WO2021146910A1 - 基于图计算的车辆驾驶数据处理方法、装置和计算机设备 - Google Patents

基于图计算的车辆驾驶数据处理方法、装置和计算机设备 Download PDF

Info

Publication number
WO2021146910A1
WO2021146910A1 PCT/CN2020/073492 CN2020073492W WO2021146910A1 WO 2021146910 A1 WO2021146910 A1 WO 2021146910A1 CN 2020073492 W CN2020073492 W CN 2020073492W WO 2021146910 A1 WO2021146910 A1 WO 2021146910A1
Authority
WO
WIPO (PCT)
Prior art keywords
data
node
processed
index data
sequence
Prior art date
Application number
PCT/CN2020/073492
Other languages
English (en)
French (fr)
Inventor
方闽杰
Original Assignee
深圳元戎启行科技有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 深圳元戎启行科技有限公司 filed Critical 深圳元戎启行科技有限公司
Priority to PCT/CN2020/073492 priority Critical patent/WO2021146910A1/zh
Priority to CN202080003165.8A priority patent/CN113677583B/zh
Publication of WO2021146910A1 publication Critical patent/WO2021146910A1/zh

Links

Images

Classifications

    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B60—VEHICLES IN GENERAL
    • B60W—CONJOINT CONTROL OF VEHICLE SUB-UNITS OF DIFFERENT TYPE OR DIFFERENT FUNCTION; CONTROL SYSTEMS SPECIALLY ADAPTED FOR HYBRID VEHICLES; ROAD VEHICLE DRIVE CONTROL SYSTEMS FOR PURPOSES NOT RELATED TO THE CONTROL OF A PARTICULAR SUB-UNIT
    • B60W50/00—Details of control systems for road vehicle drive control not related to the control of a particular sub-unit, e.g. process diagnostic or vehicle driver interfaces

