passing time and a system total throughput after the control instruction is executed, and performing adaptive optimization processing to generate a model parameter correction instruction for correcting the risk prediction unit and the strategy adjustment unit; the application significantly improves the vehicle passing efficiency of a large parking lot during a peak period and the overall intelligent management level.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent control of parking lot entrances and exits, in particular to a parking lot entrance and exit control system based on the Internet of Things. BACKGROUND

[0002] In the field of intelligent traffic management today, the parking lot entrance and exit control system is facing the risk of systemic congestion collapse caused by factors such as traffic overload and network channel deterioration. Traditional systems mostly use passive response control, that is, they take measures only after congestion occurs, which not only leads to low traffic efficiency, but also easily causes congestion avalanches. This lagging control mode lacks the ability to predict potential risks in advance and cannot take differentiated intervention strategies for different risk sources.

[0003] The prior art has obvious limitations in solving these problems. Some systems can monitor traffic flow, but usually rely only on single-dimensional data and fail to build a cross-scale, multi-dimensional system state perception model. In addition, traditional queuing theory models rely on linear steady-state assumptions and are difficult to capture the complex precursors of the system's nonlinear phase transition from stable state to congestion state. At the control level, existing methods mostly use single and extensive flow limiting strategies, lack fine-grained dynamic adjustment mechanisms according to risk levels, and more importantly, these systems generally lack the ability to adaptively learn and optimize, and cannot self-calibrate according to actual operation results, making their performance difficult to maintain optimal in the face of changing external environments. Therefore, how to provide an intelligent parking lot entrance and exit control system that can actively predict systemic collapse risks, dynamically adjust control strategies and continuously optimize itself is a technical problem that needs to be solved by those skilled in the art. SUMMARY

[0004] To solve the above technical problems, the present application provides a parking lot entrance and exit control system based on the Internet of Things. Specifically, the technical solution of the present application includes:

[0005] A data acquisition unit for real-time acquisition of authentication signaling timestamps, entrance queuing vehicle numbers, parking space occupancy states and network channel quality parameters of the parking lot to generate multi-dimensional dynamic data;

[0006] A risk prediction unit for receiving multi-dimensional dynamic data and performing risk prediction processing to generate a system collapse risk index;

[0007] A strategy adjustment unit for receiving the system collapse risk index and comparing and analyzing it with preset warning state thresholds and dangerous state thresholds to generate a risk level signal. The strategy adjustment unit is also used to generate control instructions based on the risk level signal and the system collapse risk index;

[0008] An adaptive optimization unit is configured to collect the average access travel time and the total system throughput after the execution of the control instruction, and perform adaptive optimization processing to generate model parameter correction instructions for correcting the risk prediction unit and the strategy adjustment unit.

[0009] Preferably, the process of the risk prediction unit performing risk prediction processing to generate the system collapse risk index comprises:

[0010] Based on the multi-dimensional dynamic data, the jitter variance of the authentication signaling timestamp, the instantaneous growth acceleration of the access queue length, the global entropy of the parking space turnover rate, and the Lyapunov index of the expected arriving vehicle flow are calculated.

[0011] Based on the calculated jitter variance, instantaneous growth acceleration, global entropy, and Lyapunov index, nonlinear coupling calculation is performed to generate the system collapse risk index.

[0012] Preferably, the process of the strategy adjustment unit generating the risk level signal comprises:

[0013] When the system collapse risk index is less than or equal to the warning state threshold, a safe risk level signal is generated.

[0014] When the system collapse risk index is greater than the warning state threshold and less than or equal to the dangerous state threshold, a warning risk level signal is generated.

[0015] When the system collapse risk index is greater than the dangerous state threshold, a dangerous risk level signal is generated.

[0016] Preferably, the control instruction is an authentication request admission rate adjustment instruction.

[0017] The strategy adjustment unit generates an authentication request admission rate adjustment instruction based on the logistic function model and the current input system collapse risk index in response to the warning risk level signal or the dangerous risk level signal.

[0018] Preferably, the control instruction is a multi-modal authentication protocol switching instruction.

