CN-121686600-B - Vehicle dynamic state evaluation method and system integrating multiple sensing information

CN121686600BCN 121686600 BCN121686600 BCN 121686600BCN-121686600-B

Abstract

The invention discloses a vehicle dynamic state evaluation method and system integrating multiple sensing information, and relates to the technical field of vehicle state evaluation, wherein the method comprises the steps of acquiring vehicle CAN data and intelligent driving system data, independently analyzing and integrating analysis, and carrying out safety response decomposition based on analysis data relationship to obtain safety influence factors; the method comprises the steps of carrying out time sequence mode identification based on a safety influence factor, establishing a time sequence trend relation, carrying out state time sequence evaluation through the time sequence trend relation to obtain dynamic risk evaluation information, mapping the vehicle state according to the dynamic risk evaluation information, and generating a vehicle dynamic evaluation result. The method solves the technical problems that in the prior art, vehicle multisource information is difficult to fuse, dynamic risk evolution trend of a vehicle cannot be captured through static evaluation, and accuracy and foresight of vehicle state evaluation are insufficient, and achieves the technical effects of achieving dynamic safety evaluation of the vehicle state and improving the accuracy and foresight of the state evaluation.

Inventors

QIAO YUN
BIAN CHENTONG
ZOU RAN

Assignees

江苏大块头智驾科技有限公司

Dates

Publication Date: 20260512
Application Date: 20260211

Claims (7)

