Bus station man-machine interaction system

Abstract

The invention relates to a bus station man-machine interaction system, which relates to the technical field of traffic information processing, and comprises an interaction device arranged at a bus station, a server side and a server side, analyzing the interaction information to obtain query target information; and determining corresponding feedback information based on the query target information, and feeding back the feedback information to the interaction equipment. By deploying the interactive equipment at the bus station, the defect that the technology depends on manual updating and static display and cannot respond to the dynamic change of the bus and the diversified requirements of passengers in real time is overcome.

Description

Technical Field

[0001] The present invention relates to the technical field of traffic information processing, and particularly to a human-computer interaction system for bus stops. Background Art

[0002] The human-computer interaction display at bus stops emerged against the backdrop of the accelerating urbanization process and the rapid development of intelligent transportation systems, aiming to meet the needs of modern passengers for real-time traffic information and diverse services. With the increase in urban population density and the rising demand for public transportation, the traditional static display method of bus information can no longer meet the passengers' need for timely and accurate information. Summary of the Invention

[0003] The main object of the present invention is to provide a human-computer interaction system for bus stops to solve the deficiencies in the related technologies.

[0004] To achieve the above object, according to the first aspect of the present invention, a human-computer interaction system for bus stops is provided, including an interaction device disposed at the bus stop. The interaction device includes a touch module, a controller, a sensor, a display module, and a data transmission interface. The touch module is used to detect the touch operation of the user; the controller generates specified interaction information based on the interaction operation; the sensor is used to detect the change parameters of the environment around the interaction device and send the change parameters to the controller for the controller to adjust the display module. Among them, the touch operation includes a manual query operation and a voice query operation; the interaction device obtains real-time bus data from a third-party database through the data transmission interface and displays it through the display module; a server, which obtains interaction information from the interaction device; obtains query purpose information parsed from the interaction information; determines corresponding feedback information based on the query purpose information and feeds back the feedback information to the interaction device.

[0005] Optionally, the system further includes a positioning device disposed on the bus for real-time positioning of the bus; if the query purpose information is train number query information, real-time traffic data on the line of the train number to be queried is obtained through a preset data transmission mechanism; the real-time traffic data is input into a prediction model for prediction to obtain the current arrival time of the train number to be queried. Among them, the real-time traffic data includes the real-time positioning information of the bus obtained from the positioning device on the bus of the train number to be queried, the departure time of the bus of the train number to be queried obtained from a specified database, the real-time traffic flow and road condition data obtained from a specified database, and data on specified dimensions affecting traffic.

[0006] Optionally, the interaction device further includes an image acquisition module, configured to acquire images of the bus stop through a camera device, and transmit the image acquisition device to the server through a preset data transmission mechanism for the server to perform anomaly discrimination based on the acquired images; and / or, the image acquisition module acquires images inside the bus stop, and calculates the real-time number of people inside the stop through the controller.

[0007] Optionally, the interaction device further includes a key module. When the key module is triggered, the controller generates an alarm message and uploads the alarm message to the alarm system.

[0008] Optionally, when training the prediction model, the method includes: obtaining the historical traffic data of all buses on the route of the train to be queried; processing the historical traffic data to obtain time series data as sample data; inputting the time series data into the prediction model to be trained, and using the arrival times corresponding to different data at each time point as outputs to train the prediction model to be trained. Wherein, the time series includes the historical arrival times corresponding to the specified time feature points and the specified factor information corresponding to the feature time points; the loss function is where k is the given sample data, P is the number of neurons in the output layer of the prediction model; n = 1, 2, 3......P; X nk is the actual arrival time of sample k; t nk is the target expected value; q is a constant, and the value of q is 1 / 2.

[0009] Optionally, after the server continuously obtains the new arrival times of different trains and the information of different factors affecting the new arrival times, an incremental learning method is used to optimize the prediction model.

[0010] Optionally, if the query purpose is the passenger flow query information of the bus stop, the server calls the detection device set inside the bus stop or on the interaction device to detect the number of passengers at the bus stop; obtains the weather data, the current corresponding event information, the current corresponding seasonal data, and the current traffic flow data; inputs the passenger data, the weather data, the current corresponding event information, the current corresponding seasonal data, and the current traffic flow data into the traffic flow prediction model to predict the passenger flow at the bus stop in a future period of time; and sends the prediction result to the interaction device.

[0011] Optionally, when training the traffic prediction model, the method includes: obtaining the actual passenger flow data at different time points, as well as the time feature data corresponding to each different time point, the weather data affecting the passenger flow corresponding to each different time point, the event information corresponding to each different time point, the seasonal change data corresponding to each different time point, and the historical traffic flow data corresponding to each different time point, so as to obtain the passenger flow sample data; preprocessing the passenger flow sample data and inputting it into a machine learning model, and using the actual passenger flow within a specified time period after each different time point as the output to train the machine learning model.

[0012] Optionally, the system further includes an edge computing device disposed near the bus stop to assist the interaction device in interacting with the server.