[0019] The strategy adjustment unit generates a multi-modal authentication protocol switching instruction when the contribution of the jitter variance of the authentication signaling timestamp to the system collapse risk index is greater than a preset contribution threshold in response to the dangerous risk level signal.

[0020] Preferably, the multi-modal authentication protocol switching instruction is used to switch the authentication mode from an Internet of Things protocol to a video stream-based license plate recognition protocol.

[0021] Preferably, the process of the adaptive optimization unit performing adaptive optimization processing comprises:

[0022] The cost function is calculated based on the average passage time at the entrance, the total system throughput, and the preset target passage time and target throughput.

[0023] Based on the cost function, a numerical perturbation gradient approximation method is used for minimization to generate model parameter correction instructions.

[0024] Preferably, the model parameter correction instruction is used to correct the weight coefficients used by the risk prediction unit and the response center point threshold and slope factor used by the strategy adjustment unit.

[0025] Compared with the prior art, the present invention has the following beneficial effects:

[0026] 1. This system uses a data acquisition unit to acquire multi-dimensional data in real time, such as authentication signaling timestamps, the number of vehicles queuing at the entrance, parking space occupancy status, and network channel quality parameters, forming a comprehensive snapshot of the system status. The risk prediction unit uses a nonlinear coupling model that integrates core features at the micro, meso, macro, and dynamic time-varying levels to generate a system collapse risk index. This method can predict the systemic congestion risk caused by computing resource bottlenecks and network channel deterioration earlier and more accurately, effectively avoiding congestion avalanches.

[0027] 2. This system uses a strategy adjustment unit to compare the risk index with preset warning and danger thresholds to generate three risk level signals: safe, warning, or dangerous. This graded response mechanism replaces the traditional one-size-fits-all control method. When the system is in a warning state, it initiates dynamic adjustment of the authentication request acceptance rate; when it is in a dangerous state and a network problem is diagnosed, it switches to a license plate recognition protocol based on video stream, thereby achieving targeted intelligent control.

[0028] 3. This system introduces an adaptive optimization unit, enabling the system to self-iterate and improve. After the control command is executed, this unit continuously monitors the average passage time at the entrance and the total system throughput. It quantifies the deviation between the actual performance and the target through a cost function and minimizes it using a numerical perturbation gradient approximation method. This generates model parameter correction commands, which are used to fine-tune the parameters of the risk prediction and strategy adjustment units online, ensuring that the system performance remains optimal in the face of constantly changing environments.

[0029] 4. This system constructs a complete closed-loop control system of perception, prediction, decision-making, and optimization. By proactively predicting and intervening in the risk of systemic collapse caused by computing resource bottlenecks, it effectively avoids congestion at parking lot entrances. At the same time, the adaptive optimization mechanism enables the system to continuously learn and adapt to environmental changes, maintaining a highly efficient and stable operating state in the long term. This forward-looking, adaptive, and robust design significantly improves the vehicle throughput efficiency and overall intelligent management level of large parking lots during peak hours. Attached Figure Description

[0030] The present invention will be further explained below with reference to the accompanying drawings and embodiments:

[0031] Figure 1 This is a structural diagram of the system of the present invention. Detailed Implementation

[0032] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0033] Example 1:

[0034] Please see Figure 1 The IoT-based parking lot entrance and exit control system includes:

[0035] The data acquisition unit is used to collect the authentication signaling timestamps, the number of vehicles queuing at the entrance, the occupancy status of parking spaces, and network channel quality parameters of the parking lot in real time to generate multi-dimensional dynamic data.

[0036] The risk prediction unit is used to receive multidimensional dynamic data and perform risk prediction processing to generate a system crash risk index.

[0037] The strategy adjustment unit is used to receive the system crash risk index and compare it with the preset warning state threshold and danger state threshold to generate a risk level signal; the strategy adjustment unit is also used to generate control commands based on the risk level signal and the system crash risk index.

[0038] The adaptive optimization unit is used to collect the average passage time at the entrance and the total system throughput after the control command is executed, and to perform adaptive optimization processing to generate model parameter correction commands for the risk prediction unit and the strategy adjustment unit.