1. A vehicle dynamic state evaluation method integrating multiple sensing information, comprising: Acquiring vehicle CAN data and intelligent driving system data, independently analyzing and fusing the vehicle CAN data and intelligent driving system data, and carrying out safety response decomposition based on analysis data relationship to acquire safety influence factors; Performing time sequence mode identification based on the safety influence factors, and establishing a time sequence trend relation of the safety influence factors; performing state time sequence evaluation on the vehicle CAN data and intelligent driving system data through the time sequence trend relationship to obtain dynamic risk evaluation information; Mapping the vehicle state according to the dynamic risk evaluation information to generate a vehicle dynamic evaluation result; wherein the secure response decomposition is performed based on the parsed data relationships, obtaining a security impact factor comprising: Extracting risk data according to the analysis data relation, and obtaining a plurality of risk events by carrying out cluster analysis on the risk data, wherein each risk event represents a security threat or potential danger scene; performing sensing data relationship tracing based on the analysis data relationship by taking the risk event as a center to obtain a sensing data path; Based on the sensing data path, performing sample fitting on the sensing data of each node, establishing a functional relation between the sensing data of each node and a risk event, performing safety influence factor decomposition, obtaining a minimum cut set of influence factor combinations, and determining the safety influence factors; the method for identifying the time sequence mode based on the safety influence factors establishes the time sequence trend relationship of the safety influence factors and comprises the following steps: Extracting multi-scene security events of independent parameters and fusion parameters based on the security influence factors, and fitting time sequence response relations of the security influence factors to each scene security event to obtain a multi-scene security time sequence model; Based on the multi-scene security event, collecting vehicle operation time sequence data, and constructing a positive example verification set and a negative example verification set; Performing parameter verification on the multi-scene security time sequence model through the positive example verification set and the negative example verification set, and performing time sequence iterative updating on the multi-scene security time sequence model by using verification feedback data; the method for carrying out state time sequence evaluation on the vehicle CAN data and the intelligent driving system data through the time sequence trend relationship to obtain dynamic risk evaluation information comprises the following steps: dividing the vehicle CAN data and intelligent driving system data according to multiple time scales to form a multi-scale data sequence, and associating the data sequence under each time scale with the corresponding real-time gross weight of the vehicle, the historical health data of key components and the current driving scene to form an analysis unit with context information; the analysis units are respectively input into time sequence trend prediction models corresponding to time scales, and risk identification prediction data of each time scale are obtained; And performing time sequence alignment arrangement on the time scale risk identification prediction data, and performing risk response calculation according to the multi-scene security event based on the aligned data to obtain the dynamic risk evaluation information, wherein the dynamic risk evaluation information is a quantized dynamic risk value and comprises structural dynamic risk evaluation information of main risk sources, risk evolution tracks and recommended intervention measures.
2. The method for evaluating the dynamic state of the vehicle by fusing the multi-sensor information according to claim 1, wherein the steps of acquiring the vehicle CAN data and the intelligent driving system data, and independently analyzing and fusing the vehicle CAN data and the intelligent driving system data, comprise: extracting vehicle control sensing data according to a time dimension and a driving scene state dimension based on the vehicle CAN data; acquiring automatic driving module sensing data, environment sensing data, path planning data and vehicle state monitoring data based on the intelligent driving system data; respectively carrying out risk state analysis on the vehicle control sensing data, the automatic driving module sensing data, the environment sensing data, the path planning data and the vehicle state monitoring data, and determining independent analysis risk data; And analyzing the cooperative risk interaction relationship among the vehicle control sensing data, the automatic driving module sensing data, the environment sensing data, the path planning data and the vehicle state monitoring data, and establishing fusion analysis risk data.
3. The vehicle dynamic state evaluation method of fusing multisensory information of claim 2, further comprising: based on the independent analysis risk data, establishing an independent analysis mapping topological relation according to the mapping relation between the risk data and the sensing data; Based on the fusion analysis risk data, adding the fusion risk relationship into the independently analyzed mapping topological relationship according to the mapping relationship between the risk data and the sensing data, and carrying out hierarchical expansion on the mapping topological relationship to obtain the analysis data relationship.
4. The method for evaluating a dynamic state of a vehicle incorporating multiple sensing information according to claim 3, wherein obtaining the analytical data relationship further comprises: and introducing a generated countermeasure network model, dynamically adjusting the mapping weight and the topological structure between the risk data and the sensing data based on the actual behavior feedback and the environmental state change of the vehicle, and optimizing the analysis data relation.
5. The method for evaluating the dynamic state of a vehicle incorporating multiple sensing information according to claim 4, wherein optimizing the analytical data relationship comprises: respectively constructing a generator network and a discriminator network, and forming the generator network and the discriminator network into an countermeasure training frame; Training and converging the countermeasure training frame by using historical sample data to obtain a countermeasure network, wherein weak links in a mapping topology are identified based on data feedback generated by the countermeasure network, wherein simulation data generated by a generator network are injected into an analysis data relationship to perform risk assessment, if the risk assessment confidence is lower than a set risk threshold, the weak links of the current mapping topology are judged to exist when the current mapping topology processes such data patterns, the contribution degree of each sensor data in risk identification is calculated according to the judging accuracy of a discriminator network, and the analysis data relationship is optimized by dynamically adjusting the connection weight in the topology based on the contribution degree.
6. The method for evaluating the dynamic state of a vehicle with fused multi-sensor information according to claim 5, wherein training the challenge training framework with historical sample data converges to obtain a challenge network, comprising: Collecting a sensing data sequence in a normal driving state from historical data as a positive sample, and collecting a sensing data sequence before a known risk event occurs as a negative sample; The vehicle state parameters are used as condition information, the condition information and the data sequence are input into a network, and the generator network and the discriminator network are trained alternately, so that the generator network can generate simulation data which accords with real data distribution and contains potential risk characteristics; And injecting high-quality simulation data generated by the generator into a mapping topology training process, enhancing the recognition capability of the model on the edge cases until reaching a convergence target, and obtaining the countermeasure network.
7. A vehicle dynamic state evaluation system that fuses multisensory information, wherein the system is configured to implement a vehicle dynamic state evaluation method that fuses multisensory information according to any one of claims 1 to 6, the system comprising: The data analysis and decomposition module is used for acquiring vehicle CAN data and intelligent driving system data, independently analyzing and fusing the vehicle CAN data and intelligent driving system data, and carrying out safety response decomposition based on analysis data relationship to acquire safety influence factors; the time sequence pattern recognition module is used for performing time sequence pattern recognition based on the safety influence factors and establishing a time sequence trend relationship of the safety influence factors; the state time sequence evaluation module is used for performing state time sequence evaluation on the vehicle CAN data and the intelligent driving system data through the time sequence trend relationship to obtain dynamic risk evaluation information; And the vehicle state mapping module is used for mapping the vehicle state according to the dynamic risk evaluation information to generate a vehicle dynamic evaluation result.