[0013] The bus stop human-computer interaction system in this embodiment includes: an interaction device disposed at the bus stop, where the interaction device includes a touch module, a controller, a sensor, a display module, and a data transmission interface. The touch module is used to detect the user's touch screen operation; the controller generates specified interaction information based on the interaction operation; the sensor is used to detect the environmental change parameters around the interaction device and send the change parameters to the controller so that the controller can adjust the display module. Among them, the touch screen operation includes a manual query operation and a voice query operation; the interaction device obtains real-time bus data from the database through the data transmission interface and displays it through the display module; a server, which obtains interaction information from the interaction device; query purpose information obtained by parsing the interaction information; determines corresponding feedback information based on the query purpose information, and feeds back the feedback information to the interaction device. By deploying the interaction device at the bus stop, it overcomes the technical dependence on manual updates and static displays and cannot respond in real time to the dynamic changes of buses and the diverse needs of passengers. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following will briefly introduce the drawings required for use in the description of the specific embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0015] Figure 1 It is a schematic diagram of the interaction principle within the bus stop human-computer interaction system according to an embodiment of the present invention;

[0016] Figure 2 It is a schematic diagram of the functional framework of the bus stop human-computer interaction system according to an embodiment of the present invention;

[0017] Figure 3 It is a schematic diagram of the system working process of the bus stop human-computer interaction system according to an embodiment of the present invention. Specific embodiments

[0018] In order to enable those skilled in the art to better understand the solution of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative work shall fall within the protection scope of the present invention.

[0019] It should be noted that the terms "first", "second", etc. in the specification of the present invention and the above-mentioned drawings are used to distinguish similar objects, and do not necessarily need to be used to describe a specific order or sequence. It should be understood that such data can be interchanged under appropriate circumstances so as to implement the embodiments of the present invention described here. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device that includes a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0020] It should be noted that, without conflict, the embodiments in the present invention and the features in the embodiments can be combined with each other. The present invention will be described in detail below with reference to the drawings and in conjunction with the embodiments.

[0021] According to an embodiment of the present invention, a bus stop human-computer interaction system is provided, as Figure 1 shown, including an interaction device disposed on a bus stop, the interaction device including a touch module, a controller, a sensor, a display module, and a data transmission interface. The touch module is used to detect a user's touch screen operation; the controller generates specified interaction information based on the interaction operation; the sensor is used to detect change parameters of the environment around the interaction device and send the change parameters to the controller so that the controller adjusts the display module. Among them, the touch screen operation includes a manual query operation and a voice query operation; the interaction device obtains real-time bus data from a database through the data transmission interface and displays it through the display module; a server, obtains interaction information from the interaction device; query purpose information obtained by parsing the interaction information; determines corresponding feedback information based on the query purpose information, and feeds back the feedback information to the interaction device.

[0022] In this embodiment, referring to Figure 1, The touch module may include a touch screen that allows users to directly interact with the system. For example, for query operations, bus information can be queried, route maps can be viewed, etc. The information queried can be information pre-stored in the storage device of the interaction device or real-time bus data called from a database, such as a third-party database.

[0023] In the user interface, the interface design needs to be simple and intuitive. It usually includes a touch screen operation interface and a button operation interface to facilitate users to query bus information. When displaying information, the content that can be displayed includes the arrival time of buses, route information, station locations, transfer suggestions, etc.

[0024] The controller is responsible for processing input requests from users, managing the displayed content, and exchanging data with the server. The controller in this embodiment is an embedded system or an industrial computer. The operating system running on the controller may be Linux, Windows or an embedded operating system, which is responsible for the basic management functions of the system.

[0025] The display module may include a display screen. Generally, LED or LCD display screens are used to display information. The size and resolution of the display screen determine the display effect and readability of the information. Real-time data of each bus can be obtained from a third-party database, such as the database of the bus company, through the data transmission interface. The display module can display specified information and information feedback from the server; it is also used to obtain advertisement information to be delivered from the server and display the advertisement information to be delivered.

[0026] The server belongs to the background management system and can be used to update bus information, maintain data, and monitor the system status. The display device needs to be maintained and calibrated regularly to ensure the accuracy of the information and the normal operation of the device.

[0027] In implementation, an infrared sensor can be utilized to detect the presence and actions of passengers and automatically activate the display screen or touch screen. Camera: Monitor the platform environment, identify passenger behaviors and conduct data analysis. Touch screen using capacitive screen or infrared touch screen: Provide an intuitive user interface where passengers can query bus information through touch operations. Wireless communication technologies, Wi-Fi, 4G / 5G: Achieve real-time data transmission between the bus platform and the central server to ensure timely information updates. Bluetooth: Conduct short-distance communication with passengers' mobile devices, such as ticket information, personalized services, etc. Information display technologies, LED display screen: Used to display real-time bus arrival information, weather forecasts, advertisements, etc. Electronic ink screen: Can clearly display even under direct sunlight and is suitable for energy-saving display scenarios. Data processing and cloud computing, data collection and analysis: Collect data through sensors and cameras, conduct big data analysis, and predict bus arrival times, passenger flow, etc. Cloud computing: Provide powerful computing and storage capabilities to support real-time data processing and information release. Speech recognition and synthesis technologies, speech recognition: Passengers can query information through voice commands, improving the usability of the system. Speech synthesis: The system can announce real-time bus information, stop reminders, etc. through voice, facilitating the use by visually impaired people. Navigation system: Assist passengers in planning travel routes and provide information on services around stops. These technologies together constitute the human-machine interaction display method for bus platforms, improving the efficiency of bus services and the satisfaction of passengers through intelligent and user-friendly designs.

[0028] As an optional implementation manner of this embodiment, the system further includes a positioning device disposed on the bus for real-time positioning of the bus. Among them, the information of the real-time positioning is uploaded to a specified database; if the query destination information is train number query information, the real-time traffic data on the line of the train to be queried is obtained through a preset data transmission mechanism; the real-time traffic data is input into a prediction model for prediction to obtain the current arrival time of the train to be queried. Among them, the real-time traffic data includes the real-time positioning information of the bus detected by the positioning device on the bus of the train to be queried, the departure time of the bus of the train to be queried obtained from the specified database, the real-time traffic flow and road condition data obtained from the specified database, and the data of the specified dimension affecting traffic.

