A scheduling method and system based on AI prediction of intelligent parking space idle state

Abstract

The application discloses a kind of based on AI prediction intelligent parking stall idle state scheduling method and its scheduling system, belong to big data technical field, including following steps: S1.data acquisition and pre-processing;S2.feature engineering construction;S3.AI model training;S4.operation and maintenance window recommendation.The application improves prediction accuracy by multi-source data fusion, extracts time characteristics, environmental characteristics and social characteristics from pre-processed data, fully depicts the space-time dependence of parking space and the correlation of external factors.High-reliability idle window is filtered by probability threshold, directly reducing the invalid attendance caused by temporary parking of vehicles, effectively improving the utilization rate of operation and maintenance resources.Based on the idle probability generated by AI prediction, the operation time window can be generated, and the path optimization algorithm is automatically matched to generate task scheduling instructions, realizing end-to-end automation from prediction to execution.

Description

Technical Field

[0001] This invention belongs to the field of big data technology, specifically relating to a scheduling method and system for predicting the vacancy status of intelligent berths based on AI. Background Technology

[0002] With the advancement of smart city construction, intelligent parking systems have become an important component of urban traffic management. Currently, mainstream intelligent parking facilities utilize geomagnetic sensors, Internet of Things (IoT) communication modules, and cloud management platforms to collect and transmit parking space occupancy data in real time. These systems, coupled with mobile applications, provide users with functions such as parking space lookup, navigation, and contactless payment. Such systems significantly improve parking space utilization and driver service experience, and are widely used in commercial areas, residential areas, and public parking lots, forming a complete technological ecosystem covering data collection, transmission, analysis, and application.

[0003] However, for parking space operation and maintenance management (such as geomagnetic equipment repair, parking space number restoration, and road maintenance), existing technologies still heavily rely on manual inspections or real-time geomagnetic monitoring data to schedule maintenance tasks. This model has significant drawbacks: First, real-time geomagnetic data can only reflect the current parking space occupancy status. Maintenance personnel may be unable to work due to temporary parking after arriving on-site, leading to frequent ineffective attendance. Second, frequent temporary scheduling forces repeated adjustments to maintenance plans, resulting in wasted manpower and time costs and overall low efficiency. Third, existing technologies lack the ability to deeply mine and analyze historical parking data, and cannot combine dynamic factors such as time, date, holidays, and special events to predict future available idle periods. Current solutions include geomagnetic induction-based monitoring systems that only provide static occupancy status and lack predictive capabilities; manual recording of parking patterns relies on experience and is difficult to adapt to dynamic changes; while video surveillance technology can acquire multi-dimensional information, it suffers from high deployment costs, susceptibility to environmental interference, and privacy risks.

[0004] In summary, existing parking space operation and maintenance scheduling technologies are insufficient to meet the needs of efficient management in smart cities. There is an urgent need to provide a method for predicting idle time periods based on historical geomagnetic data and artificial intelligence. By accurately predicting future work windows, this method can optimize the scheduling paths and timing of operation and maintenance tasks, thereby reducing ineffective attendance and improving the utilization rate of urban operation and maintenance resources. Summary of the Invention

[0005] To address the problems in related technologies, this invention provides a scheduling method and system based on AI-based prediction of intelligent parking space vacancy status, thereby solving the technical problems of high difficulty in the operation and maintenance management and scheduling of existing parking spaces, and difficulty in predicting the timing of operations.

[0006] The technical solution of this invention is implemented as follows: a scheduling method based on AI-predicted intelligent berth vacancy status, comprising the following steps:

[0007] S1. Data Acquisition and Preprocessing: Each smart berth is equipped with a geomagnetic sensor to acquire historical geomagnetic data of the target area, including occupancy / idle status, timestamp, and signal strength. External environmental data is also collected, including weather data, holiday information, and surrounding events. Outlier filtering and noise reduction are performed on the historical geomagnetic data.

