CN-121973803-A - Automatic driving decision planning cooperative method and system based on fault prediction

CN121973803ACN 121973803 ACN121973803 ACN 121973803ACN-121973803-A

Abstract

The invention relates to the technical field of automatic driving and discloses an automatic driving decision planning cooperative method and system based on fault prediction, wherein the automatic driving decision planning cooperative method based on fault prediction comprises the steps of collecting gateway operation data, detecting and analyzing time sequence characteristics by abnormal trend, and generating a fault probability prediction curve; the method comprises the steps of acquiring a load trigger signal and dynamic response data, estimating quality parameters on line by a recursive least square algorithm, calculating a centroid position by combining suspension displacement data with a statics equation, fusing to generate a dynamic parameter update package, comparing the gateway fault probability with a threshold value to judge risk, sending the update package with low risk to generate a check code, presynchronizing and storing the snapshot with high risk, updating a coefficient matrix of the dynamic equation to be output to a decision domain and a chassis domain, independently running each domain based on the snapshot when the gateway is in fault, comparing the check consistency of versions and outputting a cooperative identification, and ensuring that each functional domain can still cooperatively run based on the same dynamic parameter standard during the gateway fault.

Inventors

LI BOQI
CHEN PENG
CAO TING
ZHANG JIFU
CHEN CHAO
WANG YANPENG
WANG SHAOZUO

Assignees

北京博易恒华科技有限公司

Dates

Publication Date: 20260505
Application Date: 20260123

Claims (9)