[0029] As an optional implementation manner of this embodiment, when training the prediction model, the method includes: obtaining historical traffic data of all buses of the train on the route to be queried; processing the historical traffic data to obtain time series data as sample data; inputting the time series data into the prediction model to be trained, and using the arrival times corresponding to different data at each time point as outputs to train the prediction model to be trained, where the time series data includes the historical arrival times corresponding to specified time feature points and the specified factor information corresponding to the feature time points; the loss function of the prediction model is where k is the given sample data, P is the number of neurons in the output layer of the prediction model; n = 1, 2, 3......P; X nk is the actual arrival time of sample k; t nk is the target expected value; q is a constant, and the value of q is 1 / 2.

[0030] In this optional implementation manner, when constructing the time series data, core data is obtained, including basic data, bus line number, station identifier, section to which the line belongs, planned arrival time, actual arrival time (timestamp format); time data, such as date, hour, minute, day of the week, holiday / activity day; traffic flow data, such as average speed of the section, congestion index, etc.; environmental data such as weather condition data, traffic event data, etc.; special event data such as construction data, accident data, etc.

[0031] After obtaining the above data, space-time alignment is performed, with X minutes as the time slice granularity, to align the basic data, traffic flow data, environmental data, and special event data. After alignment, specified features in the data are encoded, and the model is trained based on the encoded specified features and the encoded actual arrival times.

[0032] Use multiple LSTM layers or combine convolutional layers to extract features, and finally use a fully connected layer to output the prediction results. Additionally, the attention mechanism may help the model focus on important time points or external factors. During the training process, it is necessary to pay attention to cross-validation of time series to avoid data leakage. For example, divide the training set and test set in chronological order instead of randomly. It is also necessary to consider early stopping and learning rate adjustment to prevent overfitting. When evaluating the model, in addition to common metrics such as MAE and RMSE, business-related metrics need to be considered, such as the percentage of prediction error within a few minutes, which is more important for the practical application of the bus system. Model deployment may need to consider the acquisition of real-time data and the model update mechanism, such as retraining the model regularly with new data. Potential problems include the real-time acquisition delay of external data and how to handle missing real-time data, such as a sudden interruption of traffic flow. It may be necessary to design a fault tolerance mechanism, such as filling with the latest data or using predicted values as input. Additionally, whether the models for different bus lines or stations need to be trained separately or a global model with shared parameters is also something to consider. In summary, the entire process includes data collection and preprocessing, feature engineering, model design, training and validation, deployment and optimization. Each step needs to be carefully handled, especially when dealing with multi-source heterogeneous data and the structural design of the model to ensure the accuracy and real-time nature of the prediction.

[0033] As an optional implementation manner of this embodiment, after the server continuously obtains the new arrival times of buses of different trips and the information of different factors affecting the new arrival times, an incremental learning method is used to optimize the prediction model.

[0034] In the above optional implementation manner, bus stops usually need to display real-time arrival times. However, due to various factors such as traffic congestion and vehicle scheduling delays, real-time arrival time prediction faces challenges. The traditional fixed departure schedule cannot accurately reflect the real situation of bus arrivals, resulting in lagging passenger information and affecting the user experience.

[0035] In this embodiment, each bus is equipped with a GPS device to upload the vehicle position to the server in real time. By obtaining data such as the position, speed, and route of the bus in real time, the estimated arrival time can be calculated dynamically. The mobile communication network (such as 4G / 5G) is used to transmit data in real time. In this way, the accuracy of the arrival time is improved, and it can dynamically respond to changes in traffic conditions.

[0036] The solution of this system: 1) Combine traffic flow and road condition data: By accessing the real-time traffic flow and road condition data provided by the intelligent transportation system (ITS) and combining machine learning algorithms such as neural networks, the arrival time can be predicted dynamically. Considering factors such as traffic signals, congestion, and accidents, the model can automatically adjust the predicted value. The specific steps are as follows:

[0037] Step 1: Data collection and preprocessing: When collecting data, real-time data is obtained from the specified database in the Intelligent Transportation System (ITS) database, including: bus location data (GPS positioning data), real-time traffic flow (traffic volume on the road, traffic signal status, etc.), road conditions (such as traffic accidents, construction sections, weather impacts, etc.), historical data such as the departure time of the bus and the planned arrival time. Other relevant data is collected: historical bus arrival times (historical data), external factors such as holidays and weather (which can affect passenger flow and traffic conditions).

[0038] During data preprocessing, data cleaning can be performed, including handling missing values and outliers to ensure the integrity and accuracy of the data. Feature engineering, according to the target task, extracts relevant features from the original data, such as: traffic flow on the road section, vehicle speed, traffic signal status, etc.; compares time and date (peak hours, holidays, etc.) with historical data (such as previous arrival times). And data standardization / normalization is performed on the cleaned data, including normalizing the feature data to ensure that the scales of different features are consistent and to avoid the excessive influence of certain features during model training.

[0039] Select a machine learning algorithm. According to the data characteristics and the complexity of the problem, select an appropriate machine learning algorithm. Common algorithms include: Regression Analysis, used to predict continuous variables (such as arrival time), and algorithms such as linear regression and ridge regression are adopted, considering variables such as traffic flow, road conditions, and traffic signals to establish linear or non-linear relationships.

[0040] Neural Networks can also be used: Deep learning (such as multi-layer perceptron, LSTM (Long Short-Term Memory Network)) is used to capture complex patterns in time-series data. Neural networks are particularly effective in dealing with complex non-linear relationships (such as congestion, accidents, etc.) and dynamically changing features. In this application, a neural network model can be used to predict the arrival time.