[0008] S2. Feature Engineering Construction: Extract multidimensional features from the preprocessed data, including time features, environmental features, and social features. The time features include date type, time period segmentation, and periodic trends. The environmental features include temperature and rainfall. The social features include holiday markers and event markers.

[0009] S3. AI Model Training: A time-series prediction model is used to learn historical parking patterns. The time-series prediction model includes at least one of LSTM, Prophet, or XGBoost-Time. The multi-dimensional features are input to output the vacancy probability of each time period in the next N days.

[0010] S4. Operation and Maintenance Window Recommendation: Based on the idle probability, a workable time window is generated. When the idle probability is greater than a preset threshold, the corresponding time period is marked as a high-probability idle window, and an operation and maintenance task scheduling scheme is generated based on the path optimization algorithm.

[0011] As a further improvement to the above scheme, the data preprocessing in step S1 includes:

[0012] Abnormal occupancy status is filtered out. If the occupancy status of the same parking space continues for more than 24 hours and the signal strength is lower than the preset fault threshold, it is determined to be sensor fault data and removed.

[0013] Kalman filtering is used to reduce noise in signal strength data. The observation noise covariance matrix of the Kalman filter is dynamically adjusted according to the electromagnetic interference intensity to eliminate the influence of electromagnetic interference.

[0014] As a further improvement to the above scheme, the social characteristics in step S2 also include a dynamic adjustment factor. This dynamic adjustment factor is calculated based on the influence radius of the activity event and the parking mode offset caused by similar events in historical data. Specifically:

[0015] Δ=γ·R+(1-γ)·δ;

[0016] Δ is a dynamic adjustment factor used to correct the prediction results; R is the influence radius of the event; δ is the standard deviation of the historical parking pattern offset; γ is the weighting coefficient; where 0.6≤γ≤0.8.

[0017] As a further improvement to the above scheme, the training of the time series prediction model in step S3 further includes:

[0018] The dataset is divided into training, validation and test sets, and time series samples are generated using a sliding window mechanism.

[0019] During model training, an early stopping mechanism is introduced, which terminates training when the validation set loss does not decrease for 5 consecutive epochs to prevent overfitting, and the hyperparameters are optimized through cross-validation.

[0020] As a further improvement to the above scheme, step S3 uses an LSTM model to predict the idle probability, and its output formula is:

[0021] P t =σ(W p ·[h t-1 ,x t ]+b p );

[0022] Among them, P t Let h be the idle probability during time period t. t-1 Let x be the state of the LSTM hidden layer at the previous time step. t Let W be the feature vector for time period t. p Let b be the weight matrix. p σ is the bias term, and σ is the Sigmoid activation function.

[0023] As a further improvement to the above scheme, the feature vector x t The construction includes: time-encoded vector T t Environmental feature vector E t Social feature vector S t And satisfy:

[0024] x t =T t ⊕E t ⊕S t ;

[0025] Where ⊕ represents the vector concatenation operation, T t Including the day of the week, hour, and whether it is a peak period, E t This includes normalized values ​​of temperature and rainfall, with a normalization range of 0-1, S t This includes binary tags for holidays and events.

[0026] As a further improvement to the above scheme, the path optimization algorithm in step S4 adopts a genetic algorithm, and its fitness function is defined as:

[0027]

[0028] Where F is the fitness value, P i Let D be the idle probability of the i-th maintenance point. j Let be the driving distance of the j-th segment of the path, α and β be weighting coefficients, and satisfy α+β=1; M is the total number of operation and maintenance points to be scheduled, and N is the total driving distance of the segments in the operation and maintenance path.

[0029] A scheduling system based on AI-predicted intelligent berth vacancy status applies the aforementioned scheduling method based on AI-predicted intelligent berth vacancy status; including:

[0030] The data acquisition module is used to acquire historical data from the geomagnetic sensor and external environmental data, and to perform data cleaning and noise reduction.

