CN-121982881-A - Traffic event monitoring and handling method, device, equipment and storage medium

CN121982881ACN 121982881 ACN121982881 ACN 121982881ACN-121982881-A

Abstract

The application discloses a method, a device, equipment and a storage medium for monitoring and handling traffic events. The method comprises the steps of constructing a scene state vector based on system multi-source data, inputting the scene state vector into a preset monitoring cognitive model to obtain risk values of various types of events and key scene elements of the events, carrying out early warning on the high-risk events when the risk values of the types of events are larger than a preset threshold value, outputting key scene element prompt information of the high-risk events, generating a feasible treatment strategy candidate set of the high-risk events based on system constraint, carrying out optimization solution by adopting the preset treatment cognitive model to obtain an optimal treatment strategy, decomposing the optimal treatment strategy into specific control instructions, and sending the specific control instructions to relevant equipment for execution. The method and the device can predict the occurrence probability of the event, change the event monitoring from a simple passive response to an active pre-judgment, perform risk early warning, identify key treatment factors and quantify treatment parameters, thereby forming a better treatment scheme.

Inventors

GUO SHENGMIN
YU XINYU
SHAO YANG
ZHANG RUILONG
WANG MINGYU
SUN ZELIN
ZHANG XIN
LI YANG
XIA SHUDONG

Assignees

北京掌行通信息技术有限公司

Dates

Publication Date: 20260505
Application Date: 20251231

Claims (11)

1. A method of monitoring and handling traffic events, comprising: acquiring a scene state vector constructed based on multi-source perception data, and inputting the scene state vector into a monitoring cognitive model preset by a system to obtain risk values of various events and key scene elements of the events; when the risk value of the event of the type is larger than a preset threshold value, carrying out early warning on the event of the high risk, and outputting key scene element prompt information of the event of the high risk type; Generating a feasible treatment strategy candidate set of the risk event under the current scene state vector based on system constraint and a preset knowledge base, and carrying out optimization solution by adopting a preset treatment cognitive model to obtain an optimal treatment strategy; And decomposing the optimal treatment strategy into specific control instructions, and issuing the specific control instructions to relevant equipment for execution.
2. The method according to claim 1, wherein inputting the scene state vector into a monitoring cognitive model preset by a system to obtain risk values of each type of event comprises: Inputting the scene state vector into the monitoring cognitive model to obtain probability vectors of occurrence of various types of events and intensity vectors of occurrence of various types of events; and obtaining the risk value of each type of event based on the product of the probability vector of each type of event and the intensity vector of each type of event.
3. The method of claim 1, wherein deriving key scene elements of an event comprises: calculating an average value of the occurrence probability of the current event in the reference set as a baseline score based on the traffic scene data reference set under the normal working condition and an event occurrence probability prediction function in the monitoring cognitive model; Calculating a contribution value of each feature in the scene state vector by adopting a preset attribution operator to obtain an attribution vector; sorting from large to small based on the absolute value of the contribution value of each feature in the attribution vector, and taking the preset number of features arranged in front as key scene elements of the event; Wherein the sum of all feature contribution values is equal to the difference of the current event occurrence probability prediction value and the baseline score.
4. The method of claim 1, wherein the treatment cognitive model is preconfigured with a constraint condition set, a feasible treatment strategy candidate set of each type of risk event under the current scene vector is generated based on system constraint and a preset knowledge base, and the optimal treatment strategy is obtained by adopting the preset treatment cognitive model for optimization solution, wherein the method comprises the following steps: Based on the constraint condition set and the current scene state vector, obtaining a feasible treatment strategy candidate set under the current scene state vector; inputting the scene state vector and probability vectors of occurrence of various types of events into a treatment cognitive model; Evaluating the effects of different treatment actions in the feasible treatment strategy candidate set to obtain the optimal treatment strategy, wherein the optimal treatment strategy comprises an optimal treatment action and parameters thereof, and the optimal treatment strategy comprises an intervention action before an event and a treatment action after the event; further comprising determining an expected indicator vector of treatment effect and a target aggregate value based on the optimal treatment strategy.
5. The method of claim 4, wherein decomposing the optimal treatment policy into specific control instructions, and issuing to the relevant devices for execution, comprises: decomposing the optimal treatment strategy into specific control instructions, and issuing the specific control instructions to related equipment for execution to generate an auditable work order; Judging whether the optimal treatment action is successfully executed or not, and judging whether an execution effect meets the expected index vector or not; triggering up-going treatment or manual takeover when the execution action fails or the execution effect differs from the expected indicator vector by more than a threshold.
6. The method as recited in claim 1, further comprising: after treatment is completed, acquiring a treatment effect expected index vector and a target aggregation value corresponding to the optimal treatment strategy, and acquiring a treatment effect observation index vector and an actual aggregation value; Constructing a first residual error based on the expected index vector and the observed index vector, and constructing a second residual error based on the target aggregate value and the actual aggregate value; and performing feedforward training and parameter fine adjustment on the monitoring cognitive model and the treatment cognitive model based on the first residual error and the second residual error to obtain an updated monitoring cognitive model and an updated treatment cognitive model.
7. The method as recited in claim 1, further comprising: The knowledge base comprises event monitoring treatment knowledge obtained by the system based on a monitoring cognitive model and a treatment cognitive model, wherein the event monitoring treatment knowledge comprises scene predicates, suggested actions and parameters corresponding to each scene predicate, forbidden actions and parameters, index preference and rule constraint and expected treatment effect parameters; When the treatment cognitive model is subjected to optimization solution, matching a current scene with scene predicates in the knowledge base to obtain a treatment rule applicable to the current scene; After event handling is completed, rules of the knowledge base are updated according to handling effects.
8. The method of claim 4, wherein performing an optimization solution using a treatment cognitive model preset by the system to obtain an optimal treatment strategy comprises: When the monitoring cognitive model monitors that a high-risk event exists, acquiring key scene elements of the high-risk event; And inputting the scene state vector, the probability vector of occurrence of various types of events and the key scene elements of the high-risk event into a treatment cognitive model to obtain the optimal treatment strategy.
9. A traffic event monitoring and handling device, comprising: The event monitoring module is used for acquiring scene state vectors constructed based on the multi-source perception data, inputting the scene state vectors into a monitoring cognitive model preset by the system, and obtaining risk values of various types of events and key scene elements of the events; the event early warning module is used for carrying out early warning on the high-risk event when the risk value of the type event is larger than a preset threshold value, and outputting key scene element prompt information of the high-risk type event; The event handling module is used for generating a feasible handling strategy candidate set of the risk event under the current scene state vector based on system constraint and a preset knowledge base, and carrying out optimization solution by adopting a preset handling cognitive model to obtain an optimal handling strategy; and the control module is used for decomposing the optimal treatment strategy into specific control instructions and issuing the specific control instructions to relevant equipment for execution.
10. An electronic device comprising a processor and a memory storing program instructions, the processor being configured, when executing the program instructions, to perform the method of monitoring and handling traffic events according to any one of claims 1 to 8.
11. A computer readable medium having stored thereon computer readable instructions for execution by a processor to implement the method of monitoring and handling traffic events according to any one of claims 1 to 8.