[0041] During model training, the dataset is pre-divided into a training set, a validation set, and a test set. Usually, 70% of the data is used for training, 15% for validation, and 15% for testing. During model training, the training set is used to train the model. During the training process, the model learns the relationship between the arrival time and these features by inputting features (such as real-time traffic flow, road condition information, etc.). For a regression model, the loss function (such as Mean Squared Error (MSE)) is mainly optimized during training to reduce the prediction error.

[0042] For a neural network model, the backpropagation algorithm is used to update the weights in the network and optimize the prediction effect. During the training process, the model continuously adjusts the weights based on historical data and real-time data to learn how to handle different traffic conditions, road signals, etc.

[0043] When validating and adjusting the model, it includes: validating the model performance and evaluating the accuracy of the model using a validation set. For example, the mean squared error (MSE) or mean absolute error (MAE) is used to measure the error between the predicted result and the actual arrival time. The generalization ability of the model is further improved through cross-validation methods (such as K-fold cross-validation) to avoid overfitting.

[0044] Model adjustment: Adjust the model by selecting more appropriate features, optimizing algorithms, adjusting hyperparameters, etc. to improve the performance of the model. For a neural network, the accuracy of the model can be improved by adjusting hyperparameters such as the learning rate, the number of hidden layers, and the number of nodes.

[0045] Based on the trained model, real-time prediction and dynamic adjustment are carried out, including real-time data input. When the model training is completed, real-time traffic data is accessed, including information such as vehicle positions, traffic flow, and traffic signals. The model uses the real-time data to dynamically calculate the predicted arrival time.

[0046] Dynamic adjustment of the prediction result: If the road conditions change (such as a traffic accident, road congestion, etc.), the model will adjust the predicted arrival time value according to the real-time data. In a neural network, time series models such as LSTM can be dynamically adjusted according to the historical data of the time series to ensure that the model can adapt to changes in different time periods (such as the difference between peak hours and off-peak hours). By continuously receiving new real-time data, the system can optimize the model through a feedback mechanism. For example, the model is adjusted based on the error between the actual arrival time and the predicted arrival time each time. When implementing, real-time data can be obtained and window aggregation is performed on the real-time data stream. If there is an emergency in the real-time data, the timeliness weight is set to 3 times that of the regular feature to force the model to respond quickly. If there are peak hours and off-peak hours in the real-time data, then the model switches between different models for different time periods. During peak hours, a congestion propagation layer is added, and a specified number of hidden units, such as 256, are set in this layer; during off-peak hours, a lightweight architecture is adopted, such as 64 hidden units, to improve the inference speed. Further, after receiving a sample composed of N pieces of real-time data, such as about 15 minutes of data, online fine-tuning can be triggered, such as adjusting the learning rate and reducing it to 10% of the initial value. At the same time, elastic weight consolidation can be used to prevent catastrophic forgetting.

[0047] Feedback optimization and error compensation mechanism. The model includes a multi-level feedback loop design and short-term (second-level) compensation. When the deviation between the actual arrival time and the predicted value is greater than k, for example, 5 minutes, the compensation coefficient is immediately injected: Δtnew = Δtpred + 0.7(Δtreal - Δtpred). The MAE is statistically calculated by time period, and the weights of the loss function are automatically adjusted. For medium-term parameter tuning (hour-level), the MAE is statistically calculated by time period, and the weights of the loss function are automatically adjusted. For example, the error weight during peak hours is increased, peak_weight = 3.0 if 7 <= hour < 9 else 1.0.

[0048] Long-term reconstruction (month-level): When the cumulative error exceeds the threshold (such as monthly average MAE > 4 minutes), trigger model structure reconstruction (such as adding a GRU branch).

[0049] Uncertainty quantification, using Monte Carlo Dropout (retaining the Dropout layer during inference), and outputting an 80% confidence interval. For abnormal error tracing, analyze through SHAP values to locate high-contribution features (such as a 30% sudden increase in the speed weight of a certain section of the road).

[0050] If certain factors (such as traffic accidents) lead to excessive prediction errors, the system can update the model through online learning or incremental learning to enable it to adapt to new changes.

[0051] Display the prediction result, and display the predicted arrival time of the bus in real time on the human-computer interaction display screen at the bus stop. Users can adjust their travel plans according to the predicted arrival time.

[0052] Dynamic update: Before the actual bus arrives at the station, the system will continuously update the predicted arrival time to ensure the accuracy and real-time nature of the information. When implementing, continuously input the real-time changing input data into the prediction model and obtain the current latest prediction result in real time. This method can provide more accurate predictions under complex traffic conditions, especially during the urban peak period.

[0053] 2) Prediction model based on historical data: By collecting historical bus data (such as the historical departure time, route, climate, road conditions, etc. of vehicles) and traffic flow, use big data analysis and time series analysis methods to establish a prediction model to estimate the arrival time of the bus in advance. It can be divided into the following key steps:

[0054] a. Data collection and integration: Collect historical bus data. First, historical data related to bus arrival times needs to be collected. The main data sources include: Historical departure times: Record the departure time and actual arrival time of each bus. Route information: The specific route and stop information of the bus line, including the distances between stations and the dwell times. Weather data: Factors such as temperature, precipitation, and wind speed can affect traffic flow and the driving speed of buses, so weather data is an important factor in prediction. Traffic flow data: Historical traffic flow data at different time periods, the congestion level of road sections, etc. Road conditions: For example, whether there is road construction, traffic accidents, special traffic control, etc. Real-time data collection: To enhance the accuracy of the prediction model, real-time data needs to be accessed, including: Real-time traffic flow data: Information such as the current location of the bus, road conditions, and traffic density. Weather data: Real-time weather conditions. Road condition data: Such as real-time traffic accidents, road closure information, etc. All these data will form the input of the prediction model.