[0031] The feature extraction module, connected to the data acquisition module, is used to construct multi-dimensional features of time, environment, and society.

[0032] The AI ​​prediction module is connected to the feature extraction module and uses a time-series prediction model to output the probability of future idle time.

[0033] The scheduling optimization module is connected to the AI ​​prediction module and generates operation and maintenance task scheduling instructions based on idle probability and path constraints.

[0034] As a further improvement to the above solution, the AI ​​prediction module further includes a model update unit, which periodically receives new data and triggers incremental training. The incremental training uses an online learning algorithm to update the model parameters, with an update frequency of once a week.

[0035] As a further improvement to the above solution, the scheduling optimization module also includes a visualization interaction unit. The visualization interaction unit is used to mark high-probability idle windows and recommended paths on the map, and supports manual adjustment of scheduling priorities. The priority adjustment is dynamically configured based on the urgency of the operation and maintenance task and the idle probability threshold.

[0036] Compared with the prior art, the present invention has the following advantages:

[0037] (1) Multi-source data fusion improves prediction accuracy: Historical geomagnetic data such as occupancy / idle status, timestamps and signal strength are collected by geomagnetic sensors and external environmental data such as weather, holidays and surrounding events are integrated to build a multi-dimensional data input foundation. Combined with outlier filtering and Kalman filtering noise reduction, sensor noise and environmental interference are effectively eliminated, ensuring the integrity and reliability of training data, providing high-precision input for AI models, significantly enhancing the prediction confidence of future idle periods, and avoiding misjudgment problems caused by data deviation.

[0038] (2) Multidimensional feature engineering adapts to dynamic parking patterns: Time features, environmental features, and social features are extracted from preprocessed data to comprehensively depict the spatiotemporal dependence of parking spaces and their correlation with external factors. Through multidimensional feature combination, the AI ​​model can accurately capture parking behavior patterns in different scenarios, such as commuting peak hours and holiday traffic fluctuations, dynamically adapt to complex changes in urban operation and maintenance, and improve the robustness of idle window prediction.

[0039] (3) Time-series prediction model drives intelligent operation and maintenance decision-making: Using time-series prediction models such as LSTM, Prophet, or XGBoost-Time, based on historical patterns and multi-dimensional feature learning, the idle probability of each time period in the next N days is output. The traditional passive operation and maintenance is upgraded to an active prediction mode. High-reliability idle windows are selected through probability thresholds, which directly reduces invalid attendance caused by temporary vehicle parking and effectively improves the utilization rate of operation and maintenance resources.

[0040] (4) Closed-loop scheduling optimization reduces operation and maintenance costs: Based on the idle probability predicted by AI, a workable time window is generated, and the task scheduling instructions are automatically generated by matching the path optimization algorithm, realizing end-to-end automation from prediction to execution. It breaks through the inefficiency bottleneck of traditional manual experience scheduling, reduces the frequency of temporary plan adjustments, effectively shortens the execution time of a single operation and maintenance task, significantly reduces manpower, fuel and time costs, and meets the refined needs of smart cities for efficient operation and maintenance. Attached Figure Description

[0041] Figure 1 A flowchart illustrating the steps of a scheduling method based on AI-predicted intelligent berth vacancy status provided in an embodiment;

[0042] Figure 2 This is an architecture diagram of a scheduling system based on AI-predicted intelligent berth vacancy status, provided as an example. Detailed Implementation

[0043] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0044] Example:

[0045] like Figure 1 As shown, a scheduling method based on AI-predicted intelligent berth vacancy status includes the following steps:

[0046] S1. Data Acquisition and Preprocessing: Each smart berth is equipped with a geomagnetic sensor. In this embodiment, a geomagnetic sensor of model MLX90393 is used to acquire historical geomagnetic data of the target area, including occupancy / idle status (e.g., 0 / 1, where 0 represents occupancy and 1 represents idleness), timestamp (accurate to the second), signal strength (dBm), and sensor temperature (°C). The sampling frequency is configured to 1Hz. External environmental data is also collected, including weather data, holiday information, and surrounding events. Outlier filtering and noise reduction are performed on the historical geomagnetic data. Specifically, weather data such as rainfall and temperature from the meteorological bureau, holiday lists published by the government, and information on large-scale surrounding events such as concerts and sporting events provided by the city's event management platform can be accessed through open APIs.