Definitions

  • This application relates to a method, device and computer equipment for processing vehicle driving data based on graph calculation.
  • a graph is composed of a set of vertices and a set of edges that can connect two vertices, and is a data structure.
  • DAG Directed Acyclic Graph
  • a graph is composed of a set of vertices and a set of edges that can connect two vertices, and is a data structure.
  • DAG Directed Acyclic Graph
  • a method, device, and computer device for processing vehicle driving data based on graph calculation are provided.
  • a method for processing vehicle driving data based on graph calculation includes:
  • graph calculation is performed according to the node sequence and the node information through the data analysis model to obtain multiple sub-graphs and corresponding sub-graph values;
  • a vehicle driving data processing device based on graph calculation includes:
  • a data acquisition module for acquiring vehicle driving data, and extracting index data to be processed in the vehicle driving data
  • a data storage module configured to input the to-be-processed index data into a data analysis model, and store the to-be-processed index data according to the data type of the to-be-processed index data;
  • a data verification module for extracting the node sequence and node information corresponding to the index data to be processed, and verifying the index data to be processed according to the node sequence and the node information;
  • the calculation processing module is used to perform graph calculations based on the node sequence and the node information through the data analysis model after the verification is passed, to obtain multiple sub-graphs and corresponding sub-graph values; continue according to the nodes
  • the sequence performs iterative calculations on multiple subgraphs and corresponding subgraph values until the target index value corresponding to the index data to be processed is output.
  • a computer device includes a memory and a processor, the memory stores a computer program, and the processor implements the steps of the remote takeover-based vehicle control method provided in any one of the embodiments of the present application when the computer program is executed.
  • One or more non-volatile computer-readable storage media storing computer-readable instructions.
  • the computer-readable instructions are executed by one or more processors, one or more processors execute the readable storage medium to realize the present invention. Apply for the steps of the vehicle control method based on remote takeover provided in any of the embodiments.
  • Fig. 1 is an application scenario diagram of a method for processing vehicle driving data based on graph calculation according to one or more embodiments.
  • Fig. 2 is a schematic flowchart of a method for processing vehicle driving data based on graph calculation in one or more embodiments.
  • Fig. 3 is a schematic flowchart of a graph calculation step performed by a data analysis model according to one or more embodiments.
  • Fig. 4 is a schematic flow diagram of a graph calculation step performed by a data analysis model in another embodiment.
  • Fig. 5 is a schematic flowchart of multi-tasking processing steps according to one or more embodiments.
  • Fig. 6 is a block diagram of a device for processing vehicle driving data based on graph calculation according to one or more embodiments.
  • Figure 7 is a block diagram of a computer device according to one or more embodiments.
  • the method for processing vehicle driving data based on graph calculation provided in this application can be applied to a variety of application environments. For example, it can be applied to the application environment of automatic driving as shown in FIG. 1.
  • a sensor 102 and a processor 104 are included in the vehicle.
  • the sensor 102 can communicate with the processor 104 via a network.
  • the sensor 102 can collect vehicle driving data of the vehicle, the processor 104 extracts the to-be-processed index data from the vehicle driving data and then inputs the to-be-processed index data into the data analysis model, and stores the to-be-processed index data according to the data type of the to-be-processed index data; Extract the node sequence and node information corresponding to the index data to be processed, and verify the index data to be processed according to the node sequence and node information; when the verification is passed, the processor 104 uses the data analysis model according to the node sequence and node information Perform graph calculation to obtain multiple subgraphs and corresponding subgraph values; continue to iteratively calculate multiple subgraphs and corresponding subgraph values according to the node sequence until the target index value corresponding to the index data to be processed is output.
  • the sensor 102 may be a variety of sensors carried by the automatic driving device, and may specifically include a laser radar, a laser scanner, and the like.
  • a method for processing vehicle driving data based on graph calculation is provided. Taking the method applied to the vehicle in FIG. 1 as an example for description, the method includes the following steps:
  • Step 202 Obtain vehicle driving data, and extract index data to be processed in the vehicle driving data.
  • the vehicle can be an unmanned car, which is a smart car that senses the road environment through the on-board sensor system, automatically plans the driving route, and controls the vehicle to reach the predetermined goal.
  • the vehicle records the vehicle driving data in real time during the automatic driving process.
  • the vehicle driving data may include road image information and vehicle status information.
  • the road image information may include collected video data and road screen video data.
  • the road image information may be continuous video frames.
  • the vehicle status information may include vehicle status information such as vehicle positioning information, vehicle navigation information, in-vehicle temperature information, and vehicle instrument information. Transmission of road image information and vehicle status information is used for remote takeover personnel to remotely take over the vehicle according to road conditions and vehicle status.
  • the vehicle After the vehicle obtains the vehicle driving data, it performs index feature extraction on the vehicle driving data, thereby extracting multiple index data to be processed.
  • the index data to be processed may include data corresponding to multiple indexes such as vehicle speed, mileage, and fuel consumption, and the index data to be processed may also be index data corresponding to each frame of road image.
  • the index data to be processed may include index data to be processed corresponding to multiple task types.
  • the task type can include multiple types such as traffic rule recognition and vehicle risk assessment.
  • Step 204 Input the index data to be processed into the data analysis model, and store the index data to be processed according to the data type of the index data to be processed.
  • the data analysis model can be a DAG-based calculation model of a directed acyclic graph.
  • the directed acyclic graph is constructed in advance using historical sample data according to a preset algorithm.
  • the directed acyclic graph model includes multiple pre-built Node structure.
  • the vehicle After the vehicle extracts multiple indicator data to be processed, it inputs the multiple indicator data to be processed into the preset data analysis model. Specifically, the multiple indicator data to be processed can be filled into a directed acyclic ring according to the corresponding indicator type. In the corresponding node structure in the graph model.
  • the data analysis model can identify the data types of multiple to-be-processed indicator data, determine the storage structure of multiple to-be-processed indicator data based on the data type, and store the multiple to-be-processed data in the allocated storage structure.
  • the storage structure may be a storage structure in a hard disk, which can effectively perform persistent storage of the index data to be processed, so as to ensure the consistency and integrity of the index data to be processed in the processing process.
  • the data analysis model can further identify the operation type of the index data to be processed, and then determine the operation structure of the index data to be processed according to the operation type and data type, so as to cache the operation data generated during the processing of the node information.
  • Step 206 Extract the node sequence and node information corresponding to the index data to be processed, and verify the index data to be processed according to the node sequence and node information.
  • the data analysis model After the vehicle inputs the index data to be processed into the data analysis model and stores it, the data analysis model further extracts the node sequence corresponding to multiple index data to be processed and the node information of each node, and according to the node sequence and the node The information is verified by the index data to be processed.
  • the data analysis model can extract the node sequence, the number of nodes, and also obtain the node information of each node.
  • the node information can include information such as data type, variable scope, and the number of child nodes of the node. .
  • the data analysis model detects and verifies the validity of the node sequence and node information, such as detecting the wrong number of sub-nodes of the node, the wrong data type, and the error of the variable scope. Only when the node sequence and node information meet the legal and valid conditions, it means that the data to be processed has passed the verification.
  • the node sequence may be a topological sequence obtained by topologically sorting multiple nodes.
  • the vertices in the graph represent activities
  • the directed edges in the graph represent the sequence of activities. That is, the activities at the starting point of the directed edges are the predecessor activities of the ending activities, and only after the starting activities are completed, The end activity can only be carried out.
  • a topological sequence can be constructed for all nodes in a directed acyclic graph through the active network of vertices, a vertex with zero in-degree is selected and output, and this vertex and all outgoing edges are deleted from the network of the directed acyclic graph.
  • Step 208 After the verification is passed, graph calculation is performed according to the node sequence and node information through the data analysis model to obtain multiple sub-graphs and corresponding sub-graph values.
  • the vehicle uses the data analysis model to perform graph calculations on multiple nodes in the data analysis model according to the node sequence and node information. Specifically, the data analysis model calculates the subgraph corresponding to the current node according to a preset logic algorithm, and evaluates the subgraph. The data analysis model can also use the obtained value to update the subgraph to obtain the updated simplified current subgraph and the corresponding subgraph value.
  • Step 210 Continue to iteratively calculate multiple sub-graphs and corresponding sub-graph values according to the node sequence until the target index value corresponding to the index data to be processed is output.
  • the data analysis model then continues to iteratively calculate multiple sub-graphs and corresponding sub-graph values according to the node sequence. In each round, the corresponding updated sub-graphs and corresponding sub-graph values are obtained, and then based on the updated sub-graphs and sub-graph values. Perform graph calculation on the next node information according to the node sequence until all node calculations are completed, and then output the calculated result through the data analysis model, which is the target index value corresponding to the index data to be processed.
  • the vehicle may use the data analysis model to perform graph calculations on the index data to be processed, and obtain the result values corresponding to multiple target indexes such as the vehicle traffic index value and the vehicle risk index value.
  • the data analysis model is a model based on graph calculation, by using the data analysis model to calculate and process a large amount of vehicle driving data generated by the vehicle, the calculation ability of the data analysis model can be effectively improved, so that the vehicle driving data can be calculated and processed quickly and accurately. , And then calculate the target index value with higher accuracy.
  • the vehicle obtains the vehicle driving data and extracts the index data to be processed in the vehicle driving data, and then inputs the index data to be processed into the data analysis model.
  • the node sequence and node information corresponding to the index data to be processed are extracted through the data analysis model, and the index data to be processed is verified according to the node sequence and node information to ensure the validity of the input data.
  • graph calculation is performed based on the node sequence and node information through the data analysis model. Since the data analysis model is a model based on graph calculation, multiple subgraphs and corresponding subgraph values can be effectively calculated.
  • the data analysis model continues to iteratively calculate multiple subgraphs and corresponding subgraph values according to the node sequence until the target index value corresponding to the index data to be processed is output.
  • storing the index data to be processed according to the data type of the index data to be processed includes: determining the storage structure of the index data to be processed according to the data type of the index data to be processed, and storing the index data to be processed according to the storage structure;
  • the operating structure corresponding to the index data to be processed is determined according to the node type of the node information; the operating structure is used to cache the operating data generated during the processing of the node information.
  • index data to be processed needs to be stored persistently to ensure the consistency of the data during transmission and calculation. Sex and effectiveness.
  • the vehicle determines the data type of the index data to be processed and the corresponding multiple node information through the data analysis model.
  • the data analysis model can include multiple pre-built nodes, and the vehicle can directly transfer the index data to be processed. Fill in the corresponding node.
  • the vehicle determines the storage structure of the to-be-processed index data according to the data type of the to-be-processed index data, and then stores the to-be-processed index data according to the storage structure.
  • ProtoBuf Protocol Buffers, scalable serialized structured data method
  • ProtoBuf is a language-independent, platform-independent, and extensible serialized data structure. It can be used for data communication protocols and Data storage, etc.
  • data types such as Graph, Node, and Operation can be defined.
  • Graph can include the node field of the repeated attribute and the field used to describe the ID of the root node.
  • Operation is an enumerated type that can include all arithmetic operations of the data analysis model.
  • the Node type can include: an ID of any data type, used to identify the node; a field of Operation type, used to describe the operation type corresponding to the node; two repeated attributes of the same field as the ID data type, used respectively List the predecessor and successor nodes of the current node. Fields required for operations corresponding to all enumeration values of the Operation class. For example, the UNIVERSAL operation requires one field to store the name of the variable declared by the universal quantifier, and another field to describe the variable defined by the universal quantifier Ranges.
  • the vehicle After the vehicle stores the index data to be processed according to the storage structure, it can further determine the operation structure corresponding to the index data to be processed according to the node type of the node information.
  • the operation structure is used to cache the operation data generated during the processing of the node information. .
  • the graph computing model and all the nodes it contains can be encapsulated as instances of custom types, so that the nodes can be easily read and operated to complete verification , Calculation processing and other operations.
  • an operational structure with strong logic, versatility, and scalability can be obtained. It can manage various types of nodes conveniently while effectively performing specific operations on different computing nodes, thereby effectively improving the Processing efficiency of processing indicator data.
  • Graph type when implementing the running structure corresponding to the index data to be processed, you can first define a Graph type, and each instance of this type corresponds to a DAG directed acyclic graph.
  • Graph type should provide a means to quickly find nodes based on ID.
  • a corresponding Node type to inherit NodeBase to represent the node belonging to the operation.
  • each virtual function can be provided according to the actual semantics of the operation.
  • these types can also be defined as template classes.
  • a node type that represents a multiplication operation can be used in an imaginary number operation, or it can represent a matrix multiplication.
  • the Protobuf library can be used to implement the conversion between the Protobuf file and the above data types.
  • the above method can be implemented using the C++ language.
  • verifying the index data to be processed according to the node sequence and node information includes: extracting multiple node feature parameters of the index data to be processed according to the node sequence and node information; When the characteristic parameters all meet the preset condition threshold, it is determined that the verification of the index data to be processed has passed.
  • the data analysis model After the vehicle inputs the index data to be processed into the data analysis model and stores it, the data analysis model further extracts the node sequence corresponding to multiple index data to be processed and the node information of each node, and according to the node sequence and the node The information is verified by the index data to be processed.
  • the data analysis model can extract multiple node feature parameters of the index data to be processed according to the node sequence and node information.
  • the node characteristic parameters can include multiple parameter fields such as whether there is a loop, the in-degree of the root node, the free node, the scope, the data type, and the succession relationship, and the corresponding parameter values.
  • the data analysis model then verifies the validity of multiple node feature parameters, and detects whether each node feature parameter is legal. For example, when there is a loop in a directed acyclic graph, the in-degree of the root node is 1, there is a free node, there is a scope error or a variable name is wrong, there is a data type error or a succession relationship error, etc. In this case, it means that the verification of the index data to be processed has not passed and further evaluation calculations cannot be performed. Only when all the node characteristic parameters meet the legal and valid conditions, it means that the data to be processed has passed the verification. By verifying the validity of the index data to be processed, the processing efficiency of the index data to be processed and the accuracy of the processing results can be effectively guaranteed.
  • the method further includes: when the verification fails, generating prompt information according to the characteristic parameters of the node that has not passed the verification; and sending the prompt information to the corresponding monitoring terminal.
  • Each vehicle can be bound with a corresponding monitoring terminal to monitor the vehicle when there is a fault or abnormal situation in the vehicle.
  • the vehicle In the process of verifying the validity of multiple node feature parameters by the data analysis model, when there is a node feature parameter with an error condition, it means that the verification of the index data to be processed has not passed.
  • the vehicle then generates corresponding prompt information according to the characteristic parameters of the node that has not passed the verification, and sends the prompt information to the corresponding monitoring terminal, so that the monitoring terminal monitors the driving condition of the vehicle according to the prompt information, thereby effectively ensuring the normal operation of the vehicle.
  • the node sequence is a topological sequence
  • the steps of graph calculation based on the node sequence and node information through the data analysis model specifically include the following:
  • Step 302 Extract the node knowledge representation of the node according to the node type and node information.
  • Step 304 Perform graph calculation on the initial layer nodes according to the topology sequence and the node knowledge representation to obtain the corresponding first subgraph and first subgraph values.
  • Step 306 Perform graph calculation on the nodes of the next layer according to the topological sequence and the corresponding node knowledge representation according to the first subgraph and the first subgraph values to obtain the corresponding second subgraph and second subgraph values.
  • Step 308 Continue to perform iterative graph calculations on multiple nodes according to the topology sequence and corresponding node knowledge representations according to the obtained current subgraph and current subgraph values.
  • step 310 until all nodes have completed graph calculation, the target index value corresponding to the index data to be processed is output.
  • Knowledge representation refers to a symbolic representation of all relevant aspects involved in a certain field.
  • Knowledge representation can include several representations such as first-order predicate logic representation, production rules, frame representation, script representation, semantic web representation, and knowledge graph representation.
  • the node knowledge representation can be the knowledge representation between each node and the node.
  • the vehicle inputs the index data to be processed into the data analysis model, and after the index data to be processed is verified, the data analysis model further performs graph calculation processing on multiple nodes corresponding to the index data to be processed, and iteratively calculates the index data to be processed The corresponding target index value.
  • the vehicle can extract the node knowledge representation of the node according to the node type and node information through the data analysis model.
