CN-121999622-A - Intelligent induction method and system for variable information sign networking of expressway

CN121999622ACN 121999622 ACN121999622 ACN 121999622ACN-121999622-A

Abstract

The application relates to the field of intelligent traffic systems, discloses a highway variable information sign networking intelligent induction method and system, and aims to solve the problems of information island, response lag and strategy static existing in the existing induction system. The method comprises the steps of fusing multi-source heterogeneous perception data to construct a high-precision road network traffic state digital twin model, generating a variable information mark group based on a model prediction result and an emergency event instruction, safely issuing a control instruction through a national secret SM4 encryption link, acquiring downstream response data after execution to perform deviation analysis, and triggering closed loop feedback optimization. The system comprises six modules, namely multi-source data access, digital twin modeling, strategy generation, mark grouping management, safety issuing and effect evaluation. According to the application, through space-time accurate prediction, multistage collaborative induction and closed loop self-optimization mechanisms, road network traffic efficiency, emergency response capability and driver decision support level are obviously improved.

Inventors

YUAN DONGLEI
LI NING
LV WEI
ZHENG CHENGGUANG
ZHAO YIBING
FENG CHUANG
CHEN WEIDA
ZHANG HU
LV PEIJIAN
LI JUNFENG
LI TAO
DUAN HONGYONG
DONG YANYING
WEI JINWEN
SHI JIAJI

Assignees

河南交通投资集团有限公司新乡分公司

Dates

Publication Date: 20260508
Application Date: 20260302

Claims (10)