[0047] In this embodiment, the data preprocessing in step S1 includes:

[0048] S11. Outlier Filtering: Filtering out abnormal occupancy status. If the occupancy status of the same parking space continues for more than 24 hours and the signal strength is lower than the preset fault threshold, it is determined to be sensor fault data and removed. Specifically, in this embodiment, the signal strength below -90dB is set as the preset fault threshold.

[0049] S12. Kalman Filter Noise Reduction: Kalman filtering is used to reduce noise in the signal strength data. The observation noise covariance matrix of the Kalman filter is dynamically adjusted according to the electromagnetic interference intensity to eliminate the influence of electromagnetic interference. Specifically, a state transition matrix can be set: The observation matrix H = [1 0], and the observation noise covariance matrix R is dynamically adjusted according to the electromagnetic interference intensity. For example, R = 5 when the interference is strong and R = 1 when the interference is weak. By fusing geomagnetic data with external environmental data and using dual-condition filtering of outliers and dynamic noise reduction, sensor noise and environmental interference are eliminated, improving the signal-to-noise ratio of the input data and providing a high-precision data foundation for subsequent model training. In this embodiment, historical geomagnetic data such as occupancy / idle status, timestamps, and signal strength are collected by geomagnetic sensors, and external environmental data such as weather, holidays, and surrounding events are fused to construct a multi-dimensional data input foundation. Combined with outlier filtering and Kalman filtering noise reduction, sensor noise and environmental interference are effectively eliminated, ensuring the integrity and reliability of the training data, providing high-precision input for the AI ​​model, significantly enhancing the prediction confidence of future idle periods, and avoiding misjudgment problems caused by data deviation.

[0050] S2. Feature Engineering Construction: Extracting multidimensional features from the preprocessed data, including time features, environmental features, and social features;

[0051] The time characteristics include date type, time period segmentation, and periodic trend; specifically, the date type includes weekday / weekend / holiday, using binary encoding; the time period segmentation includes morning peak, off-peak, and evening peak, using one-hot encoding; the periodic trend can be the average occupancy rate of the same time period over the past 7 days.

[0052] The environmental characteristics include temperature and rainfall; specifically, temperature is normalized to 0-1 and rainfall is normalized to 0-1.

[0053] The social features include holiday markers and event markers; specifically, holiday markers are set to 1 / 0; for event markers, such as during a concert, the marker is 1. In this embodiment, the social features in step S2 also include a dynamic adjustment factor, which is calculated based on the influence radius of the event and the parking mode offset caused by similar events in historical data, specifically:

[0054] Δ=γ·R+(1-γ)·δ;

[0055] Δ is a dynamic adjustment factor used to correct the prediction results; R is the influence radius of the event in kilometers; δ is the standard deviation of the historical parking pattern offset; γ is the weighting coefficient; where 0.6≤γ≤0.8.

[0056] In this embodiment, δ is calculated as follows:

[0057]

[0058] Where, x idenoted as , where μ is the average parking rate during similar historical events, and N is the number of data samples.

[0059] For example, if there is a large concert event in the target area, the parameters are set as follows:

[0060] R = 2 kilometers, the scope of the event's impact will be provided by the organizer;

[0061] δ = 0.5, meaning the standard deviation of parking rates during similar historical concerts is 0.5.

[0062] γ = 0.7, which focuses more on the impact of real-time activities;

[0063] The dynamic adjustment factor Δ = 0.7 × 2 + (1 - 0.7) × 0.5 = 1.4 + 0.15 = 1.55. The dynamic adjustment factor Δ = 1.55 will be added to the social characteristics, indicating that the probability of parking spaces being vacant within 2 kilometers of the concert needs to be adjusted downwards. Since the concert attracts a large number of vehicles, the adjustment range is controlled by Δ weighting.