  • the node knowledge representation of the node includes the knowledge representation of each node and the knowledge representation between nodes.
  • the data analysis model then performs graph calculation on the initial layer nodes according to the topology sequence and node knowledge representation, and obtains the corresponding first subgraph and the first subgraph value.
  • the data analysis model continues to perform graph calculations on the nodes of the next layer according to the topological sequence and the corresponding node knowledge representation according to the first subgraph and the first subgraph values to obtain the corresponding second subgraph and second subgraph values.
  • the data analysis model can use the obtained value to update the subgraph to obtain the updated simplified current second subgraph and the corresponding second subgraph value.
  • the data analysis model continues to perform iterative graph calculations on multiple nodes according to the obtained current subgraph and current subgraph values according to the topology sequence and the corresponding node knowledge representation.
  • the corresponding updated sub-graph and the corresponding sub-graph value are obtained, and then the next node information is calculated according to the node sequence according to the updated sub-graph and sub-graph value, until all the nodes have completed the graph calculation .
  • the calculated result is output through the data analysis model, and the result is the target index value corresponding to the index data to be processed.
  • the vehicle may use the data analysis model to perform graph calculations on the index data to be processed, and obtain the result values corresponding to multiple target indexes such as the vehicle traffic index value and the vehicle risk index value.
  • the data analysis model is a model based on graph calculation, by using the data analysis model to calculate and process a large amount of vehicle driving data generated by the vehicle, the calculation ability of the data analysis model can be effectively improved, so that the vehicle driving data can be calculated and processed quickly and accurately. , And then calculate the target index value with higher accuracy.
  • the step of performing graph calculation on the initial layer nodes according to the topology sequence and the node knowledge representation to obtain the corresponding first subgraph and the first subgraph value specifically includes the following content :
  • Step 402 Extract the first-order predicate logic representation of the node according to the node type and node information.
  • Step 404 using a preset predicate logic algorithm to perform graph calculation on the initial layer nodes according to the topology sequence and the node knowledge representation, and extract the current subgraph corresponding to the largest target node.
  • Step 406 Calculate the sub-picture value corresponding to the current sub-picture, update the current sub-picture with the sub-picture value, and generate the first sub-picture and the first sub-picture value corresponding to the initial layer node.
  • Knowledge data can generally be represented by a sentence or a few sentences with complete meaning.
  • This knowledge can be represented by predicate logic, for example, a predicate formula, that is, a formula formed by connecting some predicates with predicate connection symbols.
  • Predicate formulas can express factual knowledge such as the state, attributes, and concepts of things, as well as regular knowledge that determines causal relationships between things.
  • First-order predicate logic is a form of formal language representation. It is a knowledge representation composed of propositions, logical connectives, word styles, predicates and quantifiers. This knowledge representation is more accurate in data processing.
  • the vehicle inputs the index data to be processed into the data analysis model, and after the index data to be processed is verified, the data analysis model further performs graph calculation processing on multiple nodes corresponding to the index data to be processed, and iteratively calculates the index data to be processed The corresponding target index value.
  • the vehicle first uses the data analysis model to extract the node knowledge representation of the node according to the node type and node information.
  • the node knowledge representation can be a first-order predicate logic representation, that is, the data analysis model can be based on the node type and the node.
  • Point information extracts the first-order predicate logic representation of the node.
  • the first-order predicate logic representation of a node includes the first-order predicate logic representation of each node and the first-order predicate logic representation between nodes.
  • the data analysis model uses the preset predicate logic algorithm to perform graph calculations on the initial layer nodes according to the topology sequence and the node knowledge representation, and extract the current subgraph corresponding to the largest target node. Calculate the sub-graph value corresponding to the current sub-graph, update the current sub-graph with the sub-graph value, and generate the first sub-graph and the first sub-graph value corresponding to the initial layer node.
  • iterative graph calculation for the directed acyclic graph model it is necessary to find the largest subgraph in the current directed acyclic graph that contains only logical operation nodes, evaluate the subgraph, and then use the result Replace the subgraph with the value of to get a simplified subgraph.
  • the vehicle driving data can be calculated and processed quickly and accurately, so that the target index value with higher accuracy can be effectively obtained.
  • the method before inputting the index data to be processed into the data analysis model, the method further includes: acquiring historical vehicle driving data, and performing feature extraction on the historical vehicle driving data to obtain multiple feature vectors; Vectors perform characteristic analysis to obtain multiple characteristic indicators; construct multiple node structures according to multiple characteristic indicators; use preset algorithms to construct data analysis models based on multiple node structures.
  • the vehicle Before the vehicle performs calculation processing on the generated vehicle driving data, it needs to construct a data analysis model of the corresponding task type in advance. Specifically, the vehicle may obtain several historical vehicle driving data. These historical vehicle driving data may be historical data generated by the vehicle itself, or historical data generated by multiple vehicles on the vehicle platform.
  • the vehicle then performs feature extraction on historical vehicle driving data.
  • a cluster analysis algorithm can be used to perform feature clustering and feature extraction on several historical vehicle driving data, so as to obtain multiple feature vectors.
  • the vehicle further performs characteristic analysis on the extracted multiple feature vectors to obtain multiple feature indicators.
  • the vehicle can extract the characteristic index corresponding to each task type according to different task types. Therefore, multiple node structures can be constructed according to multiple characteristic indicators, and a data analysis model based on a directed acyclic graph can be effectively constructed using a preset algorithm according to multiple node structures.
  • the constructed data analysis model may include directed acyclic graph models corresponding to multiple task types.
  • the data analysis model includes multiple subtask models, and the method further includes a multitasking step, which specifically includes the following content:
  • step 502 the vehicle driving data is decomposed into multiple subtask data according to the multiple task types, and to-be-processed index data corresponding to the multiple sub-task data are extracted respectively; the to-be-processed indicator data includes the task type.
  • Step 504 Input the index data to be processed into the data analysis model, and input multiple index data to be processed into the corresponding subtask model according to the task type.
  • Step 506 Extract the node sequence and node information corresponding to the index data to be processed through the subtask model, and perform graph calculation according to the node sequence and node information to obtain the target index value corresponding to the subtask data.
  • the data analysis model can include the directed acyclic graph model corresponding to multiple task types, and the vehicle can analyze and process the vehicle driving data of multiple task types at the same time through the data analysis model.
  • the task type can include traffic rule recognition, Vehicle risk assessment and many other types.
  • the vehicle After the vehicle obtains the vehicle driving data, it decomposes the vehicle driving data into multiple subtask data according to multiple task types, and then performs index feature extraction on the multiple subtask data, and extracts the to-be-processed index data corresponding to the multiple subtask data respectively;
  • the processing index data includes the corresponding task type.
  • the vehicle After the vehicle extracts the to-be-processed indicator data corresponding to multiple task types, it inputs the multiple to-be-processed indicator data into a preset data analysis model. Specifically, the data analysis model inputs multiple to-be-processed indicator data according to the task type. To the corresponding subtask model, and fill multiple to-be-processed index data into the corresponding node structure in the directed acyclic graph model according to the corresponding index type.
  • Each subtask model stores and processes the input indicator data to be processed at the same time, extracts the node sequence corresponding to multiple indicator data to be processed and the node information of each node, and treats it according to the node sequence and node information Processing indicator data for verification. After the verification is passed, the subtask model performs graph calculation according to the node sequence and node information, and obtains multiple subgraphs corresponding to the subtask data and corresponding subgraph values. Continue to iteratively calculate multiple sub-graphs and corresponding sub-graph values according to the node sequence until the target index value corresponding to the sub-task data is output.
  • the data analysis model including multiple subtask models simultaneously processes the vehicle driving data of multiple tasks, thereby effectively calculating and processing the vehicle driving data according to different task types, effectively improving the data processing efficiency.
  • the method further includes: determining the current driving traffic state value of the vehicle according to the target index value corresponding to the index data to be processed; when the driving traffic state value indicates a violation, generating violation prompt information according to the driving traffic state value, And send the violation prompt information to the corresponding monitoring terminal.
  • the vehicle After the vehicle obtains the vehicle driving data, it extracts the vehicle driving data corresponding to the multiple subtask data according to the multiple task types to be processed index data, inputs the multiple to be processed index data into the preset data analysis model, and passes each subtask
  • the model simultaneously stores and processes the input indicator data to be processed, and obtains multiple sub-graphs and corresponding sub-graph values corresponding to the sub-task data. Continue to iteratively calculate multiple sub-graphs and corresponding sub-graph values according to the node sequence until the target index value corresponding to the sub-task data is output.
  • the vehicle can determine the current driving traffic state value of the vehicle according to the target index value corresponding to the subtask data.
  • the driving traffic state value corresponding to the target index value can be determined according to a preset index mapping table.
  • the driving traffic state value indicates a violation, it means that the vehicle currently violates traffic rules.
  • the vehicle generates violation prompt information according to the driving traffic state value, and sends the violation prompt information to the corresponding monitoring terminal, so that the monitoring terminal can monitor the vehicle in time And management.
  • the method further includes: calculating the driving risk score of the vehicle according to the target index value corresponding to the index data to be processed; when the driving risk score exceeds the risk threshold, sending an early warning message to the monitoring terminal.
  • the vehicle can determine the current driving risk score of the vehicle according to the target index value corresponding to the subtask data.
  • the current driving risk score of the vehicle can be calculated according to the preset index target value and the corresponding weight.
  • it can include the index value of multiple dimensions such as vehicle equipment risk, road obstacle risk, and external risk, so as to effectively calculate the vehicle's current driving risk score.
  • Driving risk score during driving. When the driving risk score exceeds the risk threshold, it indicates that the vehicle currently has risk factors, and the vehicle generates early warning information according to the corresponding risk indicators, and sends the early warning information to the corresponding monitoring terminal, so that the monitoring terminal can monitor and control the vehicle in time. manage.
  • a vehicle driving data processing device based on graph calculation including: a data acquisition module 602, a data storage module 604, a data verification module 606, and a calculation processing module 608, wherein:
  • the data acquisition module 602 is used to acquire vehicle driving data and extract index data to be processed in the vehicle driving data;
  • the data storage module 604 is used to input the index data to be processed into the data analysis model, and store the index data to be processed according to the data type of the index data to be processed;
  • the data verification module 606 is used to extract the node sequence and node information corresponding to the index data to be processed, and verify the index data to be processed according to the node sequence and node information;
  • the calculation processing module 608 is used to perform graph calculations based on the node sequence and node information through the data analysis model after the verification is passed to obtain multiple sub-graphs and corresponding sub-graph values; continue to perform calculations on the multiple sub-graphs and corresponding sub-graph values according to the node sequence Iterative calculation is performed on the sub-picture value of, until the target index value corresponding to the index data to be processed is output.
  • the data storage module 604 is further configured to determine the storage structure of the index data to be processed according to the data type of the index data to be processed, and store the index data to be processed according to the storage structure; determine the node type according to the node information
  • the running structure corresponding to the index data to be processed; the running structure is used to cache the running data generated in the process of processing the node information.
  • the data verification module 606 is further configured to extract multiple node feature parameters of the index data to be processed according to the node sequence and node information; when the multiple node feature parameters all meet the preset condition threshold, Confirm that the indicator data to be processed has passed the verification.
  • the data verification module 606 is further configured to generate prompt information according to the characteristic parameters of the node that has not passed the verification when the verification fails; and send the prompt information to the corresponding monitoring terminal.
  • the node sequence is a topological sequence
  • the calculation processing module 608 is also used to extract the node knowledge representation of the node according to the node type and node information; according to the topological sequence and the node knowledge representation, the initial layer node Perform graph calculation on points to obtain the corresponding first subgraph and first subgraph value; according to the first subgraph and first subgraph value, perform graph calculation on the next layer of nodes according to the topological sequence and the corresponding node knowledge representation, Obtain the corresponding second subgraph and second subgraph values; continue to perform iterative graph calculations on multiple nodes according to the obtained current subgraph and current subgraph values according to the topology sequence and the corresponding node knowledge representation; until all nodes Complete the graph calculation and output the target index value corresponding to the index data to be processed.
  • the calculation processing module 608 is also used to extract the first-order predicate logic representation of the node according to the node type and node information; use the preset predicate logic algorithm to perform the analysis of the initial layer node based on the topological sequence and the node knowledge representation.
  • Point for graph calculation extract the current subgraph corresponding to the largest target node; calculate the subgraph value corresponding to the current subgraph, update the current subgraph with the subgraph value, and generate the first subgraph and the first subgraph corresponding to the initial layer node Figure value.
  • the device further includes a model building module, which is used to obtain historical vehicle driving data, perform feature extraction on the historical vehicle driving data, and obtain multiple feature vectors; perform characteristic analysis on the multiple feature vectors to obtain multiple feature vectors.
  • Feature indicators build multiple node structures based on multiple feature indicators; use preset algorithms to build data analysis models based on multiple node structures.
  • the calculation processing module 608 is further configured to decompose the vehicle driving data into multiple sub-task data according to multiple task types, and respectively extract the to-be-processed index data corresponding to the multiple sub-task data; the to-be-processed index data includes the task Type; input the index data to be processed into the data analysis model, and input multiple index data to be processed into the corresponding subtask model according to the task type; extract the node sequence and node information corresponding to the index data to be processed through the subtask model , Perform graph calculation according to the node sequence and node information, and obtain the target index value corresponding to the subtask data.
  • the device further includes an early warning and prompting module, which is used to determine the current driving traffic state value of the vehicle according to the target index value corresponding to the index data to be processed; when the driving traffic state value indicates a violation, according to the driving traffic state value Generate the violation prompt information, and send the violation prompt information to the corresponding monitoring terminal.
  • an early warning and prompting module which is used to determine the current driving traffic state value of the vehicle according to the target index value corresponding to the index data to be processed; when the driving traffic state value indicates a violation, according to the driving traffic state value Generate the violation prompt information, and send the violation prompt information to the corresponding monitoring terminal.
  • the device further includes an early warning prompt module, which is used to calculate the driving risk score of the vehicle according to the target index value corresponding to the index data to be processed; when the driving risk score exceeds the risk threshold, send an early warning prompt message to the monitoring terminal .
  • an early warning prompt module which is used to calculate the driving risk score of the vehicle according to the target index value corresponding to the index data to be processed; when the driving risk score exceeds the risk threshold, send an early warning prompt message to the monitoring terminal .
  • the various modules in the vehicle driving data processing device based on graph calculation can be implemented in whole or in part by software, hardware, and a combination thereof.
  • the above-mentioned modules may be embedded in the form of hardware or independent of the processor in the computer equipment, or may be stored in the memory of the computer equipment in the form of software, so that the processor can call and execute the operations corresponding to the above-mentioned modules.
  • a computer device is provided.
  • the computer device may be a server, and its internal structure diagram may be as shown in FIG. 7.
  • the computer equipment includes a processor, a memory, a network interface, and a database connected through a system bus.
  • the processor of the computer device is used to provide calculation and control capabilities.
  • the memory of the computer device includes a non-volatile storage medium and an internal memory.
  • the non-volatile storage medium stores an operating system, computer readable instructions, and a database.
  • the internal memory provides an environment for the operation of the operating system and computer-readable instructions in the non-volatile storage medium.
  • the database of the computer equipment is used to store data such as vehicle driving data and prompt information.
  • the network interface of the computer device is used to communicate with an external terminal through a network connection.
  • FIG. 7 is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied.
  • the specific computer device may Including more or fewer parts than shown in the figure, or combining some parts, or having a different arrangement of parts.
  • a computer device includes a memory and one or more processors.
  • the memory stores computer readable instructions.
  • the one or more processors execute the above method embodiments. step.
  • One or more non-volatile computer-readable storage media storing computer-readable instructions.
  • the computer-readable instructions execute A step of.
  • Non-volatile memory may include read only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory.
  • Volatile memory may include random access memory (RAM) or external cache memory.
  • RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous chain Channel (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