[0055] Data integration and storage: Integrate historical data and real-time data and store them in a data warehouse or database. Ensure that all data sources can be stored in a unified format and time-aligned.

[0056] b. Data preprocessing and feature engineering: Data cleaning and processing. Data cleaning is a crucial step before building a model. Common processing methods include: Handling missing values: For example, if traffic flow or weather data is missing for certain periods, methods such as mean filling and interpolation can be used for supplementation. Handling outliers: Such as some extreme traffic flow values or abnormal GPS positioning points, which need to be detected and processed for outliers. Timestamp alignment: Ensure that the timestamps of various data (such as traffic flow, weather data, etc.) are consistent for convenient subsequent analysis. Feature selection and extraction: Extract key features from historical data, and these features will become the input of the time series analysis model. Common features include: Time features: Include date, time (hour, minute), day of the week, holidays, etc., because the arrival times of buses may vary significantly in different time periods. Traffic flow features: Include traffic flow at certain periods, road congestion index, etc. Weather features: Include temperature, precipitation, wind speed, humidity, etc. Historical arrival times: The historical arrival times in the same period are used as a reference. Data standardization and normalization: To ensure the consistency of dimensions between different features, the data needs to be standardized or normalized. Common methods include: Z-Score standardization: Make the mean of the feature 0 and the variance 1. Min-Max normalization: Scale the data values to the range of [0,1].

[0057] c. Establish a time series model: Select an appropriate time series analysis method. The core purpose of time series analysis is to model and predict the trend of data changing over time. The following methods can be selected: Autoregressive model (AR): Predicts future data based on historical data points. Suitable for stationary time series data. Moving average model (MA): Predicts future values based on past errors. ARMA model (Autoregressive Moving Average model): Combines AR and MA models, suitable for stationary time series data. ARIMA model (Autoregressive Integrated Moving Average model): Deals with non-stationary time series data, suitable for time series data that requires differencing. Seasonal ARIMA (SARIMA): When considering seasonal and periodic changes, the SARIMA model can effectively capture such fluctuations. Deep learning models (such as LSTM): For time series data with complex non-linear relationships, neural network methods such as LSTM (Long Short-Term Memory network) can handle long-term dependencies.

[0058] Feature selection and modeling: Select appropriate input features, which usually can include: Time features: Date, hour, day of the week, etc. Historical arrival times: For example, the arrival times at the same time point in the past several times are used as inputs. External factors: Such as external factors like traffic flow and weather.

[0059] Model training and optimization: Use the training data to train the selected model. During the training process, the model will learn the relationship between the input features and the bus arrival time. During training, the following methods can be used to optimize the model: Hyperparameter tuning: Adjust the hyperparameters of the model through methods such as Grid Search to find the best model configuration. Cross-validation: Use methods such as K-fold cross-validation to evaluate the generalization ability of the model and avoid overfitting.

[0060] Model evaluation: Evaluate the prediction performance of the model through indicators such as Mean Squared Error (MSE) and Mean Absolute Error (MAE). If the model performance is not ideal, other methods can be tried for improvement, such as introducing new features and adjusting the model structure.

[0061] d. Real-time prediction and dynamic adjustment: Input real-time data, input the real-time collected traffic flow, climate data, road conditions, etc. into the trained prediction model for real-time prediction. Ensure that the data can be transmitted to the model quickly and accurately for calculation.

[0062] Dynamically adjust the prediction results. The model makes dynamic adjustments based on real-time data. For example: When the road conditions change (such as a traffic accident or road closure), the model can adjust the prediction of the arrival time according to the change in traffic flow. When the weather conditions change (such as a sudden heavy rain), the model can also automatically correct the prediction results.

[0063] Feedback mechanism: The real-time prediction results can be fed back to passengers through a human-machine interaction system (such as bus stop display screens, mobile phone apps, etc.). When there are significant changes in the predicted arrival time, the system can be updated in real time to ensure that passengers obtain the latest information.

[0064] e. Performance monitoring and model update: Monitor the prediction accuracy. By monitoring the error between the prediction result and the actual arrival time in real time, the system can identify situations with large prediction errors and make adjustments. For example, if the error is large during a certain period, the reasons can be traced back and corrected.

[0065] Model update and incremental learning: Over time, the system will collect more data. The incremental learning method can be adopted to continuously update and optimize the model using new data. Incremental learning can avoid training from scratch and improve the model update efficiency.

[0066] Continuous optimization: According to user feedback and error monitoring, retrain and optimize the model regularly, add new features, adjust model parameters, etc., to improve the prediction accuracy and robustness.

[0067] This system can predict the arrival time changes in different periods, especially with high accuracy in the case of regular traffic.

[0068] As an optional implementation manner of this embodiment, if the query target information is the bus stop passenger flow query information, the server calls the detection device set in the bus stop or on the interaction device to detect the number of passengers at the bus stop; obtains the weather data, the current corresponding event information, the current corresponding seasonal data, and the current traffic flow data; inputs the passenger data, the weather data, the current corresponding event information, the current corresponding seasonal data, and the current traffic flow data into the traffic flow prediction model to predict the passenger flow at the bus stop in a future period of time; and sends the predicted result to the interaction device.

[0069] As an optional implementation manner of this embodiment, when training the traffic flow prediction model, the method includes: obtaining the actual passenger flow data at different time points, the time feature data corresponding to each different time point, the weather data affecting the passenger flow corresponding to each different time point, the event information corresponding to each different time point, the seasonal change data corresponding to each different time point, and the historical traffic flow data corresponding to each different time point, so as to obtain the passenger flow sample data; preprocessing the passenger flow sample data and then inputting it into the machine learning model, and using the actual passenger flow within a specified period after each different time point as the output to train the machine learning model.