[0039] This embodiment provides an IoT-based parking lot entrance and exit control system. The system constructs an intelligent control system that can proactively prevent congestion collapse and continuously self-optimize through cross-scale state perception, nonlinear risk prediction, dynamic strategy adjustment, and adaptive closed-loop optimization. The system includes a data acquisition unit, a risk prediction unit, a strategy adjustment unit, and an adaptive optimization unit.

[0040] The data acquisition unit functions to provide a comprehensive, real-time, and multi-dimensional snapshot of the system status for subsequent risk prediction. In this embodiment, the unit can be specifically implemented as an integrated hardware and software module. It completes data acquisition through an IoT communication module deployed at the parking lot entrance, a high frame rate video monitoring unit, and real-time parsing of the backend authentication server logs. The acquired data includes: authentication signaling timestamps of each vehicle's authentication interaction signaling, accurate to milliseconds, recorded by the authentication server; the number of vehicles queuing at the entrance, obtained in real-time through video stream image recognition algorithms; the parking space occupancy status of each zone obtained from the parking lot central management system; and network channel quality parameters provided by regional mobile communication network base station monitoring tools. These raw data are formatted and together form a multi-dimensional dynamic data stream, which is transmitted to the risk prediction unit in real time.

[0041] The risk prediction unit is designed to receive and process the aforementioned multidimensional dynamic data. Through a nonlinear coupling model, it transforms seemingly unrelated multidimensional indicators into a single quantitative indicator that can accurately predict the risk of systemic collapse, namely the system collapse risk index. This unit is the core computing engine of the system. It abandons the limitations of traditional queuing theory models under the assumption of linear steady state and instead captures subtle signs before the system transitions from a steady state to a congested state from the perspective of system dynamics and chaos theory.

[0042] The strategy adjustment unit transforms the abstract risk index output by the risk prediction unit into specific, executable control commands, serving as the decision-making center connecting prediction and control. This unit receives the real-time system crash risk index and compares it with two preset thresholds. The warning state threshold is defined as the critical value at which the system begins to show signs of instability, requiring initial intervention; the danger state threshold is defined as the critical value at which the system is about to enter an avalanche-like collapse, necessitating strong intervention. These two thresholds are derived from regression analysis of historical congestion event data. For example, the warning state threshold can be taken as the 80th percentile of the risk index distribution 5 minutes before a historical congestion collapse, while the danger state threshold can be taken as the 95th percentile of the risk index distribution 1 minute before the collapse, thus ensuring the objectivity and effectiveness of the threshold settings. Through comparison, the unit generates a clear risk level signal. Based on this risk level signal and the input system crash risk index, and according to the preset control logic model, control commands for regulating inbound traffic and authentication methods are generated.

[0043] The adaptive optimization unit introduces learning and evolution capabilities into the entire system, ensuring optimal control performance even in the face of constantly changing external traffic environments. After control commands are executed, this unit continuously monitors the system's macroscopic performance, specifically collecting two key performance indicators: average entry time and total system throughput. It compares these actual performance metrics with the system's preset optimization targets, quantifying performance deviations through a cost function. Based on this deviation, the unit executes an online optimization algorithm for adaptive optimization, outputting model parameter correction commands for fine-tuning the core parameters within the risk prediction and strategy adjustment units. This process forms a complete feedback control closed loop, enabling the system to self-calibrate and continuously improve its prediction accuracy and control effectiveness.

[0044] This embodiment constructs a complete perception-prediction-decision-optimization closed-loop control system through the collaborative work of the four units mentioned above. Compared with the passive response to congestion control methods in the prior art, this invention can predict the risk of systemic collapse caused by computing resource bottlenecks in advance and actively intervene, thereby effectively avoiding the avalanche phenomenon of congestion at parking lot entrances. Through adaptive optimization, the system can also continuously learn and adapt to changes in the external environment, maintaining a highly efficient and stable operating state in the long term, significantly improving the vehicle traffic efficiency and intelligent management level of large parking lots during peak hours. It should be noted that this embodiment mainly targets the risk of systemic congestion collapse related to computing and communication resources caused by factors such as traffic overload and network channel deterioration. Other types of system failures, such as purely physical hardware failures, are not within the direct prevention scope of this system.