Landscapes

  • Engineering & Computer Science (AREA)
  • Automation & Control Theory (AREA)
  • Human Computer Interaction (AREA)
  • Transportation (AREA)
  • Mechanical Engineering (AREA)
  • Traffic Control Systems (AREA)

Abstract

一种基于图计算的车辆驾驶数据处理方法，包括：获取车辆驾驶数据，提取所述车辆驾驶数据中的待处理指标数据；将所述待处理指标数据输入至数据分析模型，根据所述待处理指标数据的数据类型对所述待处理指标数据进行存储；提取所述待处理指标数据对应的结点序列和结点信息，根据所述结点序列和所述结点信息对所述待处理指标数据进行验证；当验证通过后，通过所述数据分析模型根据所述结点序列和所述结点信息进行图计算，得到多个子图和相应的子图值；及持续根据所述结点序列对多个子图和相应的子图值进行迭代计算，直到输出所述待处理指标数据对应的目标指标值。

Description

基于图计算的车辆驾驶数据处理方法、装置和计算机设备 技术领域
本申请涉及一种基于图计算的车辆驾驶数据处理方法、装置和计算机设备。
背景技术
图是由一组顶点和一组能够将两个顶点相连的边组成，是一种数据结构。DAG(Directed Acyclic Graph,有向无环图)是一种无回路的有向图。随着计算机技术的发展，有向无环图技术在区块链、分布式系统和高性能计算等领域中被广泛应用。例如在无人驾驶汽车场景中，需要对大量数据进行高性能计算。因此，出现了一些利用DAG技术对数据进行处理的方式。
然而传统的方式中往往需要重复进行DAG软件进行开发，数据处理过程较为复杂，对于海量数据的处理结果容易出现偏差，导致数据处理的效率较低以及准确性难以保证。
发明内容
根据本申请公开的各种实施例，提供一种基于图计算的车辆驾驶数据处理方法、装置和计算机设备。
一种基于图计算的车辆驾驶数据处理方法，包括：
获取车辆驾驶数据，提取所述车辆驾驶数据中的待处理指标数据；
将所述待处理指标数据输入至数据分析模型，根据所述待处理指标数据的数据类型对所述待处理指标数据进行存储；
提取所述待处理指标数据对应的结点序列和结点信息，根据所述结点序列和所述结点信息对所述待处理指标数据进行验证；
当验证通过后，通过所述数据分析模型根据所述结点序列和所述结点信息进行图计算，得到多个子图和相应的子图值；及
持续根据所述结点序列对多个子图和相应的子图值进行迭代计算，直到输出所述待处理指标数据对应的目标指标值。
一种基于图计算的车辆驾驶数据处理装置，包括：
数据获取模块，用于获取车辆驾驶数据，提取所述车辆驾驶数据中的待处理指标数据；
数据存储模块，用于将所述待处理指标数据输入至数据分析模型，根据所述待处理指标数据的数据类型对所述待处理指标数据进行存储；
数据验证模块，用于提取所述待处理指标数据对应的结点序列和结点信息，根据所述结点序列和所述结点信息对所述待处理指标数据进行验证；及
计算处理模块，用于当验证通过后，通过所述数据分析模型根据所述结点序列和所述结点信息进行图计算，得到多个子图和相应的子图值；持续根据所述结点序列对多个子图和相应的子图值进行迭代计算，直到输出所述待处理指标数据对应的目标指标值。
一种计算机设备，包括存储器和处理器，所述存储器存储有计算机程序，所述处理器执行所述计算机程序时实现本申请任意一个实施例中提供的基于远程接管的车辆控制方法的步骤。
一个或多个存储有计算机可读指令的非易失性计算机可读存储介质，计算机可读指令被一个或多个处理器执行时，使得一个或多个处理器执行可读存储介质时实现本申请任意一个实施例中提供的基于远程接管的车辆控制方法的步骤。
本申请的一个或多个实施例的细节在下面的附图和描述中提出。本申请的其它特征和优点将从说明书、附图以及权利要求书变得明显。
附图说明
为了更清楚地说明本申请实施例中的技术方案，下面将对实施例中所需要使用的附图作简单地介绍，显而易见地，下面描述中的附图仅仅是本申请的一些实施例，对于本领域普通技术人员来讲，在不付出创造性劳动的前提下，还可以根据这些附图获得其它的附图。
图1为根据一个或多个实施例中基于图计算的车辆驾驶数据处理方法的应用场景图。
图2为根据一个或多个实施例中基于图计算的车辆驾驶数据处理方法的流程示意图。
图3为根据一个或多个实施例中通过数据分析模型进行图计算步骤的流程示意图。
图4为另一个实施例中通过数据分析模型进行图计算步骤的流程示意图。
图5为根据一个或多个实施例中多任务处理步骤的流程示意图。
图6为根据一个或多个实施例中基于图计算的车辆驾驶数据处理装置的框图。
图7为根据一个或多个实施例中计算机设备的框图。
具体实施方式
为了使本申请的技术方案及优点更加清楚明白，以下结合附图及实施例，对本申请进行进一步详细说明。应当理解，此处描述的具体实施例仅仅用以解释本申请，并不用于限定本申请。
本申请提供的基于图计算的车辆驾驶数据处理方法，可以应用于多种应用环境中。例如，可以应用于如图1所示的自动驾驶的应用环境中。车辆中包括传感器102和处理器 104。传感器102可以通过网络与处理器104进行通信。传感器102可以采集车辆的车辆驾驶数据，处理器104提取车辆驾驶数据中的待处理指标数据后将待处理指标数据输入至数据分析模型，根据待处理指标数据的数据类型对待处理指标数据进行存储；提取待处理指标数据对应的结点序列和结点信息，根据结点序列和结点信息对待处理指标数据进行验证；当验证通过后，处理器104通过数据分析模型根据结点序列和结点信息进行图计算，得到多个子图和相应的子图值；持续根据结点序列对多个子图和相应的子图值进行迭代计算，直到输出待处理指标数据对应的目标指标值。传感器102可以是自动驾驶设备搭载的多种传感器，具体可以包括激光雷达、激光扫描仪等。
在其中一个实施例中，如图2所示，提供了一种基于图计算的车辆驾驶数据处理方法，以该方法应用于图1中的车辆为例进行说明，包括以下步骤：
步骤202，获取车辆驾驶数据，提取车辆驾驶数据中的待处理指标数据。
车辆可以为无人驾驶汽车，无人驾驶汽车是通过车载传感系统感知道路环境，自动规划行车路线并控制车辆到达预定目标的智能汽车。车辆在自动行驶过程中实时记录车辆驾驶数据，车辆驾驶数据可以包括道路影像信息和车辆状态信息，道路影像信息可以包括采集的视频数据和道路画面视频数据，道路影像信息可以为连续的视频帧。车辆状态信息可以包括车辆定位信息、车辆导航信息、车内温度信息、车辆仪表信息等车辆的状态信息。传道路影像信息和车辆状态信息用于远程接管人员根据道路情况和车辆状态对车辆进行远程接管。
车辆获取车辆驾驶数据后，对车辆驾驶数据进行指标特征提取，从而提取出多个待处理指标数据。例如，待处理指标数据可以包括车速、里程、油量等多个指标对应的数据，待处理指标数据还可以是每一帧道路影像对应的指标数据。其中，待处理指标数据可以包括多个任务类型对应的待处理指标数据。例如，任务类型可以包括交通规则识别、车辆风险评估等多个类型。
步骤204，将待处理指标数据输入至数据分析模型，根据待处理指标数据的数据类型对待处理指标数据进行存储。
其中，数据分析模型可以是基于DAG的有向无环图计算模型，有向无环图为预先利用历史样本数据按照预设算法所构建的，有向无环图模型中包括多个预先构建的结点结构。
车辆提取出多个待处理指标数据后，将多个待处理指标数据输入至预设的数据分析模型中，具体地，可以将多个待处理指标数据按照相应的指标类型填充至有向无环图模型中相应的结点结构中。
数据分析模型可以识别多个待处理指标数据的数据类型，根据数据类型确定多个待 处理指标数据的存储结构，并将多个待处理数据存储至分配的存储结构中。其中，存储结构可以是硬盘中的存储结构，由此能够有效地对待处理指标数据进行持久化存储，以保障待处理指标数据在处理过程中的一致性和完整性。
数据分析模型还可以进一步识别待处理指标数据的操作类型，进而根据操作类型和数据类型确定待处理指标数据的运行结构，以用于缓存对结点信息进行处理过程中产生的运行数据。
步骤206，提取待处理指标数据对应的结点序列和结点信息，根据结点序列和结点信息对待处理指标数据进行验证。
车辆将待处理指标数据输入至数据分析模型并存储后，进一步通过数据分析模型提取多个待处理指标数据对应的结点序列和每个结点的结点信息，并根据结点序列和结点信息对待处理指标数据进行验证。
具体地，数据分析模型可以提取结点序列、结点数目，还可以获取每个结点的结点信息，结点信息可以包括数据类型、变量作用域以及该结点的子结点数目等信息。数据分析模型进而对结点序列和结点信息的有效性进行检测和验证，例如，检测结点的子结点数目错误、数据类型错误，以及变量作用域的错误等问题。只有当结点序列和结点信息均满足合法有效的条件时，才表示待处理数据验证通过。
具体地，结点序列可以为对多个结点进行拓扑排序得到的拓扑序列。在有向无环图中，图中的顶点代表活动，图中的有向边代表活动的先后关系，即有向边的起点的活动是终点活动的前序活动，只有当起点活动完成之后，其终点活动才能进行。例如，可以通过顶点活动网对有向无环图中的所有结点构造拓扑序列，选择一个入度为零的顶点并输出，从有向无环图的网中删除此顶点及所有出边。循环结束后，若输出的顶点数小于网中的顶点数，则输出“有环路”信息，否则输出的顶点序列就是一种拓扑序列。从而输出有向无环图的拓扑排序。
步骤208，当验证通过后，通过数据分析模型根据结点序列和结点信息进行图计算，得到多个子图和相应的子图值。
对待处理指标数据验证通过后，车辆则利用数据分析模型按照结点序列和结点信息对数据分析模型中的多个结点进行图计算。具体地，数据分析模型根据预设的逻辑算法计算当前结点对应的子图，并对该子图进行求值。数据分析模型还可以利用得到的值对该子图进行更新，得到更新后的简化的当前子图以及相应的子图值。
步骤210，持续根据结点序列对多个子图和相应的子图值进行迭代计算，直到输出待处理指标数据对应的目标指标值。
数据分析模型进而持续根据结点序列对多个子图和相应的子图值进行迭代计算，每一 轮都得到相应的更新的子图以及相应的子图值，再根据更新的子图和子图值按照结点序列对下一个结点信息进行图计算，直到所有的结点计算完成，进而通过数据分析模型输出计算得到的结果，该结果即为待处理指标数据对应的目标指标值。
具体地，车辆可以利用数据分析模型对待处理指标数据进行图计算，得到车辆交通指标值和车辆风险指标值等多个目标指标对应的结果值。由于数据分析模型为基于图计算的模型，通过利用数据分析模型对车辆产生的大量车辆驾驶数据进行计算处理，能够有效提高数据分析模型的计算能力，从而能够快速准确地对车辆驾驶数据进行计算处理，进而计算得到准确性较高的目标指标值。
上述基于图计算的车辆驾驶数据处理方法中，车辆获取车辆驾驶数据并提取出车辆驾驶数据中的待处理指标数据后，将待处理指标数据输入至数据分析模型中。通过数据分析模型提取待处理指标数据对应的结点序列和结点信息，根据结点序列和结点信息对待处理指标数据进行验证，以保证输入数据的有效性。当验证通过后，通过数据分析模型根据结点序列和结点信息进行图计算，由于数据分析模型为基于图计算的模型，从而能够有效计算得到多个子图和相应的子图值。数据分析模型持续根据结点序列对多个子图和相应的子图值进行迭代计算，直到输出待处理指标数据对应的目标指标值。通过利用预先构建的基于图计算模型的数据分析模型对车辆产生的大量车辆驾驶数据进行计算处理，能够有效提高数据分析模型的计算能力，从而能够快速准确地对车辆驾驶数据进行计算处理，有效提高了车辆驾驶数据的计算处理效率和准确性。