[0064] For example, if the target area only has small community events, such as markets, then the parameters should be set as follows:

[0065] R = 0.5 km, the activity only affects a radius of 500 meters.

[0066] δ = 0.3, meaning that parking fluctuations for similar historical events are relatively small;

[0067] γ = 0.8, which is more dependent on the real-time activity range;

[0068] The dynamic adjustment factor Δ = 0.8 × 0.5 + (1 - 0.8) × 0.3 = 0.4 + 0.06 = 0.46. Δ = 0.46 indicates that the impact of the activity is limited, and the model only fine-tunes the idle probability of local areas.

[0069] As can be seen from the above examples, the dynamic adjustment factor Δ, by integrating real-time activity range and historical fluctuation data, enables the AI ​​model to dynamically correct prediction results. It is especially suitable for complex scenarios such as holidays and large-scale events, taking into account both scientific rigor and practicality, and significantly improving prediction robustness and operation and maintenance scheduling efficiency.

[0070] S3. AI Model Training: A time-series prediction model is used to learn historical parking patterns. The time-series prediction model includes at least one of LSTM, Prophet, or XGBoost-Time. The multi-dimensional features are input to output the idle probability of each time period in the next N days. In this embodiment, the training of the time-series prediction model in step S3 further includes: dividing the dataset into a training set, a validation set, and a test set, and generating time-series samples using a sliding window mechanism. During the model training process, an early stopping mechanism is introduced. Training is terminated when the validation set loss does not decrease for 5 consecutive epochs to prevent overfitting, and hyperparameters are optimized through cross-validation.

[0071] This embodiment uses an LSTM model as an example, and specifically includes the following process:

[0072] S31. Model Selection and Training:

[0073] An LSTM model is used, consisting of a 64-node input layer, a 32-node hidden layer, and a 1-node output layer. The input feature vector is x. t =T t ⊕E t ⊕S t Among them, T t The time-coded vector includes the day of the week, hour, and one-hot encoding of whether it is a peak period; E t This is an environmental feature vector, including normalized values ​​of temperature and rainfall, with a normalization range of 0-1, S... t is a social feature vector, including binary tags for holidays and events; ⊕ indicates a vector concatenation operation.

[0074] The dataset is divided into training, validation and test sets in a 7:2:1 ratio, and time series samples are generated using a sliding window mechanism (window length of 7 days, step size of 1 hour).

[0075] An early stopping mechanism is introduced, terminating training when the validation set loss fails to decrease for five consecutive epochs, and 5-fold cross-validation is used to optimize hyperparameters. It's important to note that the early stopping mechanism is a strategy to prevent overfitting. During training, by monitoring changes in the validation set loss, training is terminated early when the loss value stops decreasing within a certain number of epochs, preventing the model from overfitting on the training set and losing its generalization ability.

[0076] For example, suppose we train an LSTM model to predict the probability of a parking space being vacant:

[0077] Epoch 1: Validation set loss = 0.25;

[0078] Epoch 2: Validation set loss = 0.20;

[0079] Epoch 3: Validation set loss = 0.18;

[0080] Epoch 4: Validation set loss = 0.17;

[0081] Epoch 5: Validation set loss = 0.17;

[0082] Epoch 6: Validation set loss = 0.17;

[0083] From Epoch 4 to Epoch 6, the validation set loss did not decrease for three consecutive times (but did not reach five times), so training continued.

[0084] Epoch 7: Validation set loss = 0.18 (starting to rise);

[0085] Epoch 8: Validation set loss = 0.19;

[0086] Epoch 9: Validation set loss = 0.20;

[0087] If the model fails to decrease for five consecutive epochs from epoch 5 to epoch 9 (resulting in an increase in actual loss), training terminates at the 9th epoch, retaining the model parameters from epoch 4. At this point, the validation loss is at its lowest, 0.17. Introducing an early stopping mechanism avoids overfitting and prevents the model from excessively memorizing noise from the training data. Furthermore, it saves computational resources and reduces ineffective training epochs.