Description

Traffic event monitoring and handling method, device, equipment and storage medium Technical Field The application relates to the technical field of intelligent transportation, in particular to a method, a device, equipment and a storage medium for monitoring and handling traffic events. Background In the daily management of traffic systems, the prevention and efficient handling of traffic events is of great importance. On one hand, traffic incidents directly relate to road safety and public service quality, if the traffic incidents are improperly handled, the traffic incidents not only can cause large-scale traffic jams, but also can cause secondary accidents, and on the other hand, traffic incidents often cause low traffic efficiency, so that the operation income of a traffic system is influenced. Therefore, how to realize rapid discovery, scientific research and judgment and effective treatment of traffic incidents has become an important topic of road traffic management and operation services. Disclosure of Invention The embodiment of the application provides a method, a device, equipment and a storage medium for monitoring and disposing traffic incidents, which are used for at least solving the technical problem that the occurrence of traffic incidents is difficult to actively predict and effectively dispose in the related technology. According to an aspect of an embodiment of the present application, there is provided a method for monitoring and handling traffic events, including: acquiring a scene state vector constructed based on multi-source perception data, and inputting the scene state vector into a monitoring cognitive model preset by a system to obtain risk values of various events and key scene elements of the events; when the risk value of the event of the type is larger than a preset threshold value, carrying out early warning on the event of the high risk, and outputting key scene element prompt information of the event of the high risk type; Generating a feasible treatment strategy candidate set of the risk event under the current scene state vector based on system constraint and a preset knowledge base, and carrying out optimization solution by adopting a preset treatment cognitive model to obtain an optimal treatment strategy; And decomposing the optimal treatment strategy into specific control instructions, and issuing the specific control instructions to relevant equipment for execution. According to another aspect of the embodiments of the present application, there is also provided a monitoring and handling device for traffic events, including: The event monitoring module is used for acquiring scene state vectors constructed based on the multi-source perception data, inputting the scene state vectors into a monitoring cognitive model preset by the system, and obtaining risk values of various types of events and key scene elements of the events; the event early warning module is used for carrying out early warning on the high-risk event when the risk value of the type event is larger than a preset threshold value, and outputting key scene element prompt information of the high-risk type event; The event handling module is used for generating a feasible handling strategy candidate set of the risk event under the current scene state vector based on system constraint and a preset knowledge base, and carrying out optimization solution by adopting a preset handling cognitive model to obtain an optimal handling strategy; and the control module is used for decomposing the optimal treatment strategy into specific control instructions and issuing the specific control instructions to relevant equipment for execution. According to yet another aspect of the embodiments of the present application, there is also provided an electronic device including a memory, in which a computer program is stored, and a processor configured to execute the above-described traffic event monitoring and handling method by the above-described computer program. According to a further aspect of embodiments of the present application, there is also provided a computer readable storage medium having a computer program stored therein, wherein the computer program is arranged to perform the above-described method of monitoring and handling traffic events when run. The technical scheme provided by the embodiment of the application can have the following beneficial effects: According to the scheme, a scene state vector is constructed based on system multi-source data, the scene state vector is input into a preset monitoring cognitive model, so that risk values of various types of events and key scene elements of the events are obtained, when the risk values of the types of events are larger than a preset threshold value, high-risk event early warning is carried out, and key scene element prompt information of the high-risk type events is output. Through multi-dimension and full life cycle data acquisition and fusion, the limi