在其中一个实施例中，根据待处理指标数据的数据类型对待处理指标数据进行存储包括：根据待处理指标数据的数据类型确定待处理指标数据的存储结构，根据存储结构对待处理指标数据进行存储；根据结点信息的结点类型确待处理指标数据对应的运行结构；运行结构用于缓存对结点信息进行处理过程中产生的运行数据。
车辆在驾驶过程中会产生大量的车辆驾驶数据，车辆将提取出的待处理指标数据输入至数据分析模型后，需要对待处理指标数据进行持久化存储，以保证数据在传输和计算过程中的一致性和有效性。
具体地，车辆通过数据分析模型确定待处理指标数据的数据类型和对应的多个结点信息，例如，数据分析模型中可以包括预先构建的多个结点，车辆则可以直接将待处理指标数据填充至相应的结点中。车辆则根据待处理指标数据的数据类型确定待处理指标数据的存储结构，进而根据存储结构对待处理指标数据进行存储。
举例说明，如可以通过构建ProtoBuf(Protocol Buffers，可扩展的序列化结构数据方式)的格式文件，ProtoBuf是一种语言无关、平台无关、可扩展的序列化数据结构，它可用于数据通信协议和数据存储等。利用ProtoBuf格式文件对待处理指标数据进行存储处理 时，可以定义Graph、Node、Operation等数据类型。其中，Graph可以包括repeated属性的结点字段，以及用于说明根节点的ID的字段。Operation是一种枚举类型，可以包括数据分析模型的所有运算操作。例如至少包括LITERAL(字面值)、AND(“与”运算)、OR(“或”运算)、NOT(“非”运算)、EXISTENTIAL(存在量词)、UNIVERSAL(全称量词)、VARIABLE(变元)、FUNCTION(函数)。Node类型可以包括：一个任意数据类型的ID，用于标识结点；一个Operation类型的字段，用于说明结点对应的运算类型；两个repeated属性的与ID数据类型相同的字段，分别用于列出当前结点的前继结点和后继结点。Operation类的所有枚举值所对应的运算所需的字段，例如UNIVERSAL运算需要一个字段用于存储全称量词所声明的变元的名字，和另一个字段用于说明该全称量词所限定的变元取值范围。
车辆根据存储结构对待处理指标数据进行存储后，还可以进一步根据结点信息的结点类型确待处理指标数据对应的运行结构，运行结构用于缓存对结点信息进行处理过程中产生的运行数据。
在基于图计算的数据分析模型的实现过程中，图计算模型和所包含的所有结点可以被封装为自定义类型的实例，从而能够便捷地对其结点进行读取和操作，从而完成验证、计算处理等操作。从而可以得到一个具有很强的逻辑性、泛用性和可拓展性的运行结构，能够在便捷管理各类结点的同时有效地对不同运算结点进行特定化操作处理，从而能够有效提高待处理指标数据的处理效率。
例如，在实现待处理指标数据对应的运行结构时，可以首先定义一个Graph类型，该类型的每个实例都对应一个DAG有向无环图。Graph类型应提供根据ID来快速查找结点的手段。进一步地，还可以对每一种运算操作类型都定义一个相应的结点类型。可以通过定义一个NodeBase类型作为其他所有结点类型的基类，并在这一类型中实现所有结点的通用操作，例如获取其ID、查找其前继结点和后继结点等。然后，对于每一个类型的运算操作，定义一个相应的Node类型继承NodeBase来表示属于该运算的结点。在定义中，可以根据运算的实际语义提供各个虚函数的定义实现。为了确保高度的泛用性和支持任意的数据类型，还可以将这些类型都定义为模板类。例如，一个表示乘法运算的结点类型可以被用在虚数的运算中，也可以表示矩阵的数乘。而对于一些只针对特定数据类型的运算(如与、或、非)，则可以不必定义为模板类。例如可以利用Protobuf库实现Protobuf文件与上述数据类型的转换。在其中一个实施例中，上述方式可以利用C++语言实现。
在其中一个实施例中，根据结点序列和结点信息对待处理指标数据进行验证，包括：根据结点序列和结点信息提取待处理指标数据的多个结点特征参数；当多个结点特征参数均满足预设条件阈值时，确定待处理指标数据验证通过。
车辆将待处理指标数据输入至数据分析模型并存储后，进一步通过数据分析模型提取多个待处理指标数据对应的结点序列和每个结点的结点信息，并根据结点序列和结点信息对待处理指标数据进行验证。具体地，数据分析模型可以根据结点序列和结点信息提取待处理指标数据的多个结点特征参数。结点特征参数可以包括是否有环、根结点入度、游离结点、作用域、数据类型以及前后继关系等多个参数字段以及相应的参数值。
数据分析模型进而对多个结点特征参数的有效性进行验证，检测各个结点特征参数是否合法。例如，当有向无环图中存在环路、根结点的入度为1、存在游离结点、存在作用域错误或变量重名、存在数据类型错误或前后继关系错误等任意一种错误情况时，则表示待处理指标数据验证未通过，无法进行进一步求值计算。只有当所有的结点特征参数均满足合法有效的条件时，才表示待处理数据验证通过。通过对待处理指标数据进行合法性验证，能够有效保证待处理指标数据的处理效率和处理结果的准确性。
在其中一个实施例中，该方法还包括：当验证未通过时，根据未通过验证的结点特征参数生成提示信息；将提示信息发送至相应的监控终端。
每个车辆可以绑定有相应的监控终端，以在车辆存在故障或异常情况时对车辆进行监控等处理。
数据分析模型对多个结点特征参数的有效性进行验证过程中，当存在错误情形的结点特征参数时，表示待处理指标数据验证未通过。车辆进而根据验证未通过的结点特征参数生成相应的提示信息，并将提示信息发送至相应的监控终端，使得监控终端根据提示信息对车辆的驾驶情况进行监控，从而有效保障车辆的正常运行。
在其中一个实施例中，如图3所示，结点序列为拓扑序列，通过数据分析模型根据结点序列和结点信息进行图计算的步骤，具体包括以下内容：
步骤302，根据结点类型和结点信息提取结点的结点知识表示。
步骤304，根据拓扑序列和结点知识表示对初始层结点进行图计算，得到对应的第一子图和第一子图值。
步骤306，根据第一子图和第一子图值按照拓扑序列和相应的结点知识表示对下一层结点进行图计算，得到对应的第二子图和第二子图值。
步骤308，持续根据得到的当前子图和当前子图值按照拓扑序列和相应的结点知识表示对多个结点进行迭代图计算。
步骤310，直到所有结点完成图计算，输出待处理指标数据对应的目标指标值。
知识表示是指某领域中所涉及的各有关方面的一种符号表示。知识表示可以包括一阶谓词逻辑表示、产生式规则、框架表示法、脚本表示法、语义网表示法、知识图谱表示法等几种表示。结点知识表示可以为每个结点和结点之间的知识表示。
车辆将待处理指标数据输入至数据分析模型中，并对待处理指标数据验证通过后，数据分析模型则进一步对待处理指标数据对应的多个结点进行图计算处理，以迭代计算出待处理指标数据对应的目标指标值。
具体地，车辆可以通过数据分析模型根据结点类型和结点信息提取结点的结点知识表示，结点的结点知识表示包括每个结点的知识表示和结点之间的知识表示。数据分析模型进而根据拓扑序列和结点知识表示首先对初始层结点进行图计算，得到对应的第一子图和第一子图值。数据分析模型继续根据第一子图和第一子图值按照拓扑序列和相应的结点知识表示对下一层结点进行图计算，得到对应的第二子图和第二子图值。数据分析模型可以利用得到的值对该子图进行更新，得到更新后的简化的当前的第二子图以及相应的第二子图值。
数据分析模型持续根据得到的当前子图和当前子图值按照拓扑序列和相应的结点知识表示对多个结点进行迭代图计算。每一轮都得到相应的更新的子图以及相应的子图值，再根据更新的子图和子图值按照结点序列对下一个结点信息进行图计算，直到所有的结点都完成图计算。进而通过数据分析模型输出计算得到的结果，该结果即为待处理指标数据对应的目标指标值。
具体地，车辆可以利用数据分析模型对待处理指标数据进行图计算，得到车辆交通指标值和车辆风险指标值等多个目标指标对应的结果值。由于数据分析模型为基于图计算的模型，通过利用数据分析模型对车辆产生的大量车辆驾驶数据进行计算处理，能够有效提高数据分析模型的计算能力，从而能够快速准确地对车辆驾驶数据进行计算处理，进而计算得到准确性较高的目标指标值。
在其中一个实施例中，如图4所示，根据拓扑序列和结点知识表示对初始层结点进行图计算，得到对应的第一子图和第一子图值的步骤，具体包括以下内容：
步骤402，根据结点类型和结点信息提取结点的一阶谓词逻辑表示。
步骤404，利用预设谓词逻辑算法根据拓扑序列和结点知识表示对初始层结点进行图计算，提取最大目标结点对应的当前子图。
步骤406，计算当前子图对应的子图值，利用子图值更新当前子图，生成初始层结点对应的第一子图和第一子图值。
知识数据一般可以由具有完整意义的一句话或几句话表示出来，这些知识可以用谓词逻辑表示出来，例如可为一个谓词公式，即通过用谓词联接符号将一些谓词联接起来所形成的公式。谓词公式可以表示事物的状态、属性和概念等事实性的知识，还可以表示事物间具有确定因果关系的规则性知识。
一阶谓词逻辑是一种形式语言表示形式，通过命题、逻辑联结词、个词体、谓词与量 词等部件组成的知识表示，这种知识表示在数据处理上较为精确。
车辆将待处理指标数据输入至数据分析模型中，并对待处理指标数据验证通过后，数据分析模型则进一步对待处理指标数据对应的多个结点进行图计算处理，以迭代计算出待处理指标数据对应的目标指标值。
车辆首先通过数据分析模型根据结点类型和结点信息提取结点的结点知识表示，具体地，结点知识表示可以为一阶谓词逻辑表示，即可以通过数据分析模型根据结点类型和结点信息提取结点的一阶谓词逻辑表示。结点的一阶谓词逻辑表示包括每个结点的一阶谓词逻辑表示和结点之间的一阶谓词逻辑表示。
数据分析模型进而利用预设谓词逻辑算法根据拓扑序列和结点知识表示对初始层结点进行图计算，提取最大目标结点对应的当前子图。计算当前子图对应的子图值，利用子图值更新当前子图，生成初始层结点对应的第一子图和第一子图值。在对有向无环图模型进行迭代图计算的过程中，需要找到当前的有向无环图中的最大的只包含逻辑运算结点的子图，对该子图进行求值，而后用所得的值来代替该子图，得到一个简化的子图。然后再根据所得到的新图继续进行计算，直到得到有向无环图图的输出结果为止。通过数据分析模型利用谓词逻辑算法对车辆驾驶数据进行计算处理，能够快速准确地对车辆驾驶数据进行计算处理，从而能够有效得到准确性较高的目标指标值。
在其中一个实施例中，在将待处理指标数据输入至数据分析模型之前，该方法还包括：获取历史车辆驾驶数据，对历史车辆驾驶数据进行特征提取，得到多个特征向量；对多个特征向量进行特性分析，得到多个特征指标；根据多个特征指标构建多个结点结构；根据多个结点结构利用预设算法构建数据分析模型。
车辆在对产生的车辆驾驶数据进行计算处理之前，需要预先构建出相应任务类型的的数据分析模型。具体地，车辆可以获取若干历史车辆驾驶数据，这些历史车辆驾驶数据可以是车辆自身产生的历史数据，也可以车辆平台的多个车辆产生的历史数据。
车辆进而对历史车辆驾驶数据进行特征提取，例如可以采用聚类分析算法对若干历史车辆驾驶数据进行特征聚类和特征提取，从而得到多个特征向量。车辆进一步对提取出的多个特征向量进行特性分析，得到多个特征指标。其中，车辆可以根据不同的任务类型分别提取每个任务类型对应的特征指标。从而可以根据多个特征指标构建多个结点结构，进而能够有效地根据多个结点结构利用预设算法构建基于有向无环图的数据分析模型。其中，所构建的数据分析模型中可以包括多个任务类型对应的有向无环图模型。由此能够有效地根据不同任务类型对车辆驾驶数据进行计算处理，有效提高了数据处理效率。
