A regional risk warning method and system based on precise positioning and path prediction

Abstract

This application relates to a method and system for regional risk warning based on precise positioning and path prediction, belonging to the field of intelligent transportation. The method includes acquiring map data and electronic fence information of a first area, the map data including lane information; acquiring the positioning information and status information of a first vehicle, the positioning information and status information being used to construct an input feature vector; performing lane matching based on the positioning information and lane information to determine the first lane of the first vehicle; generating an LSTM model based on the input feature vector through training, validation, and testing, and outputting the predicted path of the first vehicle; performing traversal matching between the predicted path and the electronic fence information, and displaying the first matched information. The regional risk warning method and system based on precise positioning and path prediction of this application automatically completes the retrieval and matching of risk areas determined by temporary fences based on the high-precision positioning and path prediction results of the vehicle's location, and can automatically notify the driver, improving the timeliness and effectiveness of information transmission.

Description

Technical Field

[0001] The present application relates to the field of intelligent transportation, and in particular to a method and system for regional risk warning based on precise positioning and path prediction. Background Art

[0002] Since the outbreak, there has often been a need to demarcate some temporary irregular risk area electronic fences, which has placed high demands on the management and control of geographical areas. However, the locations of the above-mentioned areas are temporary, and their attributes (risk levels, etc.) may change at any time, causing certain troubles for personal travel and driving. The existing form of obtaining risk area information is generally for drivers to independently select some traffic information announcement systems, or pay attention to the travel areas of interest in certain travel reminder apps. They need to frequently and actively browse to obtain and confirm the latest temporary risk area electronic fence information related to themselves.

[0003] Existing technologies require drivers to frequently and actively browse information. Most of the information is not relevant to the driver, and the proportion of information that is truly valuable to the driver's travel is very small. In addition, when truly effective information appears, the driver is likely to miss the information because he did not check it in the first time and eventually enter the risk area. At the same time, existing technologies do not predict the driving route, and the relevance of the information is entirely determined by the driver based on the travel route. The above-mentioned manual judgment method can only be used in scenarios with a small amount of information (such as dozens of pieces of information). When the amount of information is large (hundreds or even thousands of pieces), manual judgment is not feasible.

[0004] Therefore, it is desired to provide a regional risk warning method and system based on precise positioning and path prediction. According to the high-precision positioning and path prediction results of the vehicle's location, the program automatically completes the retrieval and matching of the risk area determined by the temporary fence and can automatically notify the driver, thereby improving the timeliness and effectiveness of information transmission. Summary of the invention

[0005] According to a first aspect of some embodiments of the present application, a method for regional risk warning based on precise positioning and path prediction is provided, which is applied to a terminal (for example, a smart terminal, etc.), and the method may include obtaining map data and electronic fence information of a first area, the map data including lane information; obtaining positioning information and status information of a first vehicle, the positioning information and the status information are used to construct an input feature vector; performing lane matching according to the positioning information and the lane information, and determining a first lane of the first vehicle; based on the input feature vector, generating an LSTM model through training, verification, and testing, and outputting a predicted path of the first vehicle; performing traversal matching on the predicted path and the electronic fence information, and displaying the matched first information.

[0006] In some embodiments, the obtaining of map data and electronic fence information of the first area specifically includes obtaining map data of the first area sent by a roadside device through a V2X channel; obtaining electronic fence information of the first area through a 4G channel, wherein the electronic fence information includes information of irregular risk areas.

[0007] In some embodiments, the obtaining of the positioning information and status information of the first vehicle specifically includes obtaining positioning information through a GNSS module, wherein the positioning information includes high-precision positioning, heading angle, and vehicle speed information; and obtaining status information through a vehicle body CAN bus, wherein the status information includes steering wheel angle, turn signal status, throttle status, and gear status information.

[0008] In some embodiments, performing lane matching based on the positioning information and the lane information to determine the lane of the first vehicle specifically includes traversing the first node, the first child node, and the first adjacent point pair of the first child node of the map data; for each first adjacent point pair, determining the first distance, the first sign, and the first angle between the first vehicle and the first adjacent point pair; if the first distance, the first sign, and the first angle do not meet a first preset threshold, returning to traversing the first adjacent point pair.

[0009] In some embodiments, if the first distance, the first sign, and the first angle meet a first preset threshold, it specifically includes traversing the second child node associated with the first child node and the second adjacent point pairs of the second child node; for each second adjacent point pair, determining the second distance, second sign and second angle between the first vehicle and the second adjacent point pair; if the second distance, the second sign, and the second angle do not meet the second preset threshold, returning to traversing the second adjacent point pairs; if the second distance, the second sign, and the second angle meet the second preset threshold, determining the lane matched by the first vehicle.

[0010] In some embodiments, the positioning information and the state information are used to construct an input feature vector, specifically including obtaining a sampling point X of the first vehicle in the first lane. t0 To X tn ; Get the average heading angle of the first vehicle in the first lane as X t(n+1) , the average speed of the first vehicle in the first lane is X t(n+2) , the average steering wheel angle of the first vehicle in the first lane is X t(n+3) , the turn signal state of the first vehicle in the first lane is X t(n+4) , the throttle state of the first vehicle in the first lane is X t(n+5) , the gear state of the first vehicle in the first lane is X t(n+6); Construct the input feature vector corresponding to the t-th frame data, X t =[X t0 ,X t1 ,X t2 ,…,X tn ,X t(n+1) ,…,X t(n+6) ].