[0045] Example 2:

[0046] The process by which the risk prediction unit performs risk prediction processing to generate a system crash risk index includes:

[0047] Based on multidimensional dynamic data, the jitter variance of authentication signaling timestamps, the instantaneous growth acceleration of entrance queue length, the global entropy of berth space turnover rate, and the Lyapunov index of expected arrival traffic flow are calculated.

[0048] Based on the calculated jitter variance, instantaneous growth acceleration, global entropy, and Lyapunov exponent, a nonlinear coupling calculation is performed to generate a system collapse risk index.

[0049] This embodiment specifically illustrates how the risk prediction unit described in Embodiment 1 performs risk prediction processing to generate a system crash risk index. The detailed process; this process achieves a comprehensive characterization of the system state by integrating the core characteristics of the four levels: micro, meso, macro, and dynamic time-varying.

[0050] This unit performs cross-scale core feature parameter quantization calculations based on the received multi-dimensional dynamic data:

[0051] Jitter variance of authentication signaling timestamp The calculation of jitter variance is defined as the statistical measure of the dispersion of the time points when all vehicle authentication is completed and their moving average within a short time window. Its role is to characterize the computational load of the underlying authentication server and the instantaneous stability of the network channel; it is a key indicator for capturing system stress at the micro-vehicle level.

[0052] Instantaneous growth acceleration of the entry queue length The calculation of instantaneous growth acceleration is defined as the second derivative of the real-time monitored entry queue length function with respect to time. Its function is to capture the impact intensity and outbreak trend of congestion; it measures the dynamic evolution of congestion at the mesoscopic channel level.

[0053] Global entropy of berth space turnover rate The calculation of global entropy: Global entropy is defined as an indicator to measure the degree of disorder in the utilization pattern of parking spaces within a parking lot. To calculate this indicator, the parking lot must first be divided into sections based on its physical layout and management needs. Each logical region can be divided, for example, by floor, zone, or row, and statistics can be collected for each region within a unit of time T. Frequency of changes in berth status Thus, the probability of state change in this region can be obtained. Global entropy Calculated using the following formula:

[0054]

[0055] This parameter reflects the degree of chaos in internal traffic flow at the macro-level of the station; the higher the entropy value, the more chaotic the internal circulation, and the smaller the margin for the system to absorb external traffic pressure.

[0056] Lyapunov index of expected arrival traffic Calculation: The Lyapunov exponent is an indicator derived from chaos theory, defined as a quantity used to determine the future predictability of a time series. Its function is to quantify the dynamic uncertainty of short-term traffic flow. This embodiment uses historical and real-time entrance traffic flow time series and calculates the maximum Lyapunov exponent of the series through a phase space reconstruction method. The embedding dimension of the phase space reconstruction... and delay time Key parameters can be determined using standard methods such as the CC method or the spurious nearest neighbor method to ensure the validity of the calculation. Considering the complexity of real-time calculation of this index, in specific engineering implementations, alternative indicators with higher computational efficiency can be used to quantify the uncertainty of the time series, or the calculation update cycle of the Lyapunov index can be set to be longer than other parameters, such as updating once per minute, to balance the predictive foresight with the computational feasibility. When this occurs, it indicates that the traffic flow has entered a chaotic state, and the system faces significant risks due to the high uncertainty of external inputs.

[0057] After calculating the four core characteristic parameters mentioned above, the risk prediction unit performs nonlinear coupling calculations to generate the final system collapse risk index. This calculation is achieved through a heuristic model that incorporates the principles of system dynamics. Before the calculation, some parameters are normalized to eliminate the influence of dimensions. , , ;in , , These are the maximum values ​​allowed in system design or historical statistics, both of which are dimensionless; they are calculated using the following formula:

[0058]

[0059] System crash risk index It is a dimensionless scalar; the higher its value, the greater the probability that the system will enter a state of computational resource deadlock. These are preset weighting coefficients. These are constants characterizing the timescale of the system's crisis evolution; to ensure feasibility, the calibration process for these parameters is as follows: collect a calibration dataset containing M historical time segments, for each data point... The feature vector containing this time point And a true label representing whether the system has collapsed. The dataset was fitted using classification algorithms such as logistic regression to obtain a set of optimized parameters that maximized the model's prediction accuracy. ;parameter The value is determined by analyzing the characteristic time it takes for the risk index to develop from a warning state to a dangerous state in historical congestion events; the innovation of this formula lies in the fact that the linear weighted part within the parentheses integrates the internal system pressures from three levels, while the exponential term... This is used to detect the precursors of a phase transition when the expected traffic flow exhibits chaotic characteristics, indicating that the system risk will increase exponentially.

