Narrow road meeting early warning and scheduling control system and method

Abstract

The invention relates to the technical field of vehicle meeting early warning, and particularly discloses a narrow road vehicle meeting early warning and scheduling control system and method, and the system collects the position, speed and advancing direction information of a vehicle through a vehicle-mounted sensor, and transmits the vehicle state information to a rear-end platform through a vehicle-mounted communication unit. And an artificial intelligence technology based on deep learning is introduced into a rear-end platform to carry out time sequence analysis on the vehicle state information of each vehicle, and combined interactive reasoning is carried out on the driving state characteristics of different vehicles, so that intelligent prediction and scheduling control on the meeting state of the vehicles are realized. In this way, when the vehicle meeting condition of the narrow road section is recognized, the deceleration early warning prompt can be sent out in time, so that the risk of traffic accidents is effectively reduced, and the road passing efficiency is improved.

Description

Technical Field

[0001] The present application relates to the technical field of oncoming vehicle warning, and more specifically, to a narrow road oncoming vehicle warning and dispatching control system and method. Background Art

[0002] With the acceleration of urbanization and the development of social economy, road traffic volume continues to grow, and the problem of passing vehicles on the road has always been an important factor affecting traffic safety and traffic efficiency. Especially in scenes such as winding mountain roads, old urban streets, and rural roads, due to the limited width of the road, there are often problems such as limited vision and insufficient avoidance space, which makes it easy for vehicles to pass each other when driving in both directions on such roads, and may cause traffic congestion and traffic accidents, facing a high safety risk.

[0003] Although traditional traffic management measures such as adding traffic signs and road markings can alleviate such problems to a certain extent, this method mainly relies on the driver's visual judgment and reaction speed, and it is usually difficult to achieve precise time and space control, especially in complex and changeable road environments.

[0004] Therefore, an optimized narrow road vehicle meeting warning and dispatching control system and method are expected. Summary of the invention

[0005] In order to solve the above technical problems, the present application is proposed. The embodiment of the present application provides a narrow road meeting warning and dispatch control system and method, which collects the position, speed and travel direction information of the vehicle through the vehicle-mounted sensor, uses the vehicle-mounted communication unit to send the vehicle status information to the back-end platform, and introduces artificial intelligence technology based on deep learning on the back-end platform to perform time series analysis on the vehicle status information of each vehicle, and performs joint interactive reasoning on the driving status characteristics of different vehicles to achieve intelligent prediction and dispatch control of the meeting status between vehicles. In this way, when it is identified that a narrow road section is about to meet, a deceleration warning prompt can be issued in time, thereby effectively reducing the risk of traffic accidents and improving road traffic efficiency.

[0006] Accordingly, according to one aspect of the present application, a method for early warning and dispatching control of vehicles on narrow roads is provided, comprising:

[0007] Collecting vehicle status information through vehicle-mounted sensors, wherein the vehicle status information includes vehicle position, vehicle speed and travel direction;

[0008] Using the vehicle communication unit to send the vehicle status information to the back-end platform;

[0009] On the back-end platform, oncoming vehicle warning analysis is performed based on the time data set of vehicle status information of each vehicle;

[0010] On the back-end platform, scheduling control is performed based on a time data set of vehicle status information of each vehicle;

[0011] Among them, the oncoming vehicle warning analysis is performed based on the time data set of the vehicle status information of each vehicle, including: extracting the time data set of the vehicle status information of the first vehicle and the time data set of the vehicle status information of the second vehicle; performing vehicle status feature extraction on the time data set of the vehicle status information of the first vehicle and the time data set of the vehicle status information of the second vehicle respectively to obtain a first vehicle status inference representation vector and a second vehicle status inference representation vector; performing interactive fusion based on common feature guidance on the first vehicle status inference representation vector and the second vehicle status inference representation vector to obtain a first vehicle-second vehicle meeting state joint reasoning representation vector; based on the first vehicle-second vehicle meeting state joint reasoning representation vector, determining whether a narrow road meeting occurs.

[0012] According to another aspect of the present application, a narrow road vehicle meeting warning and dispatching control system is provided, which includes:

[0013] A vehicle status information collection module is used to collect vehicle status information through vehicle-mounted sensors, wherein the vehicle status information includes vehicle position, vehicle speed and travel direction;

[0014] An information transmission module, used to send the vehicle status information to a back-end platform using an on-board communication unit;

[0015] A vehicle-on-vehicle warning analysis module, used for performing vehicle-on-vehicle warning analysis based on a time data set of vehicle status information of each vehicle on the back-end platform;

[0016] The dispatch control module is used to perform dispatch control on the back-end platform based on the time data set of vehicle status information of each vehicle.

[0017] Compared with the prior art, the narrow road meeting warning and dispatch control system and method provided by the present application collects the position, speed and direction of travel information of the vehicle through the vehicle-mounted sensor, sends the vehicle status information to the back-end platform using the vehicle-mounted communication unit, and introduces artificial intelligence technology based on deep learning on the back-end platform to perform time series analysis on the vehicle status information of each vehicle, and realizes intelligent prediction and dispatch control of the meeting status between vehicles by joint interactive reasoning of the driving status characteristics of different vehicles. In this way, when it is identified that a narrow road section is about to meet, a deceleration warning prompt can be issued in time, thereby effectively reducing the risk of traffic accidents and improving road traffic efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0019] Figure 1 Flow chart of a narrow road vehicle meeting warning and dispatch control method according to an embodiment of the present application.

[0020] Figure 2 Flow chart of step S130 in the narrow road vehicle meeting warning and dispatch control method according to an embodiment of the present application.