[0011] In some embodiments, the generating of the LSTM model based on the input feature vector through training, verification, and testing to output the predicted path of the first vehicle specifically includes obtaining the turning operation of the first vehicle at the end of the first lane as h t0 The steering operations include left turn, right turn, straight ahead, and U-turn. Construct the output vector h corresponding to the t-th frame data t =[h t0 ]; Based on the historical vehicle operation data, construct the training set, validation set, and test set; construct the LSTM model:

[0012] f t =σ g (W f x t +U f h t-1 +b f )

[0013] i t =σ g (W i x t +U i h t-1 +b i )

[0014] o t =σ g (W o x t +U o h t-1 +b o )

[0015]

[0016]

[0017]

[0018] Among them, X t is the input feature vector, H t is the output vector, f t is the excitation vector used to control the forget gate, i t is the excitation vector used to control the input gate, o tis the excitation vector used to control the output gate, Input the excitation vector for the cell, c t is the cell state vector, σ is the activation function, where: σ g is the sigmoid function, σ c is the hyperbolic tangent function, σ h is the hyperbolic tangent function or σ h (x) = x, W, U, and b are weight parameters and offset parameters respectively.

[0019] In some embodiments, the method may specifically include determining the LSTM loss function; training the LSTM model based on the training set and the loss function, and determining the optimal weight parameters and offset parameters of W, U, and b; performing overfitting verification on the LSTM model based on the verification set; testing the LSTM model based on the test set; if the test meets a third preset threshold, determining the predicted path of the first vehicle.

[0020] In some embodiments, the method may specifically include obtaining a point trace sequence of the predicted path; traversing the point trace sequence and matching it with the information of the irregular risk area; if there is a match, displaying the matching first information, the first information including the risk area and risk level of the predicted path.

[0021] According to a second aspect of some embodiments of the present application, a system is provided, comprising: a memory configured to store data and instructions; a processor that establishes communication with the memory, wherein when executing instructions in the memory, the processor is configured as follows: the method may include obtaining map data and electronic fence information of a first area, the map data including lane information; obtaining positioning information and status information of a first vehicle, the positioning information and the status information being used to construct an input feature vector; performing lane matching based on the positioning information and the lane information to determine a first lane of the first vehicle; generating an LSTM model through training, verification, and testing based on the input feature vector, and outputting a predicted path of the first vehicle; performing traversal matching on the predicted path and the electronic fence information, and displaying the matched first information.

[0022] Therefore, according to the regional risk warning method and system based on precise positioning and path prediction in some embodiments of the present application, based on the high-precision positioning and path prediction results of the vehicle's location, the program automatically completes the retrieval and matching of the risk area determined by the temporary fence and can automatically notify the driver, thereby improving the timeliness and effectiveness of information transmission. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to better understand and explain some embodiments of the present application, the following description of the embodiments will be made with reference to the accompanying drawings, in which the same reference numerals indicate corresponding parts in the drawings.

[0024] Figure 1 It is an exemplary schematic diagram of a regional risk warning system based on precise positioning and path prediction provided according to some embodiments of the present application.

[0025] Figure 2 It is an exemplary flowchart of a regional risk warning method based on precise positioning and path prediction provided according to some embodiments of the present application. DETAILED DESCRIPTION

[0026] The following description with reference to the accompanying drawings is intended to facilitate a comprehensive understanding of the various embodiments of the present application as defined by the claims and their equivalents. These embodiments include various specific details for ease of understanding, but these are considered to be exemplary only. Therefore, it will be appreciated by those skilled in the art that various changes and modifications to the various embodiments described herein will not depart from the scope and spirit of the present application. In addition, in order to briefly and clearly describe the present application, the present application will omit the description of known functions and structures.

[0027] The terms and phrases used in the following specification and claims are not limited to the literal meanings, but are only to enable a clear and consistent understanding of the present application. Therefore, it will be appreciated by those skilled in the art that the description of various embodiments of the present application is provided only for the purpose of illustration, rather than to limit the present application of the appended claims and their equivalent definitions.

[0028] The following will be combined with the drawings in some embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0029] It should be noted that the terms used in the embodiments of the present application are only for the purpose of describing specific embodiments, and are not intended to limit the present application. The singular forms of "one", "one", "a kind of", "said" and "the" used in the embodiments of the present application and the appended claims are also intended to include plural forms, unless the context clearly indicates other meanings. It should also be understood that the term "and / or" used in the present application refers to and includes any or all possible combinations of one or more bound listed items. The expressions "first", "second", "the first" and "the second" are used to modify the corresponding elements without considering the order or importance, and are only used to distinguish one element from another element, without limiting the corresponding elements.