[0060] The design of this structure is based on the following considerations: Before reaching a critical point, the contributions of internal computational load, queue pressure, and loop efficiency to the overall system risk are approximately linearly additive; however, the unpredictability of external inputs due to chaotic characteristics can exponentially amplify small disturbances in the system state, leading to rapid system instability within a short period. This aligns with the characteristic of errors in chaotic systems growing exponentially over time. Therefore, combining internal pressure and external uncertainty multiplicatively can more accurately characterize this nonlinear, abrupt collapse risk.

[0061] By fusing four cross-scale, multi-dimensional feature parameters and employing a nonlinear coupling model, the complex precursors to systemic collapse can be identified earlier and more accurately. The introduction of the Lyapunov exponent allows the model to quantify the risks inherent in uncertainty itself, successfully predicting avalanche-like congestion caused by chaotic traffic flow—a scenario traditional models cannot handle—significantly improving the foresight and reliability of risk prediction. The linear weighted model used here is an effective simplification that ensures computational efficiency and model interpretability, aiming to capture the first-order principal effects of various stressors. In scenarios requiring higher accuracy, this linear component can be replaced with nonlinear models such as neural networks to capture the potential complex coupling effects between features.

[0062] Example 3:

[0063] The process by which the strategy adjustment unit generates risk level signals includes:

[0064] When the system crash risk index is less than or equal to the warning status threshold, a safety risk level signal is generated.

[0065] When the system crash risk index is greater than the warning state threshold and less than or equal to the danger state threshold, a warning risk level signal is generated.

[0066] When the system crash risk index is greater than the danger state threshold, a danger risk level signal is generated.

[0067] This embodiment is a concretization of the process by which the strategy adjustment unit generates risk level signals as described in Embodiment 1; this process is the direct basis for the execution of all subsequent control strategies, and its purpose is to provide a clear and hierarchical action guide.

[0068] The strategy adjustment unit receives the system crash risk index calculated in real time by the risk prediction unit. This unit has two preset key thresholds: a warning status threshold. and danger threshold The internal logic of this unit is based on The comparison results with these two thresholds generate the corresponding risk level signal;

[0069] When the system crash risk index Less than or equal to the warning status threshold At that time, that is This generates a safety risk level signal; under this state, all system indicators are within the normal range, and the control system does not need to intervene.

[0070] When the system crash risk index Greater than the warning status threshold And less than or equal to the danger threshold At that time, that is The system generates a warning risk level signal; this signal indicates that the system has begun to be under pressure and is trending towards congestion, and the system will initiate preventive adjustment strategies.

[0071] When the system crash risk index Greater than the danger threshold At that time, that is This generates a risk level signal; this signal indicates that the system is on the verge of collapse, congestion may break out at any time, and the system will initiate emergency intervention strategies.

[0072] This three-tiered risk classification enables refined management of system status; it avoids the traditional one-size-fits-all control approach and replaces it with a gradual and differentiated intervention strategy that matches the level of risk; this graded response mechanism not only improves the accuracy of control but also maximizes the balance between traffic efficiency and system stability.

[0073] Example 4:

[0074] The control command is an authentication request acceptance rate adjustment command;

[0075] The strategy adjustment unit responds to the early warning risk level signal or the dangerous risk level signal, and generates an authentication request acceptance rate adjustment instruction based on the logistic function model and the currently input system crash risk index;

[0076] The control command is a multimodal authentication protocol switching command;

[0077] The strategy adjustment unit responds to the danger risk level signal and performs judgment processing. When the contribution of the jitter variance of the authentication signaling timestamp to the system crash risk index is greater than the preset contribution threshold, a multimodal authentication protocol switching instruction is generated.