在其中一个实施例中，如图5所示，数据分析模型包括多个子任务模型，该方法还包括多任务处理步骤，具体包括以下内容：
步骤502，根据多个任务类型将车辆驾驶数据分解为多个子任务数据，分别提取多个子任务数据对应的待处理指标数据；待处理指标数据包括任务类型。
步骤504，将待处理指标数据输入至数据分析模型，根据任务类型分别将多个待处理指标数据输入至相应的子任务模型。
步骤506，通过子任务模型提取待处理指标数据对应的结点序列和结点信息，根据结点序列和结点信息进行图计算，得到子任务数据对应的目标指标值。
数据分析模型中可以包括多个任务类型对应的有向无环图模型，车辆则可以通过数据分析模型同时对多个任务类型的车辆驾驶数据进行分析处理，例如，任务类型可以包括交通规则识别、车辆风险评估等多个类型。
车辆获取车辆驾驶数据后，根据多个任务类型将车辆驾驶数据分解为多个子任务数据，进而对多个子任务数据进行指标特征提取，分别提取多个子任务数据对应的待处理指标数据；每种待处理指标数据则包括了相应的任务类型。
车辆提取出多个任务类型对应的待处理指标数据后，将多个待处理指标数据输入至预设的数据分析模型中，具体地，数据分析模型根据任务类型分别将多个待处理指标数据输入至相应的子任务模型中，并将多个待处理指标数据按照相应的指标类型填充至有向无环图模型中相应的结点结构中。
每个子任务模型分别同时对输入的待处理指标数据进行存储并处理，提取多个待处理指标数据对应的结点序列和每个结点的结点信息，并根据结点序列和结点信息对待处理指标数据进行验证。当验证通过后，子任务模型根据结点序列和结点信息进行图计算，得到子任务数据对应的多个子图和相应的子图值。持续根据结点序列对多个子图和相应的子图值进行迭代计算，直到输出子任务数据对应的目标指标值。通过包括多个子任务模型的数据分析模型同时对多个任务的车辆驾驶数据进行处理，由此能够有效地根据不同任务类型对车辆驾驶数据进行计算处理，有效提高了数据处理效率。
在其中一个实施例中，该方法还包括：根据待处理指标数据对应的目标指标值确定车辆当前的驾驶交通状态值；当驾驶交通状态值表示违规时，根据驾驶交通状态值生成违规提示信息，并将违规提示信息发送至相应的监控终端。
车辆获取车辆驾驶数据后，根据多个任务类型将车辆驾驶数据分别提取多个子任务数据对应的待处理指标数据，将多个待处理指标数据输入至预设的数据分析模型中，通过每个子任务模型分别同时对输入的待处理指标数据进行存储并处理，得到子任务数据对应的多个子图和相应的子图值。持续根据结点序列对多个子图和相应的子图值进行迭代计算，直到输出子任务数据对应的目标指标值。
当任务类型为交通状态分析时，车辆则可以根据该子任务数据对应的目标指标值确 定车辆当前的驾驶交通状态值。例如可以根据预设的指标映射表确定目标指标值对应的驾驶交通状态值。当驾驶交通状态值表示违规时，表示车辆当前存在违反交通规则的情况，车辆根据驾驶交通状态值生成违规提示信息，并将违规提示信息发送至相应的监控终端，使得监控终端及时对车辆进行监控和管理。
在其中一个实施例中，方法还包括：根据待处理指标数据对应的目标指标值计算车辆的驾驶风险评分；当驾驶风险评分超过风险阈值时，向监控终端发送预警提示信息。
当任务类型为驾驶风险状态分析时，车辆则可以根据该子任务数据对应的目标指标值确定车辆当前的驾驶风险评分。例如可以根据预设的指标目标值和相应的权重计算车辆当前的驾驶风险评分，如可以包括车辆设备风险、道路障碍物风险、外部风险等多个维度的指标值，从而有效地计算出车辆在驾驶过程中的驾驶风险评分。当驾驶风险评分超过风险阈值时，表示车辆当前存在风险因素，车辆根据则根据相应的风险指标生成预警提示信息，并将预警提示信息发送至相应的监控终端，使得监控终端及时对车辆进行监控和管理。
应该理解的是，虽然图2-5的流程图中的各个步骤按照箭头的指示依次显示，但是这些步骤并不是必然按照箭头指示的顺序依次执行。除非本文中有明确的说明，这些步骤的执行并没有严格的顺序限制，这些步骤可以以其它的顺序执行。而且，图2-5中的至少一部分步骤可以包括多个子步骤或者多个阶段，这些子步骤或者阶段并不必然是在同一时刻执行完成，而是可以在不同的时刻执行，这些子步骤或者阶段的执行顺序也不必然是依次进行，而是可以与其它步骤或者其它步骤的子步骤或者阶段的至少一部分轮流或者交替地执行。
在一个实施例中，如图6所示，提供了一种基于图计算的车辆驾驶数据处理装置，包括：数据获取模块602、数据存储模块604、数据验证模块606和计算处理模块608，其中：
数据获取模块602，用于获取车辆驾驶数据，提取车辆驾驶数据中的待处理指标数据；
数据存储模块604，用于将待处理指标数据输入至数据分析模型，根据待处理指标数据的数据类型对待处理指标数据进行存储；
数据验证模块606，用于提取待处理指标数据对应的结点序列和结点信息，根据结点序列和结点信息对待处理指标数据进行验证；及
计算处理模块608，用于当验证通过后，通过数据分析模型根据结点序列和结点信息进行图计算，得到多个子图和相应的子图值；持续根据结点序列对多个子图和相应的子图值进行迭代计算，直到输出待处理指标数据对应的目标指标值。
在其中一个实施例中，数据存储模块604还用于根据待处理指标数据的数据类型确定待处理指标数据的存储结构，根据存储结构对待处理指标数据进行存储；根据结点信息的 结点类型确待处理指标数据对应的运行结构；运行结构用于缓存对结点信息进行处理过程中产生的运行数据。
在其中一个实施例中，数据验证模块606还用于根据结点序列和结点信息提取待处理指标数据的多个结点特征参数；当多个结点特征参数均满足预设条件阈值时，确定待处理指标数据验证通过。
在其中一个实施例中，数据验证模块606还用于当验证未通过时，根据未通过验证的结点特征参数生成提示信息；及将提示信息发送至相应的监控终端。
在其中一个实施例中，结点序列为拓扑序列，计算处理模块608还用于根据结点类型和结点信息提取结点的结点知识表示；根据拓扑序列和结点知识表示对初始层结点进行图计算，得到对应的第一子图和第一子图值；根据第一子图和第一子图值按照拓扑序列和相应的结点知识表示对下一层结点进行图计算，得到对应的第二子图和第二子图值；持续根据得到的当前子图和当前子图值按照拓扑序列和相应的结点知识表示对多个结点进行迭代图计算；直到所有结点完成图计算，输出待处理指标数据对应的目标指标值。
在其中一个实施例中，计算处理模块608还用于根据结点类型和结点信息提取结点的一阶谓词逻辑表示；利用预设谓词逻辑算法根据拓扑序列和结点知识表示对初始层结点进行图计算，提取最大目标结点对应的当前子图；计算当前子图对应的子图值，利用子图值更新当前子图，生成初始层结点对应的第一子图和第一子图值。
在其中一个实施例中，该装置还包括模型构建模块，用于获取历史车辆驾驶数据，对历史车辆驾驶数据进行特征提取，得到多个特征向量；对多个特征向量进行特性分析，得到多个特征指标；根据多个特征指标构建多个结点结构；根据多个结点结构利用预设算法构建数据分析模型。
在其中一个实施例中，计算处理模块608还用于根据多个任务类型将车辆驾驶数据进行分解为多个子任务数据，分别提取多个子任务数据对应的待处理指标数据；待处理指标数据包括任务类型；将待处理指标数据输入至数据分析模型，根据任务类型分别将多个待处理指标数据输入至相应的子任务模型；通过子任务模型提取待处理指标数据对应的结点序列和结点信息，根据结点序列和结点信息进行图计算，得到子任务数据对应的目标指标值。
在其中一个实施例中，该装置还包括预警提示模块，用于根据待处理指标数据对应的目标指标值确定车辆当前的驾驶交通状态值；当驾驶交通状态值表示违规时，根据驾驶交通状态值生成违规提示信息，并将违规提示信息发送至相应的监控终端。
在其中一个实施例中，该装置还包括预警提示模块，用于根据待处理指标数据对应的目标指标值计算车辆的驾驶风险评分；当驾驶风险评分超过风险阈值时，向监控终端发送 预警提示信息。
关于基于图计算的车辆驾驶数据处理装置的具体限定可以参见上文中对于基于图计算的车辆驾驶数据处理方法的限定，在此不再赘述。上述基于图计算的车辆驾驶数据处理装置中的各个模块可全部或部分通过软件、硬件及其组合来实现。上述各模块可以硬件形式内嵌于或独立于计算机设备中的处理器中，也可以以软件形式存储于计算机设备中的存储器中，以便于处理器调用执行以上各个模块对应的操作。
在一个实施例中，提供了一种计算机设备，该计算机设备可以是服务器，其内部结构图可以如图7所示。该计算机设备包括通过系统总线连接的处理器、存储器、网络接口和数据库。其中，该计算机设备的处理器用于提供计算和控制能力。该计算机设备的存储器包括非易失性存储介质、内存储器。该非易失性存储介质存储有操作系统、计算机可读指令和数据库。该内存储器为非易失性存储介质中的操作系统和计算机可读指令的运行提供环境。该计算机设备的数据库用于存储车辆驾驶数据、提示信息等数据。该计算机设备的网络接口用于与外部的终端通过网络连接通信。该计算机可读指令被处理器执行时以实现一种基于图计算的车辆驾驶数据处理方法。
本领域技术人员可以理解，图7中示出的结构，仅仅是与本申请方案相关的部分结构的框图，并不构成对本申请方案所应用于其上的计算机设备的限定，具体的计算机设备可以包括比图中所示更多或更少的部件，或者组合某些部件，或者具有不同的部件布置。
一种计算机设备，包括存储器和一个或多个处理器，存储器中储存有计算机可读指令，计算机可读指令被处理器执行时，使得一个或多个处理器执行时实现上述方法实施例中的步骤。
一个或多个存储有计算机可读指令的非易失性计算机可读存储介质，计算机可读指令被一个或多个处理器执行时，使得一个或多个处理器执行时实现上述方法实施例中的步骤。
本领域普通技术人员可以理解实现上述实施例方法中的全部或部分流程，是可以通过计算机可读指令来指令相关的硬件来完成，所述的计算机可读指令可存储于一非易失性计算机可读取存储介质中，该计算机可读指令在执行时，可包括如上述各方法的实施例的流程。其中，本申请所提供的各实施例中所使用的对存储器、存储、数据库或其它介质的任何引用，均可包括非易失性和/或易失性存储器。非易失性存储器可包括只读存储器(ROM)、可编程ROM(PROM)、电可编程ROM(EPROM)、电可擦除可编程ROM(EEPROM)或闪存。易失性存储器可包括随机存取存储器(RAM)或者外部高速缓冲存储器。作为说明而非局限，RAM以多种形式可得，诸如静态RAM(SRAM)、动态RAM(DRAM)、同步DRAM(SDRAM)、双数据率SDRAM(DDRSDRAM)、增强 型SDRAM(ESDRAM)、同步链路(Synchlink)DRAM(SLDRAM)、存储器总线(Rambus)直接RAM(RDRAM)、直接存储器总线动态RAM(DRDRAM)、以及存储器总线动态RAM(RDRAM)等。
以上实施例的各技术特征可以进行任意的组合，为使描述简洁，未对上述实施例中的各个技术特征所有可能的组合都进行描述，然而，只要这些技术特征的组合不存在矛盾，都应当认为是本说明书记载的范围。
以上所述实施例仅表达了本申请的几种实施方式，其描述较为具体和详细，但并不能因此而理解为对发明专利范围的限制。应当指出的是，对于本领域的普通技术人员来说，在不脱离本申请构思的前提下，还可以做出若干变形和改进，这些都属于本申请的保护范围。因此，本申请专利的保护范围应以所附权利要求为准。