[0070] In the above optional implementation, the passenger flow at a bus stop is usually affected by various factors (such as time period, weather, holidays, etc.). Traditional manual statistics and empirical predictions cannot cope with the changing passenger flow situation, resulting in problems such as unreasonable train scheduling, overcrowding or empty buses for the bus company.

[0071] Solutions: 1) Real-time passenger flow monitoring based on sensors: Use infrared sensors, pressure sensors or video surveillance (through computer vision analysis) to monitor the passenger flow at the platform in real time. Based on the change in the number of passengers on the platform, the flow trend of passengers can be dynamically predicted. This solution can provide accurate real-time passenger flow data to ensure that the system can reflect the immediate needs of passengers.

[0072] 2) Prediction based on historical data and machine learning: Technical solution: By analyzing historical passenger flow data and seasonal changes, combined with machine learning algorithms (such as support vector machines, decision trees or deep learning), predict the future passenger flow. This method takes into account various influencing factors such as the flow pattern of passengers, weather changes, holidays, etc. It can be described in the following steps:

[0073] a. Data collection and integration: Collect historical passenger flow data, collect past passenger flow data, and the data should include the following content: Historical passenger flow: The actual passenger flow data for a certain period of time (such as per hour, per day, etc.). Time characteristics: Include date, time (hour, day of the week, holidays, etc.). For example, the passenger flow on holidays or weekdays is often different. Weather data: Weather data such as temperature, precipitation, wind speed, humidity, etc. that affect the bus passenger flow. Special event data: For example, the impact of major events, emergencies (such as accidents, road closures, etc.) on the passenger flow. Traffic flow data: Surrounding traffic conditions, congestion situations, traffic accidents and other factors will also affect the passenger flow.

[0074] Collect seasonal change data. Seasonal changes have a greater impact on the passenger flow. Therefore, the following data needs to be collected: Seasonal data: The passenger flow change rules in different seasons (spring, summer, autumn, winter), especially during the tourist season and holidays, the passenger flow usually fluctuates significantly.

[0075] Real-time data collection, obtain real-time data, including: Current weather conditions: Real-time weather data such as temperature, humidity, precipitation, etc. Real-time traffic conditions: Real-time information such as traffic flow, road congestion, etc. that affect the passenger flow.

[0076] b. Data preprocessing and feature engineering: Data cleaning and missing value processing, clean the original data, and common processing methods include: Handling missing values: If the passenger flow data is missing, methods such as interpolation or mean filling can be used to fill in the missing values. Outlier handling: Eliminate extremely abnormal passenger flow data to ensure the effectiveness of model training.

[0077] Feature selection and extraction: By analyzing historical data, features helpful for passenger flow prediction are extracted. Common features include: Time features: including year, month, day, hour, day of the week, holidays, etc. (One-Hot encoding can be used). Seasonal features: such as seasons, holidays, etc. Weather features: such as temperature, humidity, precipitation, etc. Historical passenger flow data: such as the passenger flow in the past few days or hours (can be used as lag features). Traffic condition features: such as traffic flow, traffic accidents, etc.

[0078] Data standardization and normalization: To eliminate the dimensional differences between different features, the following methods are used for data standardization: Standardization (Z-score): Convert feature data to zero mean and unit variance. Normalization: Scale the data to the range of [0, 1], which is especially suitable for the input of deep learning models.

[0079] c. Establish a prediction model: Select a machine learning model. According to the type of data features and the requirements of the model, select a suitable machine learning algorithm. Common algorithms include: Support Vector Machine (SVM): Suitable for linear and non-linear data, can effectively handle high-dimensional data, and performs well in small sample data. Steps: Use historical passenger flow and other features as input to train an SVM regression model for passenger flow prediction. Decision Tree: A decision tree can divide data into different intervals through a series of "judgment conditions" and has strong interpretability. Through the decision tree model, build a relationship tree of the impact of different features (such as time, weather, etc.) on passenger flow, so as to predict future passenger flow.

[0080] Deep learning (such as LSTM or RNN): Suitable for processing time series data, can learn long-term dependencies in time, and is especially suitable for processing complex non-linear data. Steps: Use a Long Short-Term Memory (LSTM) network for time series prediction. By inputting historical passenger flow and related features, predict future passenger flow.

[0081] Model training: After selecting the algorithm, the model needs to be trained. The main steps include: Divide the training set and the test set: Divide the data into a training set (80%) and a test set (20%) according to a certain proportion for training and verification. Train the model: Use the training data to train the model and adjust the hyperparameters of the model. Evaluate the model performance: Evaluate its prediction ability through the performance of the model on the test set. Common evaluation indicators include: Mean Squared Error (MSE), Mean Absolute Error (MAE), R2 value (coefficient of determination). Model optimization: Hyperparameter tuning: Use methods such as cross-validation and grid search to optimize the hyperparameters of the model. Model fusion: You can try to fuse multiple models (such as random forest, XGBoost, etc.) to improve the prediction effect.

[0082] d. Prediction and Application: Predict future passenger flow. After the model is trained, it can be used to predict future passenger flow. When predicting, input information such as weather data, time features, and traffic flow at the current moment, and the model will output the passenger flow within a specified future time period.