[0088] Five-fold cross-validation is a hyperparameter optimization method that divides the dataset into five equal parts, using four parts as the training set and one part as the validation set alternately, repeating the training five times to evaluate the performance of different hyperparameter combinations. The hyperparameters with the best average performance are then selected, avoiding the influence of randomness from a single validation set. Combined with early stopping, it can efficiently optimize the model and prevent overfitting, significantly improving prediction accuracy.

[0089] Specifically, the implementation steps include the following:

[0090] S311. Data partitioning: Randomly divide the training set (excluding the test set) into 5 parts (Fold 1 to 5).

[0091] S312. Cross-validation:

[0092] Round 1: Fold 1-4 training, Fold 5 verification.

[0093] Round 2: Fold 1-3 + Fold 5 training, Fold 4 verification.

[0094] Repeat this process once for each of the five samples, using each sample as a validation set.

[0095] S313. Hyperparameter selection: Calculate the average value of the results from 5 rounds of validation and select the hyperparameter combination with the best average performance.

[0096] S32. Idle probability output:

[0097] The LSTM model is used for idle probability prediction, and its output formula is as follows:

[0098] P t =σ(W p ·[h t-1 ,x t ]+b p );

[0099] Among them, P t h is the idle probability during time period t; t-1 The previous state of the LSTM hidden layer; x t W is the feature vector for time period t; p The weight matrix has dimensions 32x1; b p Here, σ is the bias term, σ is the sigmoid activation function, and the output is the idle probability P for each time period within the next 3 days. t ∈[0,1].

[0100] S4. Operation and Maintenance Window Recommendation: Based on the idle probability, a workable time window is generated. When the idle probability is greater than a preset threshold, the corresponding time period is marked as a high-probability idle window, and an operation and maintenance task scheduling scheme is generated based on the path optimization algorithm.

[0101] The path optimization algorithm in this embodiment uses a genetic algorithm, specifically including the following steps:

[0102] S41. Window Generation: Sets the idle probability threshold (default 90%). When P t A value greater than 0.9 is marked as a high-probability idle window.

[0103] S42. Path Optimization: A genetic algorithm is used (population size 50, iterations 100), with weight coefficients α and β of 0.7 and 0.3 respectively; the fitness function is:

[0104]

[0105] P i Let D be the idle probability of the i-th maintenance point. j Let M be the travel distance of the j-th segment of the path, M be the total number of maintenance points to be scheduled, and N be the total travel distance of each segment in the maintenance path.

[0106] For example, a commercial area needs to perform geomagnetic inspections on 5 parking spaces (A, B, C, D, E). The AI-predicted vacancy probability and path distance are shown in the table below:

[0107] Operation and maintenance points <![CDATA[Idle probability P i > <![CDATA[Adjacent path segment distance D j (km)]]> A 0.95 A→B: 1.2 B 0.90 B→C:0.8 C 0.85 C→D: 1.5 D 0.92 D→E:0.7 E 0.88 E→A: 2.0 (Closed-loop path)

[0108] The candidate path schemes are compared in the table below:

[0109]