[0078] The multimodal authentication protocol switching command is used to switch the authentication mode from an IoT protocol to a video stream-based license plate recognition protocol.

[0079] This embodiment further defines the specific control instructions generated by the strategy adjustment unit after receiving signals of different risk levels; these instructions aim to mitigate the predicted risks through proactive intervention, mainly including two core adjustment strategies: dynamic adjustment of authentication request acceptance rate and dynamic switching of multimodal authentication protocol;

[0080] The authentication request acceptance rate is dynamically adjusted. This adjustment is initiated when the policy adjustment unit responds to the warning risk level signal or the dangerous risk level signal. Its underlying logic is to protect the backend computing resources and prevent them from crashing due to overload by actively and smoothly limiting the inflow of requests at the source.

[0081] At this point, the generated control command is an authentication request acceptance rate adjustment command; this command is generated based on a logistic function model and uses the currently input system crash risk index. As a key variable, the specific authentication request acceptance rate Calculated using the following formula:

[0082]

[0083] Authentication request acceptance rate Defined as the upper limit of the authentication request processing rate that the system should currently execute, its unit is the number of requests per second, and its source is calculated by this formula; It represents the maximum physical processing rate of the system design, a constant preset based on hardware performance and network bandwidth; k is a dimensionless slope factor. These are the response center point thresholds; the sources of these two parameters can be determined by performing nonlinear fitting on historical control data, i.e., within a range containing different risk indices. and the corresponding optimal acceptance rate On the historical data set, the optimal solution is obtained by using the least squares method. and value;

[0084] The dynamic switching of the multimodal authentication protocol is a more targeted and powerful intervention method. When the policy adjustment unit responds to the risk level signal, it triggers an additional judgment process, the purpose of which is to diagnose the main root cause of the current high-risk state.

[0085] The judgment logic is as follows: the system calculates the contribution of each risk component to the total risk, and determines whether the contribution of the risk component corresponding to the authentication signaling timestamp jitter is greater than a preset contribution threshold. This contribution is calculated based on the risk prediction unit's role in generating the system crash risk index. The specific calculation method for the dimensionless core feature parameters used is as follows: ;in This is the risk contribution of jitter variance. These are the normalized ingress queue acceleration, jitter variance, and global entropy, all dimensionless values. This calculation method ensures that the contribution is a dimensionless relative ratio with clear physical meaning. The contribution threshold... The setting is based on the analysis of receiver operation characteristic curves of historical fault data to select the critical value that can best distinguish between crash events caused by network jitter and crash events caused by other reasons.

[0086] If the judgment result is yes, that is and Then the policy adjustment unit will generate a multimodal authentication protocol switching instruction; according to the further definition of embodiment 4, the instruction is used to switch the authentication mode from an Internet of Things protocol that is highly dependent on high-quality wireless channels to a backup mode that is not sensitive to channel quality, namely a license plate recognition protocol based on video streams.

[0087] By introducing the logistic function model for admission rate control, smooth and non-linear dynamic throttling is achieved; while the multi-modal authentication protocol switching realizes the upgrade from passive flow limiting to active diagnostic intelligent control, which can take targeted solutions to the root causes of risks. Even under extremely poor network conditions, it can ensure the basic passage capacity of the parking lot entrance, demonstrating extremely high robustness.

[0088] Example 5:

[0089] The adaptive optimization process performed by the adaptive optimization unit includes:

[0090] The cost function is calculated based on the average passage time at the entrance, the total system throughput, and the preset target passage time and target throughput.

[0091] Based on the cost function, a numerical perturbation gradient approximation method is used for minimization to generate model parameter correction instructions;

[0092] The model parameter correction command is used to correct the weighting coefficients used in the risk prediction unit and the response center point threshold and slope factor used in the strategy adjustment unit.