1. The intelligent induction method for the variable information sign networking of the expressway is characterized by comprising the following steps of: acquiring multi-source heterogeneous sensing data covering a target highway network, and performing space-time alignment and deletion compensation on the multi-source heterogeneous sensing data to generate a standardized node traffic flow parameter sequence; Abstracting a road network into a topological directed graph, and constructing a digital twin model for rolling and outputting traffic flow parameter predicted values of all nodes in a preset time window in the future based on a combined architecture of a graph neural network and a long-term and short-term memory network, wherein the graph neural network is used for aggregating space dependence characteristics of the road network topology, and the long-term and short-term memory network is used for extracting time evolution characteristics of a single node; Obtaining predicted situation information output by a digital twin model, and generating an induction strategy set facing to a variable information sign by combining a preset management and control rule base; executing a master-slave logical grouping on a plurality of variable information marks related to collaborative display according to an induction policy set, and establishing an incremental synchronization relation based on content hash check for a master mark and at least one slave mark; Issuing a control instruction to a main mark and an associated auxiliary mark through a secure encryption link, and driving to execute display refreshing; And collecting downstream traffic flow response data after the induction strategy is executed, carrying out sliding window deviation analysis on the downstream traffic flow response data and the strategy expected effect, and triggering parameter fine adjustment of the digital twin model and on-line correction of the induction strategy when the deviation exceeds a preset threshold value to form closed loop feedback control.
2. The intelligent induction method of the highway variable information sign networking according to claim 1, wherein the step of constructing the digital twin model based on the joint architecture of the graphic neural network and the long-term and short-term memory network, further comprises: Inputting a node characteristic matrix in the topological directed graph into a long-term and short-term memory network, and extracting hidden state vectors of all nodes; Calculating attention coefficients between adjacent nodes through a graph attention mechanism, and carrying out weighted aggregation on hidden state vectors of the adjacent nodes according to the attention coefficients; Inputting the aggregated feature vector to a full-connection layer, and outputting predicted values of flow, speed and occupancy rate of a plurality of time steps in the future; in the model training stage, a weighted average absolute percentage error is adopted as a loss function, and a road section importance weight factor in the loss function is determined by weighting and normalizing the historical daily average traffic volume, the accident rate and the connectivity of each road section.
3. The intelligent induction method for the networking of the highway variable information marks according to claim 1, wherein the incremental synchronization relation based on the content hash check is established for the master mark and the at least one slave mark, and the method specifically comprises the following steps: Defining a master mark and a slave mark used for strengthening induction semantics in a downstream path thereof to form a master-slave logic grouping; the main mark controller locally stores the hash value and the update time stamp of the current display content; the method comprises the steps that when a central platform issues an instruction to a main mark, a synchronous task is generated and pushed to a message queue, each slave mark controller subscribes to a synchronous theme of the corresponding main mark, and the consistency of the content hash of a local cache and the current hash value of the main mark is compared; if the verification is inconsistent, the complete display instruction is requested from the mark to the center platform and updating is executed, so that the synchronous delay of the content of the master mark and the slave mark is ensured to be less than 1 second.
4. The method for intelligent induction of variable information sign networking of highways according to claim 1, wherein generating an induction policy set for variable information signs comprises: Under an event response scene, defining an influence range comprising a primary warning area, a secondary shunting area and a tertiary prompting area; And calculating the comprehensive impedance value of each candidate detour path based on the real-time impedance function aiming at the mark in the secondary flow dividing region, and selecting the path with the smallest impedance to generate a detour proposal instruction, wherein the independent variables of the impedance function comprise the path remaining travel time, the lane number change rate and the key node waiting probability.
5. The method for intelligent induction of variable information sign networking of highways according to claim 1, wherein generating an induction policy set for variable information signs further comprises: Aiming at a speed limit sign in the three-level prompt area, obtaining a road section saturation predicted value output by the digital twin model; When the saturation predicted value exceeds a preset threshold value, the dynamic speed limit value is reversely pushed based on the design speed and the saturation deviation, and the calculation result is converted into a speed limit display instruction according to a preset rounding rule.
6. The method of intelligent induction of highway variable information sign networking of claim 1, wherein the step of driving to perform display refresh further comprises: when the instruction type is strong warning information, the primary pivot mark is controlled to flash red characters at the frequency of 2 times per second; and ensures that the visual recognition height of the variable information sign display character is not less than 40 cm.
7. The method of intelligent induction of highway variable information sign networking of claim 1, wherein the step of performing master-slave logical grouping of a plurality of variable information signs related to collaborative display according to an induction policy set further comprises: constructing a global metadata base for maintaining the whole-network variable information marks, wherein the metadata base comprises geographic space coordinates, physical performance parameters and semantic tags of the marks; in response to the geofence selection instruction, a spatial index function is invoked to retrieve all of the tokens that fall within any polygonal fence, and a batch policy binding or unification issuing operation is performed on the retrieved set of tokens.
8. The highway variable information sign networking intelligent induction method according to claim 1, wherein the issuing of the control command to the master sign and its associated slave sign via the secure encrypted link comprises: performing end-to-end encryption on the instruction load by adopting a national encryption SM4 algorithm, and generating a digital signature on the key field of the instruction by utilizing the national encryption SM2 algorithm; after receiving the instruction, the controller firstly checks and decrypts, checks whether the current time is in an executable window formed by the effective time and the dead time carried by the instruction, and can execute display refreshing after the check.
9. The method for intelligent induction of highway variable information sign networking according to claim 1, wherein the sliding window deviation analysis comprises: Setting a sliding window with a fixed length, and continuously scrolling in a step length smaller than the granularity of the predicted time; calculating the real-time average value of the travel time variance reduction rate and the detour path selection proportion in the window when each window is finished; When the continuous multiple windows of the travel time variance reduction rate are lower than a first threshold value or the continuous multiple windows of the bypass path selection proportion are lower than a second threshold value, the induction effect is not expected, and a strategy correction mechanism is automatically triggered.
10. A highway variable information sign networking intelligent guidance system for performing the method of any of claims 1 to 9, comprising: The multi-source traffic perception data access module is used for receiving and standardizing multi-source heterogeneous perception data, and is internally provided with a data quality evaluation unit and a loss compensation unit; The digital twin modeling module is constructed based on a combined architecture of a graph neural network and a long-term and short-term memory network and is used for rolling a space-time prediction sequence of traffic flow parameters of an output road network in a preset period; the strategy generation and grouping module is used for generating a multi-level induction strategy according to the prediction sequence and the emergency event instruction, executing master-slave logic grouping on the associated mark equipment and establishing a content increment synchronization mechanism based on hash value verification; the safety issuing execution module is used for packaging and issuing instructions through a SM4/SM2 cryptographic link, and the controller side comprises an instruction label checking, time window judging and breakpoint continuous transmission functional unit; The closed loop evaluation feedback module is used for collecting downstream response data, monitoring the induction effect in real time through sliding window deviation analysis, and triggering parameter updating of the digital twin modeling module and on-line iterative optimization of the strategy when the induction effect is not expected.