[0110] The above comparative analysis shows that the fitness values ​​of the sequential path and the optimal path are the same (F = 3.221), but the optimal path, by adjusting the access order and reducing the number of turns or complex road sections, such as avoiding congested areas, has a higher actual execution efficiency. The optimized path, due to its longer total distance (5.0 km), has the lowest fitness value (F = 3.210) and is therefore eliminated by the algorithm. This embodiment generates a workable time window based on the idle probability predicted by AI and automatically matches the path optimization algorithm to generate task scheduling instructions, achieving end-to-end automation from prediction to execution. This ensures that maintenance personnel can start working immediately upon arrival at the site. The configuration of the fitness function makes the algorithm naturally favor shorter paths. It also possesses a certain degree of dynamic adaptability; if the idle probability of a certain point drops sharply, such as due to temporary activities causing P... i If the value is 0.6, the algorithm will automatically remove that point, replan the route, and reduce manual intervention. The path optimization algorithm uses a genetic algorithm, which breaks through the inefficiency bottleneck of traditional manual experience scheduling, reduces the frequency of temporary plan adjustments, effectively shortens the execution time of a single operation and maintenance task, significantly reduces manpower, fuel, and time costs, and meets the refined needs of smart cities for efficient operation and maintenance.

[0111] like Figure 2 As shown, this embodiment also provides a scheduling system based on AI-predicted intelligent berth vacancy status, which applies the scheduling method based on AI-predicted intelligent berth vacancy status described above; including:

[0112] The data acquisition module is used to acquire historical data from the geomagnetic sensor and external environmental data, and to perform data cleaning and noise reduction; specifically, it can receive sensor data via the MQTT protocol and store it in a MySQL database.

[0113] The feature extraction module, connected to the data acquisition module, is used to construct multi-dimensional features of time, environment, and society.

[0114] The AI ​​prediction module, connected to the feature extraction module, outputs the future idle probability using a time-series prediction model. The AI ​​prediction module further includes a model update unit, which periodically receives new data and triggers incremental training. This incremental training uses an online learning algorithm to update model parameters, with an update frequency of once a week. Specifically, an LSTM model is deployed based on the TensorFlow framework, with incremental training triggered weekly and a learning rate of 0.001.

[0115] The scheduling optimization module, connected to the AI ​​prediction module, generates operation and maintenance task scheduling instructions based on idle probability and path constraints. The scheduling optimization module also includes a visualization interaction unit, which marks high-probability idle windows and recommended paths on a map and supports manual adjustment of scheduling priorities. The priority adjustment is dynamically configured based on the urgency of the operation and maintenance task and an idle probability threshold. Specifically, the Leaflet.js map library is used to display high-probability windows and recommended paths, and manual priority adjustment is supported, such as setting the urgency threshold to 95%.

[0116] Based on the disclosure and teachings of the foregoing specification, those skilled in the art can make changes and modifications to the above embodiments. Therefore, the present invention is not limited to the specific embodiments disclosed and described above, and some modifications and changes to the invention should also fall within the protection scope of the claims of the present invention. Furthermore, although some specific terms are used in this specification, these terms are only for convenience of explanation and do not constitute any limitation on the present invention.

Claims

1. A scheduling method based on AI-predicted intelligent berth vacancy status, characterized in that, Includes the following steps: S1. Data Acquisition and Preprocessing: Each smart berth is equipped with a geomagnetic sensor to acquire historical geomagnetic data of the target area, including occupancy / idle status, timestamp, and signal strength. External environmental data is also collected, including weather data, holiday information, and surrounding events. Outlier filtering and noise reduction are performed on the historical geomagnetic data. S2. Feature Engineering Construction: Extract multidimensional features from the preprocessed data, including time features, environmental features, and social features. The time features include date type, time period segmentation, and periodic trends. The environmental features include temperature and rainfall. The social features include holiday markers and event markers. The social characteristics also include a dynamic adjustment factor, which is calculated based on the influence radius of the event and the parking pattern offset caused by similar events in historical data. Specifically: Δ=γ·R+(1 c) ·d; Δ is a dynamic adjustment factor used to correct the prediction results; R is the influence radius of the event; δ is the standard deviation of the historical parking pattern offset; γ is the weighting coefficient; where 0.6≤γ≤0.8; S3. AI Model Training: A time-series prediction model is used to learn historical parking patterns. The time-series prediction model includes at least one of LSTM, Prophet, or XGBoost-Time. The multidimensional features are input to output the vacancy probability of each time period in the next N days. S4. Operation and maintenance window recommendation: Generate a workable time window based on the idle probability. When the idle probability is greater than a preset threshold, mark the corresponding time period as a high-probability idle window and generate an operation and maintenance task scheduling scheme based on the path optimization algorithm. The path optimization algorithm uses a genetic algorithm, and its fitness function is defined as: ； Where F is the fitness value, P i Let D be the idle probability of the i-th maintenance point. j Let be the driving distance of the j-th segment of the path, α and β be weighting coefficients, and satisfy α+β=1; M is the total number of operation and maintenance points to be scheduled, and N is the total driving distance of the segments in the operation and maintenance path.