[0021] Figure 3 Schematic diagram of data flow in step S130 of the narrow road vehicle meeting warning and dispatch control method according to an embodiment of the present application.

[0022] Figure 4 This is a flowchart of step S132 in the narrow road meeting warning and dispatch control method according to an embodiment of the present application.

[0023] Figure 5 Flow chart of step S133 in the narrow road meeting warning and dispatch control method according to an embodiment of the present application.

[0024] Figure 6 4 is a block diagram of a narrow road vehicle crossing warning and dispatch control system according to an embodiment of the present application. DETAILED DESCRIPTION

[0025] Below, the embodiments of the present application will be described in more detail in conjunction with the accompanying drawings, and the above and other purposes, features and advantages of the present application will become more apparent. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the example embodiments described herein.

[0026] Figure 1 FIG. 1 is a flow chart of a method for early warning and dispatching control of vehicles on narrow roads according to an embodiment of the present application. Figure 1As shown, according to the narrow road vehicle passing warning and dispatch control method of the embodiment of the present application, the steps include: S110, collecting vehicle status information through vehicle-mounted sensors, the vehicle status information including vehicle position, vehicle speed and travel direction; S120, using the vehicle-mounted communication unit to send the vehicle status information to the back-end platform; S130, on the back-end platform, performing vehicle passing warning analysis based on the time data set of vehicle status information of each vehicle; S140, on the back-end platform, performing dispatch control based on the time data set of vehicle status information of each vehicle.

[0027] In the above-mentioned narrow road meeting warning and dispatch control method, the step S110 collects vehicle status information through vehicle-mounted sensors, and the vehicle status information includes vehicle position, speed and direction of travel. It should be understood that vehicle position, speed and direction of travel are basic parameters that describe the driving behavior of vehicles on the road. By collecting the position, speed and direction of travel of each vehicle, the real-time status of the vehicle can be accurately grasped, so as to predict in advance whether the two vehicles will meet, and then issue a warning in advance or take measures to avoid potential conflicts.

[0028] In the process of collecting the above vehicle status information, multiple links are involved, including the selection and installation of multiple sensors, data collection and processing, data transmission and storage, and data verification and exception handling. First of all, the vehicle-mounted sensor is a key device for collecting vehicle status information. In the narrow road crossing warning and dispatch control system, commonly used vehicle-mounted sensors include GPS sensors, speed sensors, gyroscopes, and accelerometers. The selection and installation of these sensors need to comprehensively consider the type of vehicle, the use environment, and the expected accuracy requirements.

[0029] Specifically, the GPS sensor is used to obtain the precise location information of the vehicle, including longitude, latitude and altitude. The high-precision GPS module is able to receive multiple satellite signals to improve positioning accuracy. When installing, the GPS antenna should be placed on the top of the vehicle or other unobstructed locations to ensure the stability of signal reception. In addition, it is also necessary to ensure that the GPS module is properly connected to the vehicle's power and communication interfaces to transmit location data in real time.

[0030] Speed ​​sensors are usually installed on the axle or transmission of the vehicle to measure the speed of the vehicle. Common speed sensors include Hall effect sensors, magnetic induction sensors, and photoelectric sensors. These sensors calculate the vehicle speed by detecting the rotation of the axle or the speed of the transmission. When installing, ensure that the sensor is in good contact with the axle or transmission to avoid data errors caused by looseness or wear. At the same time, the sensor should be connected to the vehicle's electronic control unit (ECU) to transmit speed data in real time.

[0031] Gyroscopes and accelerometers are used to detect the attitude and motion state of the vehicle. A three-axis gyroscope can measure the angular velocity of the vehicle, while a three-axis accelerometer can measure the acceleration of the vehicle. These sensors are usually integrated into a module and installed in the center of the vehicle to reduce the impact of external interference. When installing, ensure that the sensor is firmly fixed to avoid data drift due to vibration. The sensor module should be connected to the vehicle's power and communication interfaces to transmit attitude and acceleration data in real time.

[0032] Although cameras are not used directly to collect location, speed, and direction of travel information, they can provide additional environmental perception data in some cases. Cameras can be installed in front or behind the vehicle to detect road conditions and the location of other vehicles. In addition, camera data can be used to extract useful information through image processing algorithms to assist in oncoming vehicle warning and dispatch control.

[0033] Accordingly, the raw data collected by the sensor needs to be preprocessed and converted before it can be used for subsequent analysis and control.

[0034] GPS modules usually output data in the NMEA 0183 standard format, including time, longitude, latitude, speed and other information. The data collection frequency is usually once per second or more frequently, depending on the needs of the application scenario. During the data collection process, the received data needs to be parsed to extract the required longitude, latitude and altitude information. In addition, the vehicle's heading angle, that is, the direction the vehicle is currently traveling, can also be calculated.

[0035] The speed sensor outputs a pulse signal, and the vehicle speed can be determined by counting the number of pulses per unit time. The data acquisition frequency is similar to that of a GPS sensor, usually once per second or more. During the data acquisition process, the received pulse signals need to be counted and the vehicle speed calculated based on the vehicle's wheel diameter and transmission ratio. In addition, the noise can be removed by filtering algorithms to improve the accuracy of the vehicle speed data.

[0036] Gyroscopes and accelerometers are used to collect information about the direction of travel. The gyroscope outputs angular velocity data, and the accelerometer outputs acceleration data, usually in the form of digital signals. The data collection frequency is usually dozens of times per second or higher, depending on the performance of the sensor and the requirements of the application scenario. During the data collection process, the received angular velocity and acceleration data need to be filtered and fused. A common method is the Kalman filter algorithm, which calculates the direction and attitude of the vehicle by fusing the data from the gyroscope and accelerometer.