Claims (20)

  1. 一种基于图计算的车辆驾驶数据处理方法，包括：
    获取车辆驾驶数据，提取所述车辆驾驶数据中的待处理指标数据；
    将所述待处理指标数据输入至数据分析模型，根据所述待处理指标数据的数据类型对所述待处理指标数据进行存储；
    提取所述待处理指标数据对应的结点序列和结点信息，根据所述结点序列和所述结点信息对所述待处理指标数据进行验证；
    当验证通过后，通过所述数据分析模型根据所述结点序列和所述结点信息进行图计算，得到多个子图和相应的子图值；及
    持续根据所述结点序列对多个子图和相应的子图值进行迭代计算，直到输出所述待处理指标数据对应的目标指标值。
  2. 根据权利要求1所述的方法，其特征在于，所述根据所述待处理指标数据的数据类型对所述待处理指标数据进行存储包括：
    根据所述待处理指标数据的数据类型确定所述待处理指标数据的存储结构，根据所述存储结构对所述待处理指标数据进行存储；及
    根据结点信息的结点类型确所述待处理指标数据对应的运行结构；所述运行结构用于缓存对所述结点信息进行处理过程中产生的运行数据。
  3. 根据权利要求1所述的方法，其特征在于，所述根据所述结点序列和所述结点信息对所述待处理指标数据进行验证，包括：
    根据所述结点序列和所述结点信息提取所述待处理指标数据的多个结点特征参数；及当所述多个结点特征参数均满足预设条件阈值时，确定所述待处理指标数据验证通过。
  4. 根据权利要求3所述的方法，其特征在于，所述方法还包括：
    当验证未通过时，根据未通过验证的结点特征参数生成提示信息；及
    将所述提示信息发送至相应的监控终端。
  5. 根据权利要求1所述的方法，其特征在于，所述结点序列为拓扑序列，所述通过所述数据分析模型根据所述结点序列和所述结点信息进行图计算，包括：
    根据所述结点类型和所述结点信息提取结点的结点知识表示；
    根据所述拓扑序列和所述结点知识表示对初始层结点进行图计算，得到对应的第一子图和第一子图值；
    根据所述第一子图和第一子图值按照所述拓扑序列和相应的结点知识表示对下一层结点进行图计算，得到对应的第二子图和第二子图值；
    持续根据得到的当前子图和当前子图值按照所述拓扑序列和相应的结点知识表示对多个结点进行迭代图计算；及
    直到所有结点完成图计算，输出所述待处理指标数据对应的目标指标值。
  6. 根据权利要求5所述的方法，其特征在于，所述根据所述拓扑序列和所述结点知识表示对初始层结点进行图计算，得到对应的第一子图和第一子图值，包括：
    根据所述结点类型和所述结点信息提取结点的一阶谓词逻辑表示；
    利用预设谓词逻辑算法根据所述拓扑序列和所述结点知识表示对初始层结点进行图计算，提取最大目标结点对应的当前子图；及
    计算所述当前子图对应的子图值，利用所述子图值更新所述当前子图，生成所述初始层结点对应的第一子图和第一子图值。
  7. 根据权利要求1所述的方法，其特征在于，在所述将所述待处理指标数据输入至数据分析模型之前，所述方法还包括：
    获取历史车辆驾驶数据，对所述历史车辆驾驶数据进行特征提取，得到多个特征向量；
    对所述多个特征向量进行特性分析，得到多个特征指标；
    根据所述多个特征指标构建多个结点结构；及
    根据所述多个结点结构利用预设算法构建数据分析模型。
  8. 根据权利要求1所述的方法，其特征在于，所述数据分析模型包括多个子任务模型，所述方法还包括：
    根据多个任务类型将所述车辆驾驶数据进行分解为多个子任务数据，分别提取多个子任务数据对应的待处理指标数据；所述待处理指标数据包括任务类型；
    将所述待处理指标数据输入至数据分析模型，根据所述任务类型分别将多个待处理指标数据输入至相应的子任务模型；及
    通过所述子任务模型提取所述待处理指标数据对应的结点序列和结点信息，根据所述结点序列和所述结点信息进行图计算，得到所述子任务数据对应的目标指标值。
  9. 根据权利要求1所述的方法，其特征在于，所述方法还包括：
    根据所述待处理指标数据对应的目标指标值确定车辆当前的驾驶交通状态值；及
    当所述驾驶交通状态值表示违规时，根据所述驾驶交通状态值生成违规提示信息，并将所述违规提示信息发送至相应的监控终端。
  10. 根据权利要求1所述的方法，其特征在于，所述方法还包括：
    根据所述待处理指标数据对应的目标指标值计算所述车辆的驾驶风险评分；及
    当所述驾驶风险评分超过风险阈值时，向监控终端发送预警提示信息。
  11. 一种基于图计算的车辆驾驶数据处理装置，包括：
    数据获取模块，用于获取车辆驾驶数据，提取所述车辆驾驶数据中的待处理指标数据；
    数据存储模块，用于将所述待处理指标数据输入至数据分析模型，根据所述待处理指标数据的数据类型对所述待处理指标数据进行存储；
    数据验证模块，用于提取所述待处理指标数据对应的结点序列和结点信息，根据所述结点序列和所述结点信息对所述待处理指标数据进行验证；及
    计算处理模块，用于当验证通过后，通过所述数据分析模型根据所述结点序列和所述结点信息进行图计算，得到多个子图和相应的子图值；持续根据所述结点序列对多个子图和相应的子图值进行迭代计算，直到输出所述待处理指标数据对应的目标指标值。
  12. 一种计算机设备，包括存储器及一个或多个处理器，所述存储器中储存有计算机可读指令，所述计算机可读指令被所述一个或多个处理器执行时，使得所述一个或多个处理器执行以下步骤：
    获取车辆驾驶数据，提取所述车辆驾驶数据中的待处理指标数据；
    将所述待处理指标数据输入至数据分析模型，根据所述待处理指标数据的数据类型对所述待处理指标数据进行存储；
    提取所述待处理指标数据对应的结点序列和结点信息，根据所述结点序列和所述结点信息对所述待处理指标数据进行验证；
    当验证通过后，通过所述数据分析模型根据所述结点序列和所述结点信息进行图计算，得到多个子图和相应的子图值；及
    持续根据所述结点序列对多个子图和相应的子图值进行迭代计算，直到输出所述待处理指标数据对应的目标指标值。
  13. 根据权利要求12所述的计算机设备，其特征在于，所述处理器执行所述计算机可读指令时还执行以下步骤：
    根据所述待处理指标数据的数据类型确定所述待处理指标数据的存储结构，根据所述存储结构对所述待处理指标数据进行存储；及
    根据结点信息的结点类型确所述待处理指标数据对应的运行结构；所述运行结构用于缓存对所述结点信息进行处理过程中产生的运行数据。
  14. 根据权利要求12所述的计算机设备，其特征在于，所述处理器执行所述计算机可读指令时还执行以下步骤：
    根据所述结点序列和所述结点信息提取所述待处理指标数据的多个结点特征参数；及当所述多个结点特征参数均满足预设条件阈值时，确定所述待处理指标数据验证通过。
  15. 根据权利要求12所述的计算机设备，其特征在于，所述结点序列为拓扑序列，所述处理器执行所述计算机可读指令时还执行以下步骤：
    根据所述结点类型和所述结点信息提取结点的结点知识表示；
    根据所述拓扑序列和所述结点知识表示对初始层结点进行图计算，得到对应的第一子图和第一子图值；
    根据所述第一子图和第一子图值按照所述拓扑序列和相应的结点知识表示对下一层结点进行图计算，得到对应的第二子图和第二子图值；
    持续根据得到的当前子图和当前子图值按照所述拓扑序列和相应的结点知识表示对多个结点进行迭代图计算；及
    直到所有结点完成图计算，输出所述待处理指标数据对应的目标指标值。
  16. 根据权利要求15所述的计算机设备，其特征在于，所述结点序列为拓扑序列，所述处理器执行所述计算机可读指令时还执行以下步骤：
    根据所述结点类型和所述结点信息提取结点的一阶谓词逻辑表示；
    利用预设谓词逻辑算法根据所述拓扑序列和所述结点知识表示对初始层结点进行图计算，提取最大目标结点对应的当前子图；及
    计算所述当前子图对应的子图值，利用所述子图值更新所述当前子图，生成所述初始层结点对应的第一子图和第一子图值。
  17. 根据权利要求12所述的计算机设备，其特征在于，所述处理器执行所述计算机可读指令时还执行以下步骤：
    获取历史车辆驾驶数据，对所述历史车辆驾驶数据进行特征提取，得到多个特征向量；
    对所述多个特征向量进行特性分析，得到多个特征指标；
    根据所述多个特征指标构建多个结点结构；及
    根据所述多个结点结构利用预设算法构建数据分析模型。
  18. 根据权利要求12所述的计算机设备，其特征在于，所述处理器执行所述计算机可读指令时还执行以下步骤：
    根据多个任务类型将所述车辆驾驶数据进行分解为多个子任务数据，分别提取多个子任务数据对应的待处理指标数据；所述待处理指标数据包括任务类型；
    将所述待处理指标数据输入至数据分析模型，根据所述任务类型分别将多个待处理指标数据输入至相应的子任务模型；及
    通过所述子任务模型提取所述待处理指标数据对应的结点序列和结点信息，根据所述结点序列和所述结点信息进行图计算，得到所述子任务数据对应的目标指标值。
  19. 根据权利要求12所述的计算机设备，其特征在于，所述处理器执行所述计算机 可读指令时还执行以下步骤：
    根据所述待处理指标数据对应的目标指标值计算所述车辆的驾驶风险评分；及
    当所述驾驶风险评分超过风险阈值时，向监控终端发送预警提示信息。
  20. 一个或多个存储有计算机可读指令的非易失性计算机可读存储介质，所述计算机可读指令被一个或多个处理器执行时，使得所述一个或多个处理器执行权利要求1至10任一项所述的步骤。
PCT/CN2020/073492 2020-01-21 2020-01-21 基于图计算的车辆驾驶数据处理方法、装置和计算机设备 WO2021146910A1 (zh)