[0030] According to some embodiments of the present application, the terminal may be an intelligent terminal, a platform, equipment and / or an electronic device, etc. The platform may include a cloud control platform, etc. The platform may include a system platform composed of one or more electronic devices; the equipment may include an intelligent connected vehicle (ICV); the electronic device may include a personal computer (PC, such as a tablet computer, a desktop computer, a notebook, a netbook, a handheld computer PDA), a client device, a virtual reality device (VR), an augmented reality device (AR), a mixed reality device (MR), an XR device, a rendering machine, a smart phone, a mobile phone, an e-book reader, a portable multimedia player (PMP), an audio / video player (MP3 / MP4), a camera, and a wearable device, etc. One or a combination of several. According to some embodiments of the present application, the wearable device may include an accessory type (such as a watch, a ring, a bracelet, glasses, or a head-mounted device (HMD)), an integrated type (such as electronic clothing), a decorative type (such as a skin pad, a tattoo, or a built-in electronic device), etc., or a combination of several. In some embodiments of the present application, the electronic device may be flexible, not limited to the above-mentioned devices, or may be a combination of one or several of the above-mentioned various devices. In this application, the term "user" may refer to a person using an electronic device or a device (eg, an artificial intelligence electronic device) using an electronic device.

[0031] The embodiment of the present application provides a method and system for regional risk warning based on precise positioning and path prediction. In order to facilitate understanding of the embodiment of the present application, the embodiment of the present application will be described in detail below with reference to the accompanying drawings.

[0032] Figure 1 FIG. 1 is an exemplary schematic diagram of a regional risk warning system based on precise positioning and path prediction according to some embodiments of the present application. Figure 1The regional risk warning system 100 based on precise positioning and path prediction may include a network 110, a control terminal 120, a user terminal 130, and a server 140. Specifically, the control terminal 120 and the user terminal 130 establish communication through the network. For example, the control terminal 120 and the user terminal 130 can communicate in the same local area network (for example, the network environment of the same router, etc.). Further, the control terminal 120 can be connected to the network 110 by wire (for example, a network cable, etc.) or wireless (for example, a cloud server, etc.), and the user terminal 130 can establish a communication connection with the network 110 by wire or wireless (for example, WIFI, etc.). In some embodiments, the user terminal 130 can send vehicle positioning information, status information, etc. to the control terminal 120 and the server 140. Further, the control terminal 120 and the server 140 can feedback map data, electronic fence information, etc. to the user terminal 130. According to the feedback information, the user terminal 130 can change the path planning, etc. As an example, the server 140 and / or the control terminal 120 may obtain a traversal matching result of the predicted path and the electronic fence information, and the matching result may include an irregular risk area in the predicted path.

[0033] According to some embodiments of the present application, the control terminal 120 and the user terminal 130 may be the same or different terminal devices, etc. The terminal devices may include but are not limited to intelligent terminals, cloud control platforms, mobile terminals, computers, etc. In the intelligent transportation scenario, the control terminal 120 may include electronic devices, etc., and the user terminal 130 may include display devices, etc. In some embodiments, the control terminal 120 and the user terminal 130 may be integrated into one device, for example, an intelligent electronic device of the user terminal, etc. In some embodiments, the server 140 is a type of computer, which has advantages such as faster operation and higher load than ordinary computers, and the corresponding price is higher. In a network environment, the server can provide computing or application services for other clients (for example, PCs, smart phones, ATMs and other terminals, as well as large equipment such as transportation systems). The server has high-speed CPU computing power, long-term reliable operation, powerful I / O external data throughput and better scalability. The services that the server can provide include but are not limited to the ability to respond to service requests, undertake services, and guarantee services. As an electronic device, the server has an extremely complex internal structure, including an internal structure similar to that of an ordinary computer. As an example, the internal structure of the server may include a central processing unit (CPU), a hard disk, a memory, a system, a system bus, etc.

[0034] In some embodiments of the present application, the regional risk warning system 100 based on precise positioning and path prediction may omit one or more elements, or may further include one or more other elements. As an example, the regional risk warning system 100 based on precise positioning and path prediction may include multiple user terminals 130, such as multiple intelligent equipment, etc. For another example, the regional risk warning system 100 based on precise positioning and path prediction may include one or more control terminals 120. For another example, the regional risk warning system 100 based on precise positioning and path prediction may include multiple servers 140, etc. In some embodiments, the regional risk warning system 100 based on precise positioning and path prediction may include but is not limited to a system based on urban intelligent traffic scene processing. The network 110 may be any type of communication network, and the communication network may include a computer network (for example, a local area network (LAN) or a wide area network (WAN), the Internet and / or a telephone network, etc., or a combination of several. In some embodiments, the network 110 may be other types of wireless communication networks. The wireless communication may include microwave communication and / or satellite communication, etc. The wireless communication may include cellular communication, for example, Global System for Mobile Communications (GSM), Code Division Multiple Access (CDMA), third generation mobile communication (3G), fourth generation mobile communication (4G), fifth generation mobile communication (5G), sixth generation mobile communication (6G), Long Term Evolution (LTE), LTE-A (LTE-Advanced), Wideband Code Division Multiple Access (WCDMA), Universal Mobile Telecommunications System (UMTS), wireless broadband (WiBro), etc., or a combination of several thereof.In some embodiments, the user terminal 130 may be other equipment and / or electronic devices with equivalent functional modules, and the equipment and / or electronic devices may include one or a combination of intelligent connected vehicles (ICV), virtual reality equipment (VR), rendering machines, personal computers (PCs, such as tablet computers, desktop computers, notebooks, netbooks, handheld computers (PDAs), smart phones, mobile phones, e-book readers, portable multimedia players (PMPs), audio / video players (MP3 / MP4s), cameras, and wearable devices.