[0037] In addition, the collected raw data may contain noise and outliers, which require preprocessing. Preprocessing methods include filtering, denoising, and data correction. The filtering algorithm can remove high-frequency noise and improve the smoothness of the data. The denoising algorithm can detect and remove outliers to avoid data distortion. The data correction algorithm can correct the deviation of the sensor and improve the accuracy of the data. The preprocessed data can be better used for subsequent analysis and control, providing reliable data support for oncoming vehicle warning and dispatch control.

[0038] In the above-mentioned narrow road meeting warning and dispatch control method, the step S120 uses the vehicle communication unit to send the vehicle status information to the back-end platform. That is, in order to achieve cross-vehicle information sharing and collaborative work, the vehicle status information is sent to the back-end platform using the vehicle communication unit, so as to use the powerful server resources in the back-end platform to centrally process and deeply analyze the data, and through comprehensive analysis of the driving status of each vehicle, realize intelligent prediction of the meeting status between vehicles.

[0039] It should be understood that the vehicle communication unit is a key component to achieve real-time transmission of vehicle status information. Through the vehicle communication unit, the vehicle can send the collected status information such as location, speed and direction of travel to the back-end platform, thereby realizing remote monitoring and intelligent scheduling. The following is a complete implementation method of using the vehicle communication unit to send vehicle status information to the back-end platform:

[0040] First, select and configure the communication unit. The selection and configuration of the vehicle communication unit is the first step to ensure data transmission efficiency and reliability. Common communication technologies include 4G / 5G modules, Wi-Fi modules, dedicated short-range communication (DSRC) modules, and Bluetooth modules. Each technology has its own characteristics and applicable scenarios, and needs to be selected according to specific application requirements.

[0041] Specifically, the 4G / 5G module is suitable for wide area network communication and can achieve data transmission in a large range. The 4G / 5G module has a high transmission rate and low latency, which is suitable for real-time data transmission. During installation, it is necessary to ensure that the module is correctly connected to the power and communication interface of the vehicle, and the antenna is placed on the top of the vehicle or other unobstructed locations to improve signal quality. The Wi-Fi module is suitable for local area network communication and can achieve high-speed data transmission within a short distance. The Wi-Fi module is suitable for the communication needs of vehicles in a fixed area, such as a parking lot or a park. During installation, it is necessary to ensure that the module is correctly connected to the power and communication interface of the vehicle, and set the network parameters, such as SSID and password. The dedicated short-range communication (DSRC) module is suitable for short-range communication between vehicles and can achieve real-time data exchange between vehicles. The DSRC module is commonly used in vehicle-to-everything (V2X) communication and supports multiple communication scenarios such as vehicle-to-infrastructure (V2I) and vehicle-to-pedestrian (V2P). During installation, it is necessary to ensure that the module is correctly connected to the power and communication interface of the vehicle, and set the communication parameters, such as channel and power. The Bluetooth module is suitable for low-power communication within a short distance, and can realize data exchange between the vehicle and mobile devices (such as smartphones). The Bluetooth module is suitable for short-range data transmission and control command sending. When installing, you need to ensure that the module is correctly connected to the vehicle's power and communication interfaces, and set the pairing information, such as the device name and PIN code.

[0042] In terms of hardware configuration, the vehicle communication unit needs to be connected to the vehicle's power system to ensure that it can provide stable power supply. At the same time, it needs to be connected to the vehicle's sensors and controllers to receive and send data. Common data interfaces include UART, SPI, I2C, etc. For the wireless communication module, an antenna also needs to be connected to ensure signal reception and transmission.

[0043] In terms of software configuration, you need to ensure that the firmware version of the vehicle communication unit is the latest to obtain the best performance and security. Configure the network parameters of the vehicle communication unit, such as IP address, subnet mask, gateway address, etc. For 4G / 5G modules, you also need to configure the APN (Access Point Name). Configure the security settings of the vehicle communication unit, such as encryption algorithm, authentication method, etc., to ensure the security of data transmission. Configure the data format of the vehicle communication unit to ensure that the transmitted data format is consistent with the requirements of the back-end platform. Common data formats include JSON, XML, and binary formats.

[0044] At the same time, choosing the right data transmission protocol is the key to ensuring data transmission efficiency and reliability. Common data transmission protocols include MQTT, HTTP, and TCP / IP. Each protocol has its own characteristics and applicable scenarios, and needs to be selected according to specific application requirements.

[0045] Specifically, the MQTT protocol is suitable for IoT scenarios and has the characteristics of lightweight, low power consumption and high reliability. The MQTT protocol supports the publish / subscribe mode and can realize one-to-many data transmission. Configure the broker address, port number, client ID, user name and password of the MQTT client. Set the QoS (quality of service) level to ensure the reliability of data transmission. For example, QoS 0 means at most once transmission, QoS1 means at least once transmission, and QoS2 means only once transmission. The HTTP protocol is suitable for Web applications and has wide support and good compatibility. The HTTP protocol supports the request / response mode and can realize one-to-one data transmission. Configure the URL, request method (GET / POST), request header and request body of the HTTP client. Set the timeout and retry times to ensure the stability of data transmission. The TCP / IP protocol is suitable for wide area network communication and has stable connection and reliable transmission. The TCP / IP protocol supports long connection and short connection, and you can choose the appropriate connection mode according to the application scenario. Configure the server address, port number, connection timeout and read / write timeout of the TCP client. Set the heartbeat packet to ensure the activity of the connection.