Description

Intelligent induction method and system for variable information sign networking of expressway Technical Field The invention belongs to the field of intelligent traffic systems, and particularly relates to a highway variable information sign networking intelligent induction method and system. Background The invention belongs to the field of intelligent traffic systems, and particularly relates to a highway variable information sign networking intelligent induction method and system. The variable information sign is used as a core carrier for information release of the expressway, and plays a key role in traffic guidance, event early warning and travel service. However, the current expressway induction system generally adopts a single-point independent control mode, lacks a cross-road and cross-region cooperative mechanism, causes information update lag and release content conflict, and is difficult to adapt to the rigid requirement of provincial road network one-network operation on unified control and real-time linkage of information. In view of the above problems, various technical solutions have been proposed in the industry. Patent document with publication number CN115250284A discloses a high-speed operation area safety control system based on Internet of things cloud service, and the speed-down reminding of vehicles on the upstream of an operation area is realized by dividing a plurality of control areas and setting gradient speed limit. However, the speed limit threshold of the scheme depends on static rules, lacks the rolling prediction capability for the global traffic situation, does not relate to cooperative linkage among variable information marks of different road sections, and is difficult to realize dynamic diversion and path level induction in a complex road network environment. In addition, the patent document with publication number CN113075948A relates to tunnel environment safety monitoring and control, and the control object is only local parameters such as brightness, temperature and humidity in the tunnel, and the tunnel environment safety monitoring and control device does not have the regional collaborative induction capability facing the road network level. In summary, the prior art has the following common defects that firstly, traffic state sensing and prediction capability is insufficient, a high-precision short-time prediction means which fuses multi-source data and considers space-time dependence of road network topology is lacked, secondly, variable information sign control logic is isolated, information conflict is induced across equipment, the problem of inconsistent display is highlighted, thirdly, a strategy generation mechanism is stiff, strategy parameters cannot be dynamically adjusted according to real traffic response, effect quantification assessment and self-correction capability is lacked, and fourthly, an effective safety protection mechanism is lacked in an instruction issuing process, and the risk of malicious tampering or counterfeiting of instructions exists. Therefore, there is a need for a method and a system for intelligent induction of variable information sign networking of a highway, which can realize accurate prediction of global traffic situation, coordination of information millisecond consistency of cross-equipment induction, policy closed-loop self-optimization and national security protection capability. Disclosure of Invention The invention provides a highway variable information sign networking intelligent induction method and system, and aims to solve the technical problems of low road network traffic efficiency, insufficient emergency coping capability, weakened decision support of a driver and the like caused by information island, response lag, strategy static state and synergetic deficiency in the conventional highway traffic induction system. The intelligent induction architecture integrating global perception, dynamic modeling, multistage cooperation and closed loop feedback is constructed, so that accurate, real-time and self-adaptive control of the variable information sign of the expressway is realized. According to one aspect of the present invention, there is provided a highway variable information sign networking intelligent induction method, comprising: acquiring multi-source heterogeneous sensing data covering a target highway network, and performing space-time alignment and deletion compensation on the multi-source heterogeneous sensing data to generate a standardized node traffic flow parameter sequence; Abstracting a road network into a topological directed graph, and constructing a digital twin model for rolling and outputting traffic flow parameter predicted values of all nodes in a preset time window in the future based on a combined architecture of a graph neural network and a long-term and short-term memory network, wherein the graph neural network is used for aggregating space dependence characteristics of the road networ