[0035] In some embodiments, the WIFI may be other types of wireless communication technologies. According to some embodiments of the present application, the wireless communication may include wireless LAN (WiFi, Wireless Fidelity), Bluetooth, Bluetooth Low Energy (BLE, Bluetooth Low Energy), ZigBee, Near Field Communication (NFC, Near Field Communication), magnetic security transmission, radio frequency and body area network (BAN, Body Area Network), etc., or a combination of several. According to some embodiments of the present application, the wired communication may include a global navigation satellite system (Glonass / GNSS, Global Navigation Satellite System), a global positioning system (GPS, Global Position System), a Beidou navigation satellite system or Galileo (European Global Satellite Navigation System), etc. The wired communication may include a universal serial bus (USB, Universal Serial Bus), a high-definition multimedia interface (HDMI, High-Definition Multimedia Interface), a recommended standard 232 (RS-232, Recommend Standard 232), and / or a simple old telephone service (POTS, Plain Old Telephone Service), etc., or a combination of several.

[0036] It should be noted that the above description of the regional risk warning system 100 based on precise positioning and path prediction is only for the convenience of description and does not limit the present application to the scope of the embodiments cited. It can be understood that for those skilled in the art, based on the principle of the present system, it is possible to arbitrarily combine the various components, or form a subsystem to connect with other components without deviating from the principle, and make various corrections and changes in form and details to the application field of the above-mentioned method and system. For example, the server 140 and / or the control terminal 120 can obtain map data through roadside equipment, etc. Such variations are within the scope of protection of the present application.

[0037] Figure 2 is an exemplary flow chart of a regional risk warning method based on precise positioning and path prediction according to some embodiments of the present application. Figure 2 The process 200 can be implemented by the regional risk warning system 100 based on precise positioning and path prediction. In some embodiments, the regional risk warning method 200 based on precise positioning and path prediction can be automatically started or started by instructions. The instructions can include system instructions, device instructions, user instructions, action instructions, etc., or a combination of several.

[0038] At 201, map data and electronic fence information of the first area are obtained, and the map data includes lane information. Operation 201 can be implemented by the control terminal 120 and the server 140 of the regional risk warning system 100 based on precise positioning and path prediction. In some embodiments, the control terminal 120 and / or the server 140 can obtain map data and electronic fence information of the first area, and the map data includes lane information. In some embodiments, the control terminal 120 and / or the server 140 can obtain map data of the first area sent by the roadside equipment through the V2X channel. In some embodiments, the control terminal 120 and / or the server 140 can obtain electronic fence information of the first area through the 4G channel, and the electronic fence information includes information of irregular risk areas. As an example, the first area may include an area within a preset distance of the current user terminal 130, for example, the surrounding area of ​​the current vehicle, etc. The irregular risk area may include risk area location coordinate information, risk level information, etc.

[0039] At 202, the positioning information and status information of the first vehicle are obtained, and the positioning information and the status information are used to construct an input feature vector. Operation 202 can be implemented by the control terminal 120 and the server 140 of the regional risk warning system 100 based on precise positioning and path prediction. In some embodiments, the control terminal 120 and / or the server 140 can obtain the positioning information and status information of the first vehicle, and the positioning information and the status information are used to construct an input feature vector. As an example, the control terminal 120 and / or the server 140 can obtain positioning information through a GNSS module, and the positioning information may include but is not limited to high-precision positioning, heading angle, vehicle speed information, etc. As an example, the control terminal 120 and / or the server 140 can obtain status information through the vehicle body CAN bus, and the status information may include but is not limited to steering wheel angle, turn signal status, throttle status, gear status information, etc.

[0040] In 203, lane matching is performed according to the positioning information and the lane information to determine the first lane of the first vehicle. Operation 203 may be implemented by the control terminal 120 and the server 140 of the regional risk warning system 100 based on precise positioning and path prediction. In some embodiments, the control terminal 120 and / or the server 140 may perform lane matching according to the positioning information and the lane information to determine the first lane of the first vehicle. The first vehicle may include the vehicle where the current user terminal 130 is located, and the first lane may include the lane of the vehicle where the current user terminal 130 is located, etc.