Description

Vehicle dynamic state evaluation method and system integrating multiple sensing information Technical Field The application relates to the technical field of vehicle state evaluation, in particular to a vehicle dynamic state evaluation method and system integrating multi-sensor information. Background Container transfer trucks, mining dumpers, field bridges, stacking machines and the like in closed or semi-closed scenes such as ports and mines are rapidly developed towards automation and intellectualization, complex environments such as fixed obstacles, mobile equipment, pedestrian mixing and the like exist, severe working conditions such as uneven road surfaces, large dust, variable visibility and the like and the characteristics of high operation intensity and high middle and low speed reciprocating transportation are mainly adopted, traditional vehicle state assessment is mainly dependent on bus data of a vehicle controller local area network or a limited dynamics model, and limitations such as assessment lag and poor forward looking exist when the situation faces complex and changeable actual driving scenes, particularly dynamic working conditions requiring potential risk pre-judgment. It is difficult to fully understand the "intention" of the vehicle and the context of the environment in which it is located only by CAN data, for example, the vehicle CAN data shows that the braking system is normal, but the intelligent driving system senses a front emergency obstacle and has issued an emergency braking request, and the vehicle dynamic response is delayed or deviated, so that the mismatch between the system intention and the vehicle body response may cause significant dynamic risk, collision, overturning or workflow interruption is extremely easy to occur in middle and low speed operations of ports and mines, and the characteristics capable of accurately reflecting the comprehensive dynamic safety state of the vehicle cannot be extracted, the risk accumulation and evolution trend of the vehicle safety state is difficult to capture, and slight oversteer may develop into serious sideslip or tail flicking in several continuous control periods, which affects the accuracy and reliability of vehicle state assessment. Therefore, in the related technology at the present stage, the technical problems that the vehicle multisource information is difficult to fuse, the dynamic risk evolution trend of the vehicle cannot be captured through static evaluation, and the accuracy and the foresight of the vehicle state evaluation are insufficient exist. Disclosure of Invention The application solves the technical problems that in the prior art, vehicle multisource information is difficult to fuse, dynamic risk evolution trend of a vehicle cannot be captured in static evaluation, and accuracy and foresight of vehicle state evaluation are insufficient, and achieves the technical effects of realizing dynamic safety evaluation of vehicle state and improving accuracy and foresight of state evaluation. The application provides a vehicle dynamic state evaluation method integrating multi-sensor information, which comprises the steps of obtaining vehicle CAN data and intelligent driving system data, independently analyzing and integrating the vehicle CAN data and intelligent driving system data, carrying out safety response decomposition based on analysis data relation to obtain safety influence factors, carrying out time sequence pattern recognition based on the safety influence factors, establishing time sequence trend relation of the safety influence factors, carrying out state time sequence evaluation on the vehicle CAN data and intelligent driving system data through the time sequence trend relation to obtain dynamic risk evaluation information, and mapping the vehicle state according to the dynamic risk evaluation information to generate a vehicle dynamic evaluation result. In a possible implementation manner, the vehicle dynamic state evaluation method integrating the multi-sensing information further performs the following processing of extracting vehicle control sensing data according to a time dimension and a driving scene state dimension based on the vehicle CAN data, acquiring automatic driving module sensing data, environment sensing data, path planning data and vehicle state monitoring data based on the intelligent driving system data, respectively performing risk state analysis on the vehicle control sensing data, the automatic driving module sensing data, the environment sensing data, the path planning data and the vehicle state monitoring data to determine independent analysis risk data, analyzing cooperative risk interaction relations among the vehicle control sensing data, the automatic driving module sensing data, the environment sensing data, the path planning data and the vehicle state monitoring data, and establishing integrated analysis risk data. In a possible implementation manner, the vehicle dynamic stat