[0046] At the same time, the on-board communication unit needs to package the collected vehicle status information into messages in a specific format for easy transmission and parsing. Common data packaging methods include JSON format, XML format, and binary format. Among them, the JSON format has good readability and extensibility, and is suitable for the development and debugging stages. The XML format has good readability and structured characteristics, and is suitable for complex scenarios. The binary format has high transmission efficiency and is suitable for the formal operation stage. Data in binary format requires clear protocol specifications to ensure that the receiver can parse it correctly.

[0047] After the data is packaged, the on-board communication unit sends the data to the back-end platform through the selected communication technology and transmission protocol. Common data sending methods include scheduled sending, event-triggered sending, and batch sending. Specifically, scheduled sending: set a fixed sending interval, such as sending data once a second. Scheduled sending is suitable for scenarios that require real-time monitoring. For example, a timer can be used to call the sending function every 1 second. Event-triggered sending: Send data when a specific event occurs, such as when a vehicle enters a certain area or the speed exceeds a threshold. Event-triggered sending is suitable for scenarios that need to respond to specific events. For example, data can be sent when the speed exceeds 60km / h. Batch sending: Pack data within a certain time period into a batch and send it to the back-end platform at one time. Batch sending is suitable for scenarios with large data volumes and no real-time transmission is required. For example, data within 1 minute can be packaged into a batch and sent every 1 minute.

[0048] In addition, the backend platform needs to configure the corresponding receiving program to receive the data sent by the vehicle communication unit. Common receiving programs include MQTT Broker, HTTP Server and TCP Server. Specifically, MQTT Broker: Configure MQTTBroker, listen to the specified Topic, and receive messages sent by the MQTT client. For example, Mosquitto can be used as an MQTT Broker. HTTP Server: Configure HTTP Server, listen to the specified port, and receive requests sent by the HTTP client. For example, you can use Node.js and Express framework to build HTTP Server. TCP Server: Configure TCPServer, listen to the specified port, and receive data sent by the TCP client. For example, you can use Node.js to build TCPServer.

[0049] Similarly, the received data needs to be parsed to extract the required vehicle status information. Common data parsing methods include JSON parsing, XML parsing, and binary parsing. Specifically, JSON parsing: Use a JSON parsing library (such as Python's json module) to parse a JSON string into a dictionary object. XML parsing: Use an XML parsing library (such as Python's xml.etree.ElementTree module) to parse an XML string into a tree structure. Binary parsing: According to the defined protocol specifications, write parsing code to parse binary data into structured information. For example, Python can be used to parse the binary data packet defined above.

[0050] Finally, the received data needs to be verified to ensure the integrity and accuracy of the data. Common data verification methods include data integrity verification, data consistency verification, and data rationality verification. Specifically, data integrity verification checks whether the data packet is complete and whether there is any missing or damaged data. For example, you can check whether the length and format of the data packet are as expected. Data consistency verification checks whether different data from the same data source are consistent, for example, checking whether the timestamp in the GPS data is synchronized with the vehicle's internal clock. Data rationality verification checks whether the data is within a reasonable range, for example, checking whether the vehicle speed exceeds the speed limit and checking whether the location exists on the map.

[0051] Accordingly, for abnormal data, corresponding processing measures need to be taken to ensure the normal operation of the system. Common abnormal data include GPS signal loss, sensor failure, data transmission interruption, etc. Exception handling methods include data interpolation, redundant sensors, and retransmission mechanisms. Specifically, for GPS signal loss, a data interpolation method can be used to estimate the lost data points based on the position and time information of the previous and next data points. For sensor failures, redundant sensors or backup sensors can be used to ensure data continuity and accuracy. For data transmission interruptions, a retransmission mechanism can be used to resend data packets that were not successfully transmitted to ensure data integrity.

[0052] In summary, using the vehicle communication unit to send vehicle status information to the back-end platform is a multi-step process, involving multiple links such as data packaging and sending, data receiving and processing, and data verification and exception handling. Through the above multiple links, the vehicle communication unit can efficiently and reliably send vehicle status information to the back-end platform, thus providing a solid foundation for realizing narrow road crossing warning and dispatch control.

[0053] In the above narrow road vehicle-on-road warning and dispatch control method, in step S130, on the back-end platform, vehicle-on-road warning analysis is performed based on the time data set of vehicle status information of each vehicle. Specifically, Figure 2 Flow chart of step S130 in the narrow road meeting warning and dispatch control method according to an embodiment of the present application. Figure 3 FIG. 1 is a data flow diagram of step S130 in the narrow road vehicle meeting warning and dispatch control method according to an embodiment of the present application. Figure 2 and Figure 3 As shown, the step S130 includes: S131, extracting a time data set of vehicle status information of the first vehicle and a time data set of vehicle status information of the second vehicle; S132, performing vehicle status feature extraction on the time data set of vehicle status information of the first vehicle and the time data set of vehicle status information of the second vehicle respectively to obtain a first vehicle status inference representation vector and a second vehicle status inference representation vector; S133, performing interactive fusion based on common feature guidance on the first vehicle status inference representation vector and the second vehicle status inference representation vector to obtain a first vehicle-second vehicle meeting state joint reasoning representation vector; S134, determining whether a narrow road meeting occurs based on the first vehicle-second vehicle meeting state joint reasoning representation vector.