[0041] In some embodiments, the control end 120 and / or the server 140 can traverse the first node (e.g., node node), the first child node (e.g., link node), and the first adjacent point pair (e.g., adjacent link_point pair) of the map data; for each first adjacent point pair, determine the first distance (distance), the first flag (flag), and the first angle (angle) between the first vehicle and the first adjacent point pair; if the first distance, the first flag, and the first angle do not meet the first preset threshold, return to traversing the first adjacent point pair. Further, if the first distance, the first sign, and the first angle meet a first preset threshold, the method specifically includes traversing the second child node (for example, lane node) associated with the first child node (for example, link node), and the second adjacent point pairs (for example, adjacent lane_point pairs) of the second child node; for each second adjacent point pair, determining the second distance, second sign, and second angle between the first vehicle and the second adjacent point pair; if the second distance, the second sign, and the second angle do not meet the second preset threshold, returning to traversing the second adjacent point pairs; if the second distance, the second sign, and the second angle meet the second preset threshold, determining the lane matched by the first vehicle.

[0042] As an example, step 203 may specifically include a lane matching algorithm process 300, which may perform lane matching based on the high-precision positioning, heading angle and lane information in the map data (MAP data) of the first vehicle to determine the first lane where the first vehicle is located, as follows.

[0043] At 301 , the node nodes in the MAP data are traversed. If the traversal is not completed, the process goes to operation 302 . After the traversal is completed, the process goes to operation 311 .

[0044] In 302, for each node in operation 301, traverse the link nodes under it. If the traversal is not completed, proceed to operation 303. If the traversal is completed, proceed to operation 311.

[0045] In 303, for each link node in operation 302, the first adjacent point pair (adjacent link_point pair) under it is traversed. If the traversal is not completed, the process proceeds to operation 304. If the traversal is completed, the process proceeds to operation 311.

[0046] At 304, for each link_point pair in operation 303, calculate the first distance distance between it and the two points in combination with the first vehicle GNSS positioning information, and whether the projection of the first vehicle onto the straight line determined by the two points is between the two points (yes: set the first flag flag to 0, no: set flag to 1 or -1); in combination with the first vehicle GNSS heading information, calculate the first angle angle between the vehicle and the link_point pair.

[0047] At 305, determine whether the first distance distance, the first angle angle, and the first flag flag calculated in operation 304 meet the first preset threshold: (flag == 0) && (distance < link_width / 2) && (fabs(angle) < 0.53); if the first preset threshold is not met, return to operation 303; if the first preset threshold is met, enter operation 306.

[0048] At 306, for each link node in operation 302, traverse the lane nodes under it. If the traversal is not ended, enter operation 307; if the traversal is ended, enter operation 311.

[0049] At 307, for each lane node in operation 306, traverse the second adjacent point pairs (adjacent lane_point pairs) under it. If the traversal is not ended, enter operation 308; if the traversal is ended, enter operation 311.

[0050] At 308, for each lane_point pair in operation 307, calculate the second distance distance between it and the two points in combination with the first vehicle GNSS positioning information, and whether the projection of the first vehicle onto the straight line determined by the two points is between the two points (yes: set the second flag flag to 0, no: set flag to 1 or -1); in combination with the first vehicle GNSS heading information, calculate the second angle angle between the vehicle and the lane_point pair.

[0051] At 309, determine whether the second distance distance, the second angle angle, and the second flag flag calculated in operation 308 meet the second preset threshold: (flag == 0) && (distance < lane_width / 2) && (fabs(angle) < 0.53); if the second preset threshold is not met, return to operation 307; if the second preset threshold is met, enter operation 310.

[0052] At 310, record the matching lane information.

[0053] At 311, end the lane matching algorithm.

[0054] In some embodiments, the positioning information and the state information are used to construct an input feature vector, specifically including obtaining a sampling point X of the first vehicle in the first lane. t0 To X tn ; Get the average heading angle of the first vehicle in the first lane as X t(n+1) , the average speed of the first vehicle in the first lane is X t(n+2) , the average steering wheel angle of the first vehicle in the first lane is X t(n+3) , the turn signal state of the first vehicle in the first lane is X t(n+4) , the throttle state of the first vehicle in the first lane is X t(n+5) , the gear state of the first vehicle in the first lane is X t(n+6) ; Construct the input feature vector corresponding to the t-th frame data, X t =[X t0 ,X t1 ,X t2 ,…,X tn ,X t(n+1) ,…,X t(n+6) ].

[0055] At 204, based on the input feature vector, an LSTM model is generated through training, verification, and testing, and the predicted path of the first vehicle is output. Operation 204 can be implemented by the server 140 and / or the control end 120 of the regional risk warning system 100 based on precise positioning and path prediction. In some embodiments, the server 140 and / or the control end 120 can generate an LSTM model based on the input feature vector through training, verification, and testing, and output the predicted path of the first vehicle. As an example, the server 140 and / or the control end 120 can obtain the turning operation of the first vehicle at the end of the first lane as h t0 The steering operations include left turn, right turn, straight ahead, and U-turn. Construct the output vector h corresponding to the t-th frame data t =[h t0 ]; Based on the historical vehicle operation data, construct the training set, validation set, and test set; construct the LSTM model:

[0056] f t =σ g (W f x t +U f h t-1 +b f )

[0057] i t =σ g (W i x t+U i h t-1 +b i )

[0058] o t =σ g (W o x t +U o h t-1 +b o )

[0059]

[0060]

[0061]

[0062] Among them, X t is the input feature vector, h t is the output vector, f t is the excitation vector used to control the forget gate, i t is the excitation vector used to control the input gate, o t is the excitation vector used to control the output gate, Input the excitation vector for the cell, c t is the cell state vector, and the three σ are activation functions, where: g is the sigmoid function, σ c is the hyperbolic tangent function, σ h Generally, it is the hyperbolic tangent function or σ h (x) = x, W, U, and b are weight parameters and offset parameters respectively.