[0093] This embodiment describes in detail the specific implementation of the adaptive optimization unit described in Embodiment 1 to perform adaptive optimization processing to generate model parameter correction instructions; the core of this unit lies in building a feedback mechanism that can self-iterate and improve based on actual operating results;

[0094] This process is initiated periodically after the control command is executed;

[0095] This unit continuously monitors and collects two macroscopic indicators reflecting the final control effect of the system: average passage time at the entrance. and total system throughput Meanwhile, the system has two preset performance targets: target travel time. and target throughput These target values ​​are performance benchmarks preset based on the parking lot operation strategy;

[0096] To quantify the gap between the actual effect and the expected goal, this embodiment introduces a cost function. The calculation formula is as follows:

[0097]

[0098] Cost function This is a quantitative indicator used to evaluate the quality of the current control strategy of a system. The smaller the value, the closer the system's operating state is to the ideal target. Before applying this cost function to calculate the gradient, it is normalized, for example, by dividing it by a preset benchmark value, to obtain a dimensionless cost evaluation value. To ensure the consistency of dimensions in subsequent parameter updates, These are weighting coefficients used to balance the two optimization objectives of passage time and throughput. They are derived from adjustable constants preset according to operational management needs; in the formula, the average passage time at the entrance... The unit of measurement is time / vehicle, while the total system throughput The dimension is vehicles / time, therefore its reciprocal The physical meaning of is the average time required to process a single vehicle, and the dimension is also time / vehicle; this ensures that the two square terms added on the right side of the formula have the same dimension (time / vehicle)², thus guaranteeing the rigor and consistency of the cost function in terms of physical dimensions;

[0099] After calculating the current cost function value Then, the goal of the adaptive optimization unit is to adjust the internal parameters of the system to achieve... Minimize; since there is no direct analytical expression between the cost function and the parameter to be tuned, this embodiment adopts a numerical perturbation gradient approximation method;

[0100] Specifically, in each optimization cycle, the system selects a parameter to be optimized, such as a weighting coefficient. Apply a small, known perturbation to it. To obtain new parameters The system was run under these parameters for a short period of time, and the resulting change in the cost function was observed. The gradient of this parameter It can be passed To approximate;

[0101] Based on this numerical gradient, the system makes small updates to the parameters along the gradient descent direction to generate part of the model parameter correction instructions:

[0102]

[0103] in It is a learning rate, which is a preset hyperparameter used to control the step size of each update;

[0104] The system will perform this process sequentially on all adjustable parameters, including the weighting coefficients used in the risk prediction unit. and the response center point threshold used by the strategy adjustment unit and slope factor The model parameter correction instruction updates the corresponding parameter values ​​in the memory of the risk prediction unit and the strategy adjustment unit in a secure online manner through the system's internal configuration interface, so that the corrected model can take effect immediately in the next calculation cycle, thus completing the closed loop of the entire adaptive optimization.

[0105] By introducing a cost function and a numerical perturbation gradient approximation method, the system is provided with a powerful online, model-independent adaptive optimization capability. This means that the system can improve itself based on real-world feedback during actual operation without large-scale offline retraining, and automatically adapt to various unexpected environmental changes. The specific objects to be corrected are clearly defined, forming a direct feedback path from macroscopic performance to microscopic model parameters. This makes the optimization of the system highly targeted, ensuring the synergistic evolution of the entire perception-prediction-decision chain, and ultimately achieving the continuous optimization of the overall system performance.

[0106] To ensure stable operation of the system in real-world, complex environments, this system also includes preprocessing and verification mechanisms for input data. The data acquisition unit checks the validity of the collected raw data. If key data in a certain dimension is continuously missing or shows abnormal values ​​over a period of time, the risk prediction unit will temporarily use the historical average or moving average of that dimension as a substitute input and simultaneously generate a system alarm to ensure the continuity of risk index calculation. Furthermore, in the multimodal authentication protocol switching strategy, if the backup video stream license plate recognition protocol also becomes unusable, the system will execute the final contingency plan. For example, it will forcibly adjust the authentication request acceptance rate to a preset minimum security value to ensure the most basic access capacity at the entrance and prevent complete system failure, demonstrating the multi-layered redundancy and robustness of the design.