[0054] Specifically, the step S131 extracts the time data set of the vehicle status information of the first vehicle and the time data set of the vehicle status information of the second vehicle. It should be understood that the time data set of the vehicle status information contains the dynamic behavior record of the vehicle status changing over time. By performing data time series analysis on it, the dynamic characteristics of the vehicle behavior can be captured, such as the behavior patterns of acceleration, deceleration, lane change, etc., so as to realize the prediction of the future behavior of the vehicle. To simplify the description, in the technical solution of the present application, only the data processing process of the prediction analysis of the meeting state between the first vehicle and the second vehicle is elaborated in detail, but it should be understood that in actual application, this method can be applied to the prediction of the meeting state between any two vehicles in the back-end platform, thereby realizing the intelligent scheduling of the entire vehicle group connected to the back-end platform.

[0055] Specifically, in step S132, vehicle state feature extraction is performed on the time data set of the vehicle state information of the first vehicle and the time data set of the vehicle state information of the second vehicle to obtain a first vehicle state estimation representation vector and a second vehicle state estimation representation vector. Specifically, Figure 4 FIG. 1 is a flow chart of step S132 in the narrow road vehicle meeting warning and dispatch control method according to an embodiment of the present application. Figure 4 As shown, the step S132 includes: S1321, extracting the time queue of the first vehicle position, the time queue of the first vehicle speed and the time queue of the first vehicle travel direction from the time data set of the vehicle state information of the first vehicle; S1322, extracting the time queue of the second vehicle position, the time queue of the second vehicle speed and the time queue of the second vehicle travel direction from the time data set of the vehicle state information of the second vehicle; S1323, inputting the time queue of the first vehicle position, the time queue of the first vehicle speed and the time queue of the first vehicle travel direction into the vehicle state inference module including the forward LSTM model and the Bayesian probability network to obtain the first vehicle state inference representation vector; S1324, inputting the time queue of the second vehicle position, the time queue of the second vehicle speed and the time queue of the second vehicle travel direction into the vehicle state inference module including the forward LSTM model and the Bayesian probability network to obtain the second vehicle state inference representation vector.

[0056] Here, it is considered that the position, speed and direction of travel of the vehicle usually have different data dimensions and numerical ranges. For example, the vehicle position data is expressed in the form of longitude and latitude coordinates, the speed is in kilometers per hour, and the direction of travel is expressed in angles. Therefore, in order to improve the pertinence and accuracy of data time series analysis, the present application further extracts the time queue of the first vehicle position, the time queue of the first vehicle speed and the time queue of the first vehicle travel direction from the time data set of the vehicle status information of the first vehicle, and extracts the time queue of the second vehicle position, the time queue of the second vehicle speed and the time queue of the second vehicle travel direction from the time data set of the vehicle status information of the second vehicle. That is, by dividing the time data set of the vehicle status information according to the parameter sample dimension, it is convenient to perform independent time series analysis on each parameter, thereby improving the accuracy of data analysis.

[0057] Then, a vehicle state inference module including a forward LSTM model and a Bayesian probability network is used to perform time series analysis and multimodal parameter joint reasoning on the time queues of the position, speed and direction of travel of the first vehicle and the second vehicle, respectively. Specifically, the vehicle state inference module first uses the forward LSTM model to perform time series analysis on the time queues of each vehicle parameter, so as to capture the long-term dependency and characteristic trend of each vehicle parameter over time by utilizing the long-term memory characteristics of the LSTM model, and respectively generate the time series feature representations of each vehicle parameter of the first vehicle and the second vehicle. Then, the Bayesian probability network performs joint interactive reasoning on the time series feature representations of each vehicle parameter of the first vehicle and the time series feature representations of each vehicle parameter of the second vehicle, respectively, based on the output results of the forward LSTM model. It should be known to those skilled in the art that the Bayesian probability network is a probability-based statistical model that captures and processes the mutual influence and dependency between each vehicle parameter through probabilistic reasoning, thereby calculating the state posterior probability distribution of the first vehicle and the second vehicle, respectively, and generating the first vehicle state inference representation vector and the second vehicle state inference representation vector, so as to provide important decision-making basis for the subsequent vehicle-meeting warning analysis.

[0058] Specifically, in step S133, the first vehicle state inference representation vector and the second vehicle state inference representation vector are interactively fused based on common feature guidance to obtain a first vehicle-second vehicle meeting state joint reasoning representation vector. That is, through feature fusion processing, the driving state information of the two vehicles is comprehensively analyzed to identify possible risks of meeting on a narrow road. In particular, in order to more fully capture the mutual influence between the two vehicles, the present application proposes an interactive fusion method based on common feature guidance, which extracts common features and strengthens differentiated features of the first vehicle state inference representation vector and the second vehicle state inference representation vector to identify and strengthen potential conflicts between the two vehicles, so as to improve the prediction accuracy of narrow road meeting risks.

[0059] Figure 5 FIG. 1 is a flow chart of step S133 in the narrow road vehicle meeting warning and dispatch control method according to an embodiment of the present application. Figure 5 As shown, the step S133 includes: S1331, extracting the common features between the first vehicle state inference representation vector and the second vehicle state inference representation vector to obtain the common feature representation vector of the first vehicle-second vehicle meeting state; S1332, inputting the first vehicle state inference representation vector, the second vehicle state inference representation vector and the first vehicle-second vehicle meeting state common feature representation vector into a vector probabilization unit based on a Sigmoid function to obtain a probabilistic first vehicle state inference representation vector, a probabilistic second vehicle state inference representation vector and a probabilistic first vehicle-second vehicle meeting state common feature representation vector. S1333, based on the probabilistic first vehicle-second vehicle meeting state common feature representation vector, performing differentiated feature enhancement compensation on the probabilistic first vehicle state inference representation vector and the probabilistic second vehicle state inference representation vector to obtain a first vehicle state feature compensation vector and a second vehicle state feature compensation vector; S1334, cascading the first vehicle state feature compensation vector, the second vehicle state feature compensation vector and the first vehicle-second vehicle meeting state common feature representation vector to obtain the first vehicle-second vehicle meeting state joint reasoning representation vector.