[0063] According to some embodiments of the present application, the training, verification, and testing of generating an LSTM model specifically include determining an LSTM loss function; training the LSTM model based on the training set and the loss function, and determining the optimal weight parameters and offset parameters of W, U, and b; performing overfitting verification on the LSTM model based on the verification set; and testing the LSTM model based on the test set; if the test meets a third preset threshold, determining a predicted path of the first vehicle.

[0064] As an example, step 204 may specifically include an LSTM model building process 400, and the process 400 may construct a training set, a validation set, and a test set to generate an LSTM model based on historical vehicle operation data collected over a long period of time, as follows.

[0065] At 401, construct an LSTM input feature vector. As an example, select the sampling point of the vehicle in the current lane as x t0 to x t31(Assume that the lane is represented by 32 sampling points. In other embodiments, the number of sampling points may be 64, etc.) The average heading angle of the vehicle in the current lane is selected as x t32 , select the average speed of the vehicle in the current lane as x t33 , select the average steering wheel angle of the vehicle in the current lane as x t34 , select the vehicle's turn signal status in the current lane as x t35 , select the throttle state of the vehicle in the current lane as x t36 , select the gear state of the vehicle in the current lane as x t37 , construct the input feature vector x corresponding to the t-th frame data t =[x t0 ,x t1 ,x t2 ,…,x t35 ,x t36 ,x t37 ].

[0066] At 402, an LSTM output vector is constructed. As an example, the turning operation performed by the vehicle when it reaches the end of the current lane is selected as h t0 (turn left, turn right, go straight, turn around), construct the output vector h corresponding to the t-th frame data t =[h t0 ], where h t It can be the output of the current t-th frame of LSTM, and it can also be the feedback input of the t+1-th frame.

[0067] At 403 , a training set, a validation set, and a test set are constructed. As an example, the training set, the validation set, and the test set are constructed by using historical vehicle operation data collected over a long period of time.

[0068] At 404, an LSTM model is constructed. As an example:

[0069] f t =σ g (W f x t +U f h t-1 +b f )

[0070] i t =σ g (W i x t +U i h t-1 +b i )

[0071] o t =σ g (W o x t+U o h t-1 +b o )

[0072]

[0073]

[0074]

[0075] Among them, x t The input feature vector constructed for operation 401, h t The output vector constructed for operation 402, f t is the excitation vector used to control the forget gate inside LSTM, i t is the excitation vector used to control the input gate inside LSTM, o t is the excitation vector used to control the output gate inside LSTM, Input the excitation vector for the cell, c t is the cell state vector, and the three σ are activation functions, where: g is the sigmoid function, σ c is the hyperbolic tangent function, σ h Generally, it is the hyperbolic tangent function or σ h (x) = x, W, U, and b are weight parameters and offset parameters respectively.

[0076] At 405, the LSTM loss function is determined. As an example, according to the output data, it can be determined that the present application belongs to a machine learning multi-classification problem (left turn, right turn, straight, U-turn), and preferably, the present application selects a cross entropy loss function for the training of the LSTM model.

[0077] In 406, the LSTM model is trained and verified. As an example, according to the training set constructed in operation 403 and the loss function determined in operation 405, the BPTT algorithm is applied to train the LSTM model on a high-performance supercomputing server to obtain the optimal weight parameters and offset parameters of the W, U, and b parameters in the LSTM model constructed in operation 404; at the same time, during the training process, overfitting verification is performed according to the verification set constructed in operation 403.

[0078] In operation 407, the LSTM model is tested. As an example, the LSTM model that has completed parameter training and verification is tested according to the test set constructed in operation 403, and the test result is obtained.

[0079] At 408, if the test result meets the third preset threshold (for example, meets the actual application standard), the model algorithm is integrated into the system and device code of the present application to predict the turning operation performed by the vehicle when it reaches the end of the current lane, and the predicted turning operation is used to ultimately determine the predicted path information such as the selected downstream lane.

[0080] At 205, a traversal match is performed on the predicted path and the electronic fence information, and the matched first information is displayed. Operation 205 can be implemented by the server 140 and / or the control terminal 120 of the regional risk warning system 100 based on precise positioning and path prediction. In some embodiments, the server 140 and / or the control terminal 120 can perform a traversal match on the predicted path and the electronic fence information, and display the matched first information. In some embodiments, the first information includes irregular risk area information in the predicted path of the first vehicle, etc. As an example, the server 140 and / or the control terminal 120 can obtain a point trace sequence of the predicted path; traverse the point trace sequence, and perform a match with the information of the irregular risk area; if there is a match, the matched first information is displayed, and the first information includes the risk area and risk level of the predicted path, etc.