[0083] Dynamic Adjustment and Real-time Prediction: Dynamic adjustment dynamically adjusts the model based on real-time data. For example, when sudden changes occur in weather or traffic conditions, the model can quickly adjust the prediction results according to the new input. Real-time prediction: For real-time prediction at a specific moment, real-time data streams can be accessed for dynamic prediction to update arrival information or passenger flow information. Feedback and Visualization: Feed the predicted passenger flow back to relevant platforms (such as bus stop displays, APPs, etc.) through the system for passengers to refer to. In addition, the trend changes of passenger flow prediction can be displayed through a visualization interface to facilitate better operation management.

[0084] e. Performance Monitoring and Model Update: Monitor prediction results, monitor the performance of the prediction model in real time, and compare the error between the actual passenger flow and the predicted value. If the error is large, the model or data processing method can be adjusted in a timely manner. Model retraining: Over time, as more historical data accumulates, the model may lose some accuracy, so the model needs to be updated and retrained regularly to adapt to new data. Incremental learning: To ensure the continuous improvement of the model, incremental learning can be adopted to gradually update the model when new data arrives, rather than completely retraining from scratch. Advantages: Machine learning methods can automatically identify potential passenger flow change trends, thereby improving the accuracy of prediction and avoiding overcrowding or empty buses at the platform.

[0085] 3) Integration of Social Media and External Data: By analyzing social media (such as Weibo, WeChat, etc.) and weather data, predict passenger flow changes at specific times and special events (such as concerts, festivals, etc.). Social media data can be used as real-time feedback to help optimize the passenger flow prediction model. External data can effectively supplement the input of the model and improve the reliability of prediction, especially in some irregular special situations.

[0086] As an alternative implementation of this embodiment, the preset data transmission mechanism includes: compressing data using a data compression algorithm; based on the sharding transmission technology, splitting large data packets into small pieces for transmission, where a low-latency data transmission protocol is used during the transmission process; using multiple network transmission methods for transmission; the multiple network transmission methods include automatically selecting a network transmission method according to the real-time network condition. The human-computer interaction system at the bus stop needs to efficiently transmit a large amount of data, including real-time vehicle positions, arrival times, passenger flows, traffic conditions, and other information. Problems such as data transmission latency, network instability, and data loss seriously affect the real-time performance and reliability of the system, especially in areas with incomplete network coverage or strong signal interference.

[0087] Solution: 1) Low-latency data transmission protocol:

[0088] Technical solution: Adopt low-latency data transmission protocols such as MQTT or WebSocket to reduce the time delay of data transmission and ensure real-time update of information. MQTT is a lightweight protocol based on the publish / subscribe model and is particularly suitable for low-bandwidth and high-latency network environments. Advantage: It can effectively reduce transmission latency, improve the response speed and stability of the system, and adapt to complex network environments.

[0089] 2) Hybrid network transmission and redundant backup: Technical solution: To cope with network fluctuations and instability, adopt redundant backup of multiple network transmission methods such as 4G / 5G and Wi-Fi. The system can automatically select the best transmission path according to the real-time network condition to ensure stable data transmission. Advantage: Through redundant backup, it is ensured that even if a certain network fails, other networks can continue to guarantee the stability of data transmission.

[0090] 3) Edge computing and data caching: Technical solution: Transfer some data processing tasks from the central server to edge computing devices close to the platform. This can not only reduce the pressure of data transmission but also analyze and predict the received data in real time. In addition, the platform can use local caching to temporarily store data that fails to be updated in a timely manner and synchronize it later when the network is restored. Advantage: Edge computing reduces the dependence on the central server and improves the efficiency and real-time performance of data processing. The caching mechanism can ensure the continuous display of data and maintain basic functions even when the network is unstable.

[0091] 4) Data Compression and Optimized Transmission: Technical Solution: During data transmission, data compression algorithms (such as gzip or Brotli) are used to compress the data to reduce the size of the transmitted data, thereby improving the transmission efficiency. In addition, the fragment transmission technology is adopted to split large data packets into small pieces for transmission, which helps to improve stability. Advantage: Through compression and fragment transmission, the bandwidth requirement is reduced and the data transmission efficiency is improved, especially in the case of limited network bandwidth.

[0092] The combination of these technologies not only improves the accuracy and real-time performance of the bus stop display system, but also enhances the user experience and the efficiency of bus services.

[0093] Reference Figure 2 Schematically shows the functional system diagram of the system, reference Figure 3 Schematically shows the operation flow chart of the present system when in use.

[0094] The bus stop human-computer interaction display method not only has many advantages, but also significantly improves the travel experience of passengers and the overall efficiency of the bus system. The following are the main effects and advantages of this system:

[0095] The present embodiment achieves the following beneficial effects:

[0096] Improve passenger satisfaction: By providing accurate and real-time bus information, the uncertainty and waiting time of passengers are reduced, and the passenger satisfaction is improved.

[0097] Optimize bus scheduling: Real-time monitoring and feedback on the running status of buses help bus companies optimize scheduling and management, and improve operation efficiency.

[0098] Increase advertising revenue: Use the display screen for advertising placement to open up new sources of income and relieve the operation pressure of the bus system.

[0099] Enhance safety guarantee: The integrated monitoring system and emergency alarm function improve the safety of bus stops and ensure the travel safety of passengers.

[0100] Promote the construction of smart cities: Through the linkage with other intelligent transportation systems, information sharing and resource integration are realized, and the development of smart cities is promoted.

[0101] Real-time information update: The system can display the arrival time of buses, route information, transfer suggestions, etc. in real time to help passengers make better travel decisions.

[0102] Convenient query: Passengers can conveniently query the required bus information through touch screens, voice recognition, etc., and the operation is simple and intuitive.

[0103] Multi - language support: Supports interfaces in multiple languages, facilitating the use by passengers from different language backgrounds.