[0060] In a specific example of the present application, the step S1331 includes: using a neural network layer based on a tanh function to construct a common feature extraction network to extract common features from the first vehicle state inference representation vector and the second vehicle state inference representation vector to obtain a common feature representation vector of the first vehicle-second vehicle meeting state, which can be expressed by the formula:

[0061]

[0062] Among them, V1 represents the first vehicle state estimation representation vector, V 2 represents the second vehicle state estimation representation vector, V J represents the common feature representation vector of the meeting state between the first vehicle and the second vehicle, Indicates adding by position point, W VT and b VT Respectively represent the weight parameter matrix and the bias term, and tanh represents the hyperbolic tangent function. That is, firstly, a common feature extraction network is constructed using a neural network layer based on the tanh function to learn the common pattern between the first vehicle state inference representation vector and the second vehicle state inference representation vector, and to mine the shared features between the two, such as position information, so as to generate a common feature representation vector of the first vehicle-second vehicle meeting state.

[0063] In a specific example of the present application, the step S1332 inputs the first vehicle state inference representation vector, the second vehicle state inference representation vector and the first vehicle-second vehicle meeting state common feature representation vector into a vector probabilization unit based on a Sigmoid function, which can be expressed by the formula:

[0064] V a =sigmoid(V 1 )

[0065] V b =sigmoid(V 2 )

[0066] V c =sigmoid(V J )

[0067] Among them, V a represents the probabilistic first vehicle state estimation representation vector, V b represents the probabilistic second vehicle state estimation representation vector, V c represents the common feature representation vector of the probabilistic first vehicle-second vehicle meeting state, and sigmoid represents the sigmoid function. Here, the sigmoid function is used for normalization processing to make the feature value range between 0 and 1, so as to smooth the distribution of the feature value, thereby facilitating the subsequent feature interaction fusion processing.

[0068] In a specific example of the present application, the step S1333 includes: calculating the differential features of the probabilistic first vehicle state inference representation vector relative to the probabilistic first vehicle-second vehicle meeting state common feature representation vector to obtain the first vehicle state differential feature compensation representation vector; calculating the differential features of the probabilistic second vehicle state inference representation vector relative to the probabilistic first vehicle-second vehicle meeting state common feature representation vector to obtain the second vehicle state differential feature compensation representation vector; inputting the first vehicle state differential feature compensation representation vector and the first vehicle state inference representation vector into the fine-grained feature compensation module to obtain the first vehicle state feature compensation vector; inputting the second vehicle state differential feature compensation representation vector and the second vehicle state inference representation vector into the fine-grained feature compensation module to obtain the second vehicle state feature compensation vector. The fine-grained feature compensation module obtains the first vehicle state feature compensation vector by calculating the position point multiplication between the first vehicle state differential feature compensation representation vector and the first vehicle state inference representation vector, and obtains the second vehicle state feature compensation vector by calculating the position point multiplication between the second vehicle state differential feature compensation representation vector and the second vehicle state inference representation vector.

[0069] Accordingly, the above step S1333 can be expressed by the formula:

[0070]

[0071] v 1a =V 1 ⊙v a '

[0072] v 2b =V 2 ⊙v b '

[0073] in, and denote the ith eigenvalue of the probabilistic first vehicle state inference representation vector, the ith eigenvalue of the probabilistic second vehicle state inference representation vector, and the ith eigenvalue of the probabilistic first vehicle-second vehicle meeting state common feature representation vector, log(·) denotes a logarithmic function with base 2, exp(·) denotes a natural exponential function, softmax denotes a normalized exponential function, and v a ' and v b 'represents the first vehicle state differential feature compensation representation vector and the second vehicle state differential feature compensation representation vector, ⊙ represents the position point multiplication, v 1a and v 2bRespectively represent the first vehicle state characteristic compensation vector and the second vehicle state characteristic compensation vector.

[0074] That is, by respectively calculating the differentiated features of the probabilistic first vehicle state inference representation vector and the probabilistic second vehicle state inference representation vector relative to the probabilistic first vehicle-second vehicle meeting state common feature representation vector, the unique feature information of the first vehicle state features and the second vehicle state features in addition to the common features is mined, and this is used as a weight to compensate and strengthen the original first vehicle state inference representation vector and the second vehicle state inference representation vector by position point multiplication, thereby effectively strengthening the potential conflict feature representation between the two vehicles, ensuring that during the fusion process, as much unique information of the first vehicle state and the second vehicle state features is retained as much as possible.

[0075] In a specific example of the present application, the step S1334, cascading the first vehicle state feature compensation vector, the second vehicle state feature compensation vector and the first vehicle-second vehicle meeting state common feature representation vector, can be expressed by the formula:

[0076] v f =concat[V J ;v 1a ;v 2b ]

[0077] Among them, concat[·;·;] represents the cascade operation, v f The first vehicle and the second vehicle meeting state joint reasoning representation vector is represented. That is, the feature fusion is performed through the cascade operation to comprehensively consider the common information of the state feature of the first vehicle and the state feature of the second vehicle as well as their respective unique information, and generate the first vehicle and the second vehicle meeting state joint reasoning representation vector, so as to more accurately reflect the mutual influence between the two vehicles.