[0081] As an example, operation 205 may extract and traverse a point sequence from the predicted downstream lane, and match each traversed point with the electronic fence information (irregular risk area information) obtained through the 4G channel in operation 201. The matching may be implemented using a winding number algorithm.

[0082] According to some embodiments of the present application, when a successful match between an irregular risk area and a predicted path of a first vehicle is obtained, the first information of the match is output to an HMI (human-computer interface) to prompt the user in a visual form the possible risk areas in the future driving path and their corresponding risk levels.

[0083] In some embodiments, the matched first information can be displayed in the user interface (UI) of the user terminal 130, and the display scene can include but is not limited to scene display in any form or combination of VR, AR, MR, and XR.

[0084] It should be noted that the above description of process 200 is only for the convenience of description and does not limit the present application to the scope of the embodiments. It is understandable that for those skilled in the art, based on the principle of the present system, it is possible to arbitrarily combine the various operations, or constitute a sub-process and other operation combination without deviating from the principle, and make various corrections and changes in form and details to the functions of implementing the above-mentioned processes and operations. For example, process 200 may further include outputting the matched first information to the HMI (human-computer interaction interface), and prompting the user in a visual form the risk areas and their corresponding risk levels of possible routes in the future driving path. Such variations are within the scope of protection of the present application.

[0085] To summarize, according to the regional risk warning method and system based on precise positioning and path prediction of the embodiment of the present application, based on the high-precision positioning and path prediction results of the vehicle's location, the program automatically completes the retrieval and matching of the risk area determined by the temporary fence and can automatically notify the driver, thereby improving the timeliness and effectiveness of information transmission.

[0086] It should be noted that the above-mentioned embodiments are merely used as examples, and the present application is not limited to such examples, but various changes can be made.

[0087] It should be noted that, in this specification, the terms "include", "comprises" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or device including the element.

[0088] Finally, it should be noted that the above series of processing includes not only processing executed in time series in the order described here, but also processing executed in parallel or separately rather than in time series.

[0089] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by hardware related to computer program instructions, and the program can be stored in a computer-readable storage medium, which, when executed, can include the processes in the embodiments of the above-mentioned methods. The storage medium can be a disk, an optical disk, a read-only memory (ROM) or a random access memory (RAM), etc.

[0090] The above disclosures are only some preferred embodiments of the present application, which cannot be used to limit the scope of rights of the present application. Ordinary technicians in this field can understand that all or part of the processes of implementing the above embodiments and equivalent changes made according to the claims of the present application are still within the scope of the invention.

Claims

1. A regional risk warning method based on precise positioning and path prediction, characterized in that: include: Acquire map data and electronic fence information of the first area, wherein the map data includes lane information; Acquire positioning information and state information of the first vehicle, wherein the positioning information and the state information are used to construct an input feature vector; Performing lane matching according to the positioning information and the lane information to determine a first lane of the first vehicle; Based on the input feature vector, generate an LSTM model through training, verification, and testing, and output a predicted path of the first vehicle; Performing traversal matching on the predicted path and the electronic fence information, and displaying matching first information; The obtaining of the map data and electronic fence information of the first area specifically includes: Acquire map data of the first area sent by the roadside equipment through the V2X channel; Acquire electronic fence information of the first area through a 4G channel, where the electronic fence information includes information of an irregular risk area; The obtaining of the positioning information and status information of the first vehicle specifically includes: Acquire positioning information through the GNSS module, the positioning information including high-precision positioning, heading angle, and vehicle speed information; Acquire status information through the vehicle body CAN bus, the status information including steering wheel angle, turn signal status, throttle status, and gear status information; The positioning information and the state information are used to construct an input feature vector, which specifically includes: Get the sampling point of the first vehicle in the first lane as X t0 To X tn ; The average heading angle of the first vehicle in the first lane is obtained as X t(n+1) , the average speed of the first vehicle in the first lane is X t(n+2) , the average steering wheel angle of the first vehicle in the first lane is X t(n+3) , the turn signal state of the first vehicle in the first lane is X t(n+4) , the throttle state of the first vehicle in the first lane is X t(n+5) , the gear state of the first vehicle in the first lane is X t(n+6) ; Construct the input feature vector corresponding to the t-th frame data, Xt = [X t0 ,X t1 ,X t2 ,…,X tn ,X t(n+1) ,…,X t(n+6) ]; The generating of the LSTM model based on the input feature vector through training, verification, and testing, and outputting the predicted path of the first vehicle specifically includes: Obtain the turning operation of the first vehicle at the end of the first lane as h t0 The steering operations include left turn, right turn, straight ahead, and U-turn. Construct the output vector h corresponding to the t-th frame data t =[h t0 ]; Based on historical vehicle operation data, construct training sets, validation sets, and test sets; Construct LSTM model: f t =σ g (W f x t +U f h t-1 +b f ) i t =σ g (W i x t +U i h t-1 +b i ) the t =s g (W o x t +U o h t-1 +b o ) h t =o t oσ h (c t ) Among them, X t is the input feature vector, h t is the output vector, f t is the excitation vector used to control the forget gate, i t is the excitation vector used to control the input gate, o t is the excitation vector used to control the output gate, Input the excitation vector for the cell, c t is the cell state vector, σ is the activation function, where: σ g is the sigmoid function, σ c is the hyperbolic tangent function, σ h is the hyperbolic tangent function or σ h (x) = x, W, U, and b are weight parameters and offset parameters respectively.