[0104] Advertisement display: Displays advertisement content through the display screen, increasing revenue sources while providing commercial information to passengers.

[0105] Interactive functions: Provides functions such as route planning, weather forecast, and query of surrounding facilities, enriching the travel experience of passengers.

[0106] Safety monitoring: The system integrates surveillance cameras and emergency alarm devices to enhance the safety of bus stops.

[0107] Information sharing: Can be networked with other urban transportation systems to achieve information sharing and resource integration, enhancing the intelligence level of the entire urban transportation system.

[0108] Energy conservation and environmental protection: Adopts highly energy - efficient display technologies and intelligent control systems to reduce energy consumption and meet environmental protection requirements.

[0109] In summary, the human - machine interaction display method for bus stops greatly improves the efficiency of the bus system and the travel experience of passengers through intelligent management and services, providing strong support for the construction of smart cities. Its diverse functions and remarkable benefits make it an indispensable part of modern public transportation systems.

[0110] The above are only examples of this application and are not used to limit this application. For those skilled in the art, this application can have various changes and modifications. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of this application shall be included within the scope of the claims of this application.

Claims

1. A bus stop human-computer interaction system, characterized in that: include: An interactive device provided at a bus stop, the interactive device comprising a touch module, a controller, a sensor, a display module, and a data transmission interface, the touch module being used to detect a user's touch screen operation; The controller generates designated interaction information based on the interaction operation; the sensor is used to detect the change parameters of the environment around the interaction device, and send the change parameters to the controller, so that the controller can adjust the display module, wherein the touch screen operation includes manual query operation and voice query operation; the interaction device obtains the real-time bus data from the database through the data transmission interface and displays it through the display module; The server obtains interaction information from the interaction device; parses the interaction information to obtain query purpose information; determines corresponding feedback information based on the query purpose information, and feeds the feedback information back to the interaction device.

2. The bus stop human-computer interaction system according to claim 1, characterized in that: The system also includes a positioning device arranged on the bus, which is used to locate the bus in real time, wherein the real-time positioning information is uploaded to a designated database; If the query purpose information is train query information, the real-time traffic data on the route of the train to be queried is obtained through a preset data transmission mechanism; the real-time traffic data is input into the prediction model for prediction to obtain the current arrival time of the train to be queried, wherein the real-time traffic data includes the real-time positioning information of the bus detected by the positioning device on the bus of the train to be queried, the departure time of the bus of the train to be queried obtained from the designated database, the real-time traffic flow and road condition data obtained from the designated database, and data of designated dimensions affecting traffic.

3. The bus stop human-computer interaction system according to claim 1, characterized in that: The interactive device also includes an image acquisition module, which is used to capture images of the bus stop through a camera device, and transmit the image acquisition device to a server through a preset data transmission mechanism, so that the server can perform abnormality judgment based on the captured image; and / or, the image acquisition module captures images inside the bus stop, and calculates the real-time number of people in the station through the controller.

4. The bus stop human-computer interaction system according to claim 1, characterized in that: The interactive device also includes a button module. When the button module is triggered, the controller generates alarm information; and uploads the alarm information to the alarm system.

5. The bus stop human-computer interaction system according to claim 2, characterized in that: When training the prediction model, the method includes: obtaining historical traffic data of all buses under the bus number on the route to be queried; processing the historical traffic data to obtain time series data as sample data; inputting the time series data into the prediction model to be trained, and taking the arrival time corresponding to different data at each time point as output, and training the prediction model to be trained, wherein the time series data includes the historical arrival time corresponding to the specified time feature point, and the corresponding specified factor information; The loss function of the prediction model is Wherein, k is the given sample data, P is the number of neurons in the output layer of the prediction model; n = 1, 2, 3...P; X nk is the actual arrival time of sample k; t nk is the target expected value; q is a constant, and the value of q is 1 / 2.

6. The bus stop human-computer interaction system according to claim 5, characterized in that: The server optimizes the prediction model by using an incremental learning method based on the continuously acquired information about the new arrival times of buses of different train numbers and the different factors that affect the new arrival times.

7. The bus stop human-computer interaction system according to claim 1, characterized in that: If the query purpose information is bus station passenger flow query information, the server calls a detection device set in the bus station or on the interactive device to detect the number of passengers at the bus station; Obtained weather data, current corresponding event information, current corresponding seasonal data, current traffic flow data; The passenger data, the weather data, the current corresponding event information, the current corresponding seasonal data, and the current traffic flow data are input into a traffic prediction model to predict the passenger flow of the bus station in the future; and the prediction result is sent to the interactive device.

8. The bus stop human-computer interaction system according to claim 7, characterized in that: When training the traffic prediction model, the method includes: Obtain actual passenger flow data at different time points, as well as time characteristic data corresponding to each different time point, weather data affecting passenger flow at each different time point, event information corresponding to each different time point, seasonal change data corresponding to each different time point, and historical traffic flow data corresponding to each different time point, and then obtain passenger flow sample data; The passenger flow sample data is preprocessed and then input into a machine learning model, and the actual passenger flow in a specified time period after different time points is used as output to train the machine learning model.

9. The bus stop human-computer interaction system according to claim 2 or 3, characterized in that: The preset data transmission mechanism includes: A data compression algorithm is used to compress data. Based on the fragmentation transmission technology, large data packets are split into small blocks for transmission, wherein a low-latency data transmission protocol is used during the transmission process. A variety of network transmission methods are used for transmission. The multiple network transmission methods include automatically selecting a network transmission method according to the real-time network status.

10. The bus stop human-computer interaction system according to claim 1, characterized in that: The system also includes an edge computing device arranged near the bus stop to assist the interactive device in interacting with the server.

Images