2. The scheduling method based on AI-predicted intelligent berth vacancy status according to claim 1, characterized in that, The data preprocessing in step S1 includes: Abnormal occupancy status is filtered out. If the occupancy status of the same parking space continues for more than 24 hours and the signal strength is lower than the preset fault threshold, it is determined to be sensor fault data and removed. Kalman filtering is used to reduce noise in signal strength data. The observation noise covariance matrix of the Kalman filter is dynamically adjusted according to the electromagnetic interference intensity to eliminate the influence of electromagnetic interference.

3. The scheduling method based on AI-predicted intelligent berth vacancy status according to claim 1, characterized in that, Step S3, training the time series prediction model, further includes: The dataset is divided into training, validation and test sets, and time series samples are generated using a sliding window mechanism. During model training, an early stopping mechanism is introduced, which terminates training when the validation set loss does not decrease for 5 consecutive epochs to prevent overfitting, and the hyperparameters are optimized through cross-validation.

4. The scheduling method based on AI-predicted intelligent berth vacancy status according to claim 1, characterized in that, In step S3, an LSTM model is used to predict the idle probability, and its output formula is: P t =σ(W p ·[h t 1,x t ]+b p ); Among them, P t Let h be the idle probability during time period t. t 1 represents the previous state of the LSTM hidden layer, x t Let W be the feature vector for time period t. p Let b be the weight matrix. p σ is the bias term, and σ is the Sigmoid activation function.

5. The scheduling method based on AI-predicted intelligent berth vacancy status according to claim 4, characterized in that, The feature vector x t The construction includes: time-encoded vector T t Environmental feature vector E t Social feature vector S t And satisfy: x t =T t ⊕E t ⊕S t ; Where ⊕ represents the vector concatenation operation, T t Including the day of the week, hour, and whether it is a peak period, E t This includes normalized values ​​of temperature and rainfall, with a normalization range of 0-1, S t This includes binary tags for holidays and events.

6. A scheduling system based on AI-predicted intelligent berth vacancy status, employing the scheduling method based on AI-predicted intelligent berth vacancy status as described in any one of claims 1-5; characterized in that, include: The data acquisition module is used to acquire historical data from the geomagnetic sensor and external environmental data, and to perform data cleaning and noise reduction. The feature extraction module, connected to the data acquisition module, is used to construct multi-dimensional features of time, environment, and society. The AI ​​prediction module is connected to the feature extraction module and uses a time-series prediction model to output the probability of future idle time. The scheduling optimization module is connected to the AI ​​prediction module and generates operation and maintenance task scheduling instructions based on idle probability and path constraints.

7. A scheduling system based on AI-predicted intelligent berth vacancy status according to claim 6, characterized in that, The AI ​​prediction module further includes a model update unit, which periodically receives new data and triggers incremental training. The incremental training uses an online learning algorithm to update the model parameters, with an update frequency of once a week.

8. A scheduling system based on AI-predicted intelligent berth vacancy status according to claim 7, characterized in that, The scheduling optimization module also includes a visualization interaction unit, which is used to mark high-probability idle windows and recommended paths on the map, and supports manual adjustment of scheduling priorities. The priority adjustment is dynamically configured based on the urgency of the operation and maintenance task and the idle probability threshold.

Images