2. The method according to claim 1, characterized in that The performing lane matching according to the positioning information and the lane information to determine the lane of the first vehicle specifically includes: Traversing the first node, the first child node, and the first adjacent point pair of the first child node of the map data; For each first adjacent point pair, determining a first distance, a first mark, and a first angle between the first vehicle and the first adjacent point pair; If the first distance, the first mark, and the first angle do not satisfy a first preset threshold, return to traversing the first adjacent point pair.

3. The method according to claim 2, characterized in that If the first distance, the mark, and the angle meet a first preset threshold, specifically including: Traversing the second child node associated with the first child node and the second adjacent point pair of the second child node; For each second adjacent point pair, determining a second distance, a second mark, and a second angle between the first vehicle and the second adjacent point pair; If the second distance, the second mark, and the second angle do not meet a second preset threshold, return to traverse the second adjacent point pair; If the second distance, the second sign, and the second angle meet a second preset threshold, the lane matched by the first vehicle is determined.

4. The method according to claim 1, characterized in that Specifically include: Determine the LSTM loss function; Based on the training set and the loss function, train the LSTM model to determine the optimal weight parameters and offset parameters of W, U, and b; Based on the verification set, performing overfitting verification on the LSTM model; Based on the test set, testing the LSTM model; If the test satisfies a third preset threshold, a predicted path of the first vehicle is determined.

5. The method according to claim 4, characterized in that Specifically include: Obtaining a point trace sequence of the predicted path; Traversing the point trace sequence and performing matching with the information of the irregular risk area; If there is a match, the matching first information is displayed, where the first information includes the risk area and risk level of the predicted path.

6. A system, characterized in that, include: a memory configured to store data and instructions; A processor in communication with a memory, wherein when executing instructions in the memory, the processor is configured to: Acquire map data and electronic fence information of the first area, wherein the map data includes lane information; Acquire positioning information and state information of the first vehicle, wherein the positioning information and the state information are used to construct an input feature vector; Performing lane matching according to the positioning information and the lane information to determine a first lane of the first vehicle; Based on the input feature vector, generate an LSTM model through training, verification, and testing, and output a predicted path of the first vehicle; Performing traversal matching on the predicted path and the electronic fence information, and displaying matching first information; The obtaining of the map data and electronic fence information of the first area specifically includes: Acquire map data of the first area sent by the roadside equipment through the V2X channel; Acquire electronic fence information of the first area through a 4G channel, where the electronic fence information includes information of an irregular risk area; The obtaining of the positioning information and the state information of the first vehicle specifically includes: Acquire positioning information through the GNSS module, the positioning information including high-precision positioning, heading angle, and vehicle speed information; Acquire status information through the vehicle body CAN bus, the status information including steering wheel angle, turn signal status, throttle status, and gear status information; The positioning information and the state information are used to construct an input feature vector, which specifically includes: Get the sampling point of the first vehicle in the first lane as X t0 To X tn ; The average heading angle of the first vehicle in the first lane is obtained as X t(n+1) , the average speed of the first vehicle in the first lane is X t(n+2) , the average steering wheel angle of the first vehicle in the first lane is X t(n+3) , the turn signal state of the first vehicle in the first lane is X t(n+4) , the throttle state of the first vehicle in the first lane is X t(n+5) , the gear state of the first vehicle in the first lane is X t(n+6) ; Construct the input feature vector corresponding to the t-th frame data, Xt = [X t0 ,X t1 ,X t2 ,…,X tn ,X t(n+1) ,…,X t(n+6) ]; The generating of the LSTM model based on the input feature vector through training, verification, and testing, and outputting the predicted path of the first vehicle specifically includes: Obtain the turning operation of the first vehicle at the end of the first lane as h t0 The steering operations include left turn, right turn, straight ahead, and U-turn. Construct the output vector h corresponding to the t-th frame data t =[h t0 ]; Based on historical vehicle operation data, construct training sets, validation sets, and test sets; Construct LSTM model: f t =σ g (W f x t +U f h t-1 +b f ) i t =σ g (W i x t +U i h t-1 +b i ) the t =s g (W o x t +U o h t-1 +b o ) h t =o t oσ h (c t ) Among them, X t is the input feature vector, h t is the output vector, f t is the excitation vector used to control the forget gate, i t is the excitation vector used to control the input gate, o t is the excitation vector used to control the output gate, Input the excitation vector for the cell, c t is the cell state vector, σ is the activation function, where: σ g is the sigmoid function, σ c is the hyperbolic tangent function, σ h is the hyperbolic tangent function or σ h (x) = x, W, U, and b are weight parameters and offset parameters respectively.

Images