[0107] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A parking lot entrance / exit control system based on the Internet of Things, characterized in that, include: The data acquisition unit is used to collect the authentication signaling timestamps, the number of vehicles queuing at the entrance, the occupancy status of parking spaces, and network channel quality parameters of the parking lot in real time to generate multi-dimensional dynamic data. The risk prediction unit is used to receive multidimensional dynamic data and perform risk prediction processing to generate a system crash risk index. The strategy adjustment unit is used to receive the system crash risk index and compare it with the preset warning state threshold and danger state threshold to generate a risk level signal; the strategy adjustment unit is also used to generate control commands based on the risk level signal and the system crash risk index. The adaptive optimization unit is used to collect the average passage time at the entrance and the total throughput of the system after the control command is executed, and to perform adaptive optimization processing to generate model parameter correction commands for the risk prediction unit and the strategy adjustment unit. The process by which the risk prediction unit performs risk prediction processing to generate a system crash risk index includes: Based on multidimensional dynamic data, the jitter variance of authentication signaling timestamps, the instantaneous growth acceleration of entrance queue length, the global entropy of berth space turnover rate, and the Lyapunov index of expected arrival traffic flow are calculated. Based on the calculated jitter variance Instantaneous growth acceleration Global entropy And the Lyapunov exponent is used for nonlinear coupling calculations to generate a system collapse risk index. ; The coupling calculation is performed using the following formula: System crash risk index It is a dimensionless scalar; the higher its value, the greater the probability that the system will enter a state of computational resource deadlock. These are preset weighting coefficients. It is a constant characterizing the timescale of the system's crisis evolution.

2. The parking lot entrance and exit control system based on the Internet of Things according to claim 1, characterized in that, The process by which the strategy adjustment unit generates a risk level signal includes: When the system crash risk index is less than or equal to the warning status threshold, a safety risk level signal is generated. When the system crash risk index is greater than the warning state threshold and less than or equal to the danger state threshold, a warning risk level signal is generated. When the system crash risk index is greater than the danger state threshold, a danger risk level signal is generated.

3. The parking lot entrance and exit control system based on the Internet of Things according to claim 2, characterized in that, The control command is an authentication request acceptance rate adjustment command; The strategy adjustment unit responds to the early warning risk level signal or the dangerous risk level signal, and generates an authentication request acceptance rate adjustment instruction based on the logistic function model and the currently input system crash risk index; The instruction is generated based on a logistic function model and uses the current input system crash risk index. As a key variable, the specific authentication request acceptance rate Calculated using the following formula: Authentication request acceptance rate Defined as the upper limit of the authentication request processing rate that the system should currently execute, its unit is the number of requests per second, and its source is calculated by this formula; It represents the maximum physical processing rate of the system design, a constant preset based on hardware performance and network bandwidth; k is a dimensionless slope factor. These are the response center point thresholds; the sources of these two parameters are determined by nonlinear fitting of historical control data, i.e., data containing different risk indices. and the corresponding optimal acceptance rate On the historical data set, the optimal solution is obtained by using the least squares method. and value.

4. The parking lot entrance and exit control system based on the Internet of Things according to claim 2, characterized in that, The control command is a multimodal authentication protocol switching command; The strategy adjustment unit responds to the danger risk level signal and performs judgment processing. When the contribution of the jitter variance of the authentication signaling timestamp to the system crash risk index is greater than the preset contribution threshold, a multimodal authentication protocol switching instruction is generated.

5. The parking lot entrance and exit control system based on the Internet of Things according to claim 4, characterized in that, The multimodal authentication protocol switching instruction is used to switch the authentication mode from the Internet of Things protocol to a video stream-based license plate recognition protocol.

6. The parking lot entrance / exit control system based on the Internet of Things according to claim 1, characterized in that, The adaptive optimization unit performs adaptive optimization processing including: The cost function is calculated based on the average passage time at the entrance, the total system throughput, and the preset target passage time and target throughput. Based on the cost function, a numerical perturbation gradient approximation method is used for minimization to generate model parameter correction instructions.

7. The parking lot entrance and exit control system based on the Internet of Things according to claim 3, characterized in that, The model parameter correction command is used to correct the weight coefficients used by the risk prediction unit and the response center point threshold and slope factor used by the strategy adjustment unit.

Images