Priority Applications (2)

Application Number Priority Date Filing Date Title
PCT/CN2020/073492 WO2021146910A1 (zh) 2020-01-21 2020-01-21 基于图计算的车辆驾驶数据处理方法、装置和计算机设备
CN202080003165.8A CN113677583B (zh) 2020-01-21 2020-01-21 基于图计算的车辆驾驶数据处理方法、装置和计算机设备

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2020/073492 WO2021146910A1 (zh) 2020-01-21 2020-01-21 基于图计算的车辆驾驶数据处理方法、装置和计算机设备

Publications (1)

Publication Number Publication Date
WO2021146910A1 true WO2021146910A1 (zh) 2021-07-29

Family

ID=76991934

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/073492 WO2021146910A1 (zh) 2020-01-21 2020-01-21 基于图计算的车辆驾驶数据处理方法、装置和计算机设备

Country Status (2)

Country Link
CN (1) CN113677583B (zh)
WO (1) WO2021146910A1 (zh)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114464006B (zh) * 2022-04-13 2022-06-28 新石器慧通（北京）科技有限公司 自动驾驶车辆的分配方法和装置

Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20040071069A1 (en) * 2000-06-27 2004-04-15 Fujitsu Limited Apparatus for recording and regenerating data
CN106844947A (zh) * 2017-01-18 2017-06-13 清华大学 一种基于高阶相关学习的机车节能优化自动驾驶方法
US20170185482A1 (en) * 2015-12-29 2017-06-29 Cnex Labs, Inc. Computing system with circular-shift recovery mechanism and method of operation thereof
CN108333959A (zh) * 2018-03-09 2018-07-27 清华大学 一种基于卷积神经网络模型的机车节能操纵方法
CN109835375A (zh) * 2019-01-29 2019-06-04 中国铁道科学研究院集团有限公司通信信号研究所 基于人工智能技术的高速铁路列车自动驾驶系统
CN109886079A (zh) * 2018-12-29 2019-06-14 杭州电子科技大学 一种车辆检测与跟踪方法
CN110329271A (zh) * 2019-06-18 2019-10-15 北京航空航天大学杭州创新研究院 一种基于机器学习的多传感器车辆行驶检测系统及方法
CN110442731A (zh) * 2019-07-24 2019-11-12 中电科新型智慧城市研究院有限公司 一种基于交通管理知识图谱的交通运行管理系统

Family Cites Families (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US8700296B2 (en) * 2006-03-03 2014-04-15 Inrix, Inc. Dynamic prediction of road traffic conditions
JP2018508418A (ja) * 2015-01-20 2018-03-29 ソルフィス リサーチ、インコーポレイテッド リモートセンシング及び車両制御のための実時間マシンビジョン並びに点群分析
EP3828657A1 (en) * 2016-12-23 2021-06-02 Mobileye Vision Technologies Ltd. Navigational system
KR102535540B1 (ko) * 2017-01-12 2023-05-23 모빌아이 비젼 테크놀로지스 엘티디. 차량 움직임에 기반한 항법
US11010658B2 (en) * 2017-12-22 2021-05-18 Intel Corporation System and method for learning the structure of deep convolutional neural networks
CN108984483B (zh) * 2018-07-13 2020-06-09 清华大学 基于dag及矩阵重排的电力系统稀疏矩阵求解方法和系统

Patent Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20040071069A1 (en) * 2000-06-27 2004-04-15 Fujitsu Limited Apparatus for recording and regenerating data
US20170185482A1 (en) * 2015-12-29 2017-06-29 Cnex Labs, Inc. Computing system with circular-shift recovery mechanism and method of operation thereof
CN106844947A (zh) * 2017-01-18 2017-06-13 清华大学 一种基于高阶相关学习的机车节能优化自动驾驶方法
CN108333959A (zh) * 2018-03-09 2018-07-27 清华大学 一种基于卷积神经网络模型的机车节能操纵方法
CN109886079A (zh) * 2018-12-29 2019-06-14 杭州电子科技大学 一种车辆检测与跟踪方法
CN109835375A (zh) * 2019-01-29 2019-06-04 中国铁道科学研究院集团有限公司通信信号研究所 基于人工智能技术的高速铁路列车自动驾驶系统
CN110329271A (zh) * 2019-06-18 2019-10-15 北京航空航天大学杭州创新研究院 一种基于机器学习的多传感器车辆行驶检测系统及方法
CN110442731A (zh) * 2019-07-24 2019-11-12 中电科新型智慧城市研究院有限公司 一种基于交通管理知识图谱的交通运行管理系统

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
WANG LONGDA, LIU WENSHENG, LIU XING, XIANG BOYANG: "A New Calculation Method of Urban Rail Vehicle Operation Diagram", SCIENCE & TECHNOLOGY INFORMATION, no. 5, 1 January 2013 (2013-01-01), pages 157 - 158, XP055831304 *

Also Published As

Publication number Publication date
CN113677583A (zh) 2021-11-19
CN113677583B (zh) 2023-08-15

Similar Documents

Publication Publication Date Title
CN110050257B (zh) 可执行数据流图的差分
US10394770B2 (en) Methods and systems for implementing a data reconciliation framework
Cugola et al. Introducing uncertainty in complex event processing: model, implementation, and validation
US20150186777A1 (en) Automated creation of semantically-enriched diagnosis models
CN108664241B (zh) 一种将SysML模型进行仿真验证的方法
JP2020514916A (ja) 異種製品ライフサイクルデータの同種モデル
US11580009B2 (en) Systems and methods for program code defect and acceptability for use determination
CN112202617B (zh) 资源管理系统监控方法、装置、计算机设备和存储介质
US11347864B2 (en) Ace: assurance, composed and explained
Mason et al. Assurance in reinforcement learning using quantitative verification
WO2021146910A1 (zh) 基于图计算的车辆驾驶数据处理方法、装置和计算机设备
Ramakrishna et al. Automating Pattern Selection for Assurance Case Development for Cyber-Physical Systems
Herzig et al. Bayesian Reasoning Over Models.
CN111596894B (zh) 软件需求抽取方法、装置、计算机设备和可读存储介质
CN109409411B (zh) 基于运维管理的问题定位方法、装置及存储介质
US12007978B2 (en) Validation and recommendation engine
Nazaruks et al. Joint usage of frames and the topological functioning model for domain knowledge presentation and analysis
CN116028042A (zh) 面向随机混成系统的SysML组合建模语言及转换为概率混成自动机的方法
WO2023275765A1 (en) Systems and methods for operating an autonomous system
CN114896168A (zh) 用于自动驾驶算法开发的快速调试系统、方法以及存储器
Kalinin et al. Big Data Security Evaluation by Bidirectional Analysis of Access Control Policy
Bimonte et al. Data-centric UML profile for agroecology applications: Agricultural autonomous robots monitoring case study
Duan et al. A systematic mapping study on the verification of cyber-physical systems
Deshmukh et al. Stochastic temporal logic abstractions: Challenges and opportunities
AU2021287457B2 (en) "Log Data Compliance"

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 20914845

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 20914845

Country of ref document: EP

Kind code of ref document: A1