[0078] Specifically, the step S134 determines whether a narrow road meeting occurs based on the joint reasoning representation vector of the first vehicle-second vehicle meeting state. In a specific example of the present application, the step S134 includes: inputting the joint reasoning representation vector of the first vehicle-second vehicle meeting state into a meeting warning module based on a classifier to obtain a meeting warning result, and the meeting warning result is used to indicate whether the first vehicle and the second vehicle meet on a narrow road. Specifically, the classifier learns the driving state information of the first vehicle and the second vehicle contained in the joint reasoning representation vector of the first vehicle-second vehicle meeting state through a fully connected layer, and performs probability distribution calculation on the feature representation learned by the fully connected layer through a softmax classification function, and generates the probability value of the first vehicle and the second vehicle meeting on a narrow road and the probability value of not meeting on a narrow road. Finally, by selecting the classification label with the highest probability as the final meeting warning result, real-time warning information is provided for vehicle scheduling.

[0079] In the above-mentioned narrow road meeting warning and dispatch control method, the step S140, on the back-end platform, performs dispatch control based on the time data set of the vehicle status information of each vehicle. In a specific example of the present application, the step S140 includes: in response to the meeting warning result that the first vehicle and the second vehicle meet on a narrow road, generating a meeting deceleration prompt to the first vehicle and the second vehicle. That is, when the meeting warning result indicates that the first vehicle and the second vehicle are at risk of meeting on a narrow road, a deceleration prompt will be automatically sent to the two vehicles through the vehicle communication system to ensure that the two vehicles can safely decelerate when meeting and avoid the risk of collision caused by excessive speed. In an embodiment of the present application, the meeting deceleration prompt is sent to the driver through the vehicle's on-board communication system, or directly controls the vehicle to perform a deceleration operation in an autonomous driving vehicle.

[0080] In summary, the method for warning and dispatching control of meeting vehicles on narrow roads according to the embodiment of the present application is explained, which collects the position, speed and direction of travel information of the vehicle through the on-board sensor, uses the on-board communication unit to send the vehicle status information to the back-end platform, and introduces artificial intelligence technology based on deep learning on the back-end platform to perform time series analysis on the vehicle status information of each vehicle, and realizes intelligent prediction and dispatching control of the meeting status between vehicles by joint interactive reasoning of the driving status characteristics of different vehicles. In this way, when it is identified that a narrow road section is about to meet, a deceleration warning prompt can be issued in time, thereby effectively reducing the risk of traffic accidents and improving road traffic efficiency.

[0081] Figure 6 FIG. 1 is a block diagram of a narrow road vehicle crossing warning and dispatch control system according to an embodiment of the present application. Figure 6As shown, according to the embodiment of the present application, the narrow road vehicle-crossing warning and dispatch control system 100 includes: a vehicle status information acquisition module 110, which is used to collect vehicle status information through vehicle-mounted sensors, and the vehicle status information includes vehicle position, vehicle speed and travel direction; an information transmission module 120, which is used to send the vehicle status information to a back-end platform using a vehicle-mounted communication unit; a vehicle-crossing warning analysis module 130, which is used to perform vehicle-crossing warning analysis on the back-end platform based on a time data set of vehicle status information of each vehicle; and a dispatch control module 140, which is used to perform dispatch control on the back-end platform based on a time data set of vehicle status information of each vehicle.

[0082] Here, those skilled in the art can understand that the specific operations of each module in the above narrow road vehicle crossing warning and dispatching control system have been described in the above reference. Figures 1 to 5 It has been described in detail in the description of the narrow road oncoming vehicle warning and dispatching control method, and therefore, its repeated description will be omitted.

[0083] The basic principle of the present invention is described above in conjunction with specific embodiments. However, it should be pointed out that the advantages, strengths, effects, etc. mentioned in the present invention are only examples and not limitations, and it cannot be considered that these advantages, strengths, effects, etc. must be possessed by each embodiment of the present invention. In addition, the specific details of the above embodiments are only for the purpose of illustration and facilitation of understanding, rather than limitation, and the above details do not limit the present invention to being implemented by adopting the above specific details.

Claims

1. A method for early warning and dispatching control of vehicles on narrow roads, characterized in that: include: Collecting vehicle status information through vehicle-mounted sensors, wherein the vehicle status information includes vehicle position, vehicle speed and travel direction; Using the vehicle communication unit to send the vehicle status information to the back-end platform; On the back-end platform, oncoming vehicle warning analysis is performed based on the time data set of vehicle status information of each vehicle; On the back-end platform, scheduling control is performed based on a time data set of vehicle status information of each vehicle; Among them, the oncoming vehicle warning analysis is performed based on the time data set of the vehicle status information of each vehicle, including: extracting the time data set of the vehicle status information of the first vehicle and the time data set of the vehicle status information of the second vehicle; performing vehicle status feature extraction on the time data set of the vehicle status information of the first vehicle and the time data set of the vehicle status information of the second vehicle respectively to obtain a first vehicle status inference representation vector and a second vehicle status inference representation vector; performing interactive fusion based on common feature guidance on the first vehicle status inference representation vector and the second vehicle status inference representation vector to obtain a first vehicle-second vehicle meeting state joint reasoning representation vector; based on the first vehicle-second vehicle meeting state joint reasoning representation vector, determining whether a narrow road meeting occurs.

2. The narrow road vehicle meeting warning and dispatching control method according to claim 1 is characterized in that: Extracting vehicle state features from a time data set of vehicle state information of the first vehicle and a time data set of vehicle state information of the second vehicle to obtain a first vehicle state inference representation vector and a second vehicle state inference representation vector, comprising: Extracting a time sequence of a first vehicle position, a time sequence of a first vehicle speed, and a time sequence of a first vehicle travel direction from a time data set of vehicle state information of the first vehicle; Extracting a time sequence of the second vehicle position, a time sequence of the second vehicle speed, and a time sequence of the second vehicle travel direction from the time data set of the vehicle state information of the second vehicle; Inputting the time queue of the first vehicle position, the time queue of the first vehicle speed, and the time queue of the first vehicle travel direction into a vehicle state inference module including a forward LSTM model and a Bayesian probability network to obtain the first vehicle state inference representation vector; The time queue of the second vehicle position, the time queue of the second vehicle speed and the time queue of the second vehicle travel direction are input into the vehicle state inference module including the forward LSTM model and the Bayesian probability network to obtain the second vehicle state inference representation vector.

3. The narrow road vehicle meeting warning and dispatching control method according to claim 2 is characterized in that: Interactively fusing the first vehicle state inference representation vector and the second vehicle state inference representation vector based on common feature guidance to obtain a first vehicle-second vehicle meeting state joint reasoning representation vector, including: Extracting common features between the first vehicle state inference representation vector and the second vehicle state inference representation vector to obtain a common feature representation vector of a first vehicle-second vehicle meeting state; Inputting the first vehicle state inference representation vector, the second vehicle state inference representation vector and the first vehicle-second vehicle meeting state common feature representation vector into a vector probabilization unit based on a sigmoid function to obtain a probabilistic first vehicle state inference representation vector, a probabilistic second vehicle state inference representation vector and a probabilistic first vehicle-second vehicle meeting state common feature representation vector; Based on the probabilistic first vehicle-second vehicle meeting state common feature representation vector, performing differential feature enhancement compensation on the probabilistic first vehicle state inference representation vector and the probabilistic second vehicle state inference representation vector to obtain a first vehicle state feature compensation vector and a second vehicle state feature compensation vector; The first vehicle state feature compensation vector, the second vehicle state feature compensation vector and the first vehicle-second vehicle meeting state common feature representation vector are cascaded to obtain the first vehicle-second vehicle meeting state joint reasoning representation vector.

4. The narrow road vehicle meeting warning and dispatching control method according to claim 3 is characterized in that: Extracting common features between the first vehicle state inference representation vector and the second vehicle state inference representation vector to obtain a common feature representation vector of a first vehicle-second vehicle meeting state, including: A common feature extraction network is constructed using a neural network layer based on a tanh function to extract common features from the first vehicle state inference representation vector and the second vehicle state inference representation vector to obtain a common feature representation vector of the first vehicle-second vehicle meeting state.

5. The narrow road vehicle meeting warning and dispatching control method according to claim 4 is characterized in that: Based on the probabilistic first vehicle-second vehicle meeting state common feature representation vector, performing differentiated feature enhancement compensation on the probabilistic first vehicle state inference representation vector and the probabilistic second vehicle state inference representation vector to obtain a first vehicle state feature compensation vector and a second vehicle state feature compensation vector, including: Calculating the differential features of the probabilistic first vehicle state inference representation vector relative to the probabilistic first vehicle-second vehicle meeting state common feature representation vector to obtain a first vehicle state differential feature compensation representation vector; Calculating the differential features of the probabilistic second vehicle state inference representation vector relative to the probabilistic first vehicle-second vehicle meeting state common feature representation vector to obtain a second vehicle state differential feature compensation representation vector; Inputting the first vehicle state differentiated feature compensation representation vector and the first vehicle state inference representation vector into a fine-grained feature compensation module to obtain the first vehicle state feature compensation vector; The second vehicle state differentiated feature compensation representation vector and the second vehicle state inference representation vector are input into the fine-grained feature compensation module to obtain the second vehicle state feature compensation vector.

6. The narrow road vehicle meeting warning and dispatching control method according to claim 5, characterized in that: Inputting the first vehicle state differential feature compensation representation vector and the first vehicle state inference representation vector into a fine-grained compensation module to obtain the first vehicle state feature compensation vector, including: The first vehicle state feature compensation vector is obtained by calculating the point-by-point multiplication between the first vehicle state differentiated feature compensation representation vector and the first vehicle state inference representation vector.

7. The narrow road vehicle meeting warning and dispatching control method according to claim 6, characterized in that: Determining whether a narrow road meeting occurs based on the joint reasoning representation vector of the meeting state of the first vehicle and the second vehicle includes: The first vehicle-second vehicle meeting state joint reasoning representation vector is input into a classifier-based meeting warning module to obtain a meeting warning result, wherein the meeting warning result is used to indicate whether the first vehicle and the second vehicle meet on a narrow road.

8. The narrow road vehicle meeting warning and dispatching control method according to claim 7, characterized in that: Scheduling control is performed based on the time data set of vehicle status information of each vehicle, including: In response to the meeting warning result being that the first vehicle and the second vehicle meet on a narrow road, a meeting deceleration prompt is generated to the first vehicle and the second vehicle.

9. A narrow road vehicle meeting warning and dispatching control system, characterized in that: include: A vehicle status information collection module is used to collect vehicle status information through vehicle-mounted sensors, wherein the vehicle status information includes vehicle position, vehicle speed and travel direction; An information transmission module, used to send the vehicle status information to a back-end platform using an on-board communication unit; A vehicle-on-vehicle warning analysis module, used for performing vehicle-on-vehicle warning analysis based on a time data set of vehicle status information of each vehicle on the back-end platform; The dispatch control module is used to perform dispatch control on the back-end platform based on the time data set of vehicle status information of each vehicle.

Images