1. An automatic driving decision planning cooperative method based on fault prediction is characterized by comprising the following steps: acquiring running state data of the vehicle-mounted gateway, analyzing time sequence changes of health indexes by using an abnormal trend detection algorithm, and generating a gateway fault probability prediction curve; acquiring a load change event trigger signal and vehicle dynamic response data, and estimating current quality parameters of the vehicle on line by using a recursive least square algorithm; acquiring suspension displacement sensor data, calculating a centroid position estimated value by using a statics balance equation, and combining a mass parameter and a centroid position to generate a dynamic parameter update package; the method comprises the steps of obtaining parameter version identifiers of controllers of all domains, comparing a predicted value of gateway fault probability with a threshold value, and judging a gateway risk state; under the gateway low risk state, a dynamic parameter update package is sent to each domain controller through the gateway to generate a cross-domain parameter consistency check code; based on the updated dynamic parameters, recalculating a dynamic equation coefficient matrix, and outputting the dynamic equation coefficient matrix to a decision domain and a chassis domain; When the gateway fails, each domain independently operates based on the local parameter snapshot, and the parameter consistency is checked by using the version identification comparison algorithm, so as to output the collaborative state identification.
2. The automatic driving decision planning collaboration method based on fault prediction according to claim 1, wherein the acquiring the running state data of the vehicle-mounted gateway, analyzing the time sequence change of the health index by using an abnormal trend detection algorithm, and generating a gateway fault probability prediction curve comprises: Acquiring processing delay data, packet loss rate data and temperature index data of a vehicle-mounted gateway in a current time window; respectively extracting trend characteristics from the processing delay data, the packet loss rate data and the temperature index data; weighting, fusing and calculating trend characteristics of each health index to generate a comprehensive fault probability prediction value; and arranging the comprehensive fault probability prediction values according to a time sequence to form a gateway fault probability prediction curve for representing the probability distribution of faults occurring in a network in a preset time period in the future.
3. The collaborative method for automatic driving decision planning based on failure prediction according to claim 1, wherein the obtaining the load change event trigger signal and the vehicle dynamic response data and estimating the current quality parameter of the vehicle on line using a recursive least square algorithm comprises: Acquiring a load change event trigger signal, wherein the load change event trigger signal comprises a vehicle door opening and closing state signal and a cargo space sensor state change signal; After a load change event is detected, acceleration response data and driving moment data of the vehicle in a preset time period before and after the event are obtained; establishing a longitudinal dynamics equation of the vehicle based on Newton's second law, and inputting acceleration response data and driving moment data into a recursive least square algorithm to perform parameter identification; the recursive least square algorithm adopts a forgetting factor form, and the weight proportion of the historical data and the current data is adjusted through the forgetting factor to output the estimated value of the current quality parameter of the vehicle.
4. The automated driving decision planning cooperative method based on failure prediction according to claim 1, wherein the acquiring suspension displacement sensor data, calculating a centroid position estimation value using a statics balance equation, and combining a mass parameter and a centroid position to generate a dynamic parameter update package comprises: acquiring displacement data of each wheel acquired by a suspension displacement sensor, wherein the displacement data comprise a front axle left wheel displacement value, a front axle right wheel displacement value, a rear axle left wheel displacement value and a rear axle right wheel displacement value; Inputting displacement data of each wheel and the rigidity parameter of the suspension into a static equilibrium equation, and calculating the vertical load born by each wheel; solving position coordinate estimated values of the mass center in the longitudinal direction and the transverse direction of the vehicle based on the vertical load of each wheel and the geometric parameters of the vehicle; and merging and packaging the quality parameter estimation value and the centroid position estimation value to generate a dynamic parameter update package containing the time stamp and the version identifier.
5. The automatic driving decision planning collaboration method based on fault prediction according to claim 1, wherein the performing parameter presynchronization and sending parameter snapshot storage instructions to each domain controller in the gateway high risk state comprises: judging whether the gateway is still in an available state at present; if the gateway is still available, sending a dynamic parameter update packet to the perception domain controller, the decision domain controller and the chassis domain controller through the gateway; Sending a parameter snapshot storage instruction to each domain controller; After receiving the storage instruction, each domain controller stores the current latest parameter version and the version identification into a local nonvolatile memory to form a parameter snapshot.
6. The automatic driving decision planning collaboration method based on the fault prediction according to claim 1, wherein the recalculating the coefficient matrix of the kinetic equation based on the updated kinetic parameters and outputting the coefficient matrix to the decision domain and the chassis domain comprises: Recalculating a coefficient matrix of the vehicle dynamics equation based on the updated mass parameter and centroid position parameter; inputting the updated coefficient matrix into a track planning module of a decision domain, wherein the track planning module reconstructs constraint conditions of track optimization based on the new coefficient matrix, and the constraint conditions comprise acceleration constraint, steering constraint and stability constraint; outputting a parameter update notification to a chassis domain controller, wherein the chassis domain controller adjusts a control gain parameter based on the new dynamic parameter; and generating a parameter transition zone bit, wherein a decision domain adopts a conservative track planning strategy during parameter transition, and clearing the parameter transition zone bit after the chassis domain confirms that the control gain adjustment is completed.
7. The automatic driving decision planning collaboration method based on fault prediction according to claim 1, wherein when the gateway fails, each domain independently operates based on a local parameter snapshot, and verifies parameter consistency by using a version identification comparison algorithm, and outputs a collaboration state identification, including: when detecting that the gateway breaks down to cause the cross-domain communication to be interrupted, switching each domain controller to an independent operation mode; Each domain controller reads a locally stored parameter snapshot and continues to execute respective functional tasks based on dynamic parameters in the snapshot; The decision domain controller sends a version identification inquiry request to the chassis domain controller through a standby communication link; The chassis domain controller returns the locally stored parameter version identification; the decision domain controller executes a version identification comparison algorithm to compare the version identification of the local domain with the version identification of the chassis domain; and if the version identifiers are inconsistent, outputting a parameter version inconsistent warning identifier and triggering a degraded operation strategy.
8. The automated driving decision planning co-ordination method based on fault prediction of claim 7, wherein the degraded operational policy comprises: The decision domain adopts a conservative track planning boundary, and the safety margin is increased to compensate the execution deviation caused by inconsistent parameters.
9. An automatic driving decision planning collaboration system based on failure prediction for executing the automatic driving decision planning collaboration method based on failure prediction as claimed in any one of claims 1 to 8, comprising: The gateway state monitoring module is used for acquiring the running state data of the vehicle-mounted gateway, analyzing the time sequence change of the health index by using an abnormal trend detection algorithm and generating a gateway fault probability prediction curve; The dynamic parameter estimation module is used for acquiring a load change event trigger signal and vehicle dynamic response data, estimating the current mass parameter of the vehicle on line by using a recursive least square algorithm, calculating a centroid position estimation value by using a static equilibrium equation, and generating a dynamic parameter update package; the risk judging module is used for acquiring parameter version identifiers of the controllers of each domain, comparing the predicted value of the gateway fault probability with a threshold value and judging the gateway risk state; The parameter synchronization module is used for sending a dynamic parameter update packet to each domain controller through the gateway under the gateway low risk state, generating a cross-domain parameter consistency check code, executing parameter presynchronization under the gateway high risk state and sending a parameter snapshot storage instruction to each domain controller; The planning control updating module is used for recalculating a kinetic equation coefficient matrix based on the updated kinetic parameters and outputting the kinetic equation coefficient matrix to the decision domain and the chassis domain; and the cooperative operation module is used for controlling each domain to independently operate based on the local parameter snapshot when the gateway fails, checking the parameter consistency by using the version identification comparison algorithm and outputting the cooperative state identification.

Description

Automatic driving decision planning cooperative method and system based on fault prediction Technical Field The invention relates to the technical field of automatic driving, in particular to an automatic driving decision planning cooperative method and system based on fault prediction. Background Domain controller architectures are commonly adopted for automatic driving taxis or freight vehicles, and each functional domain (perception domain, decision domain and chassis domain) performs data exchange through a vehicle-mounted gateway. The vehicle-mounted gateway serves as a cross-domain communication hub and bears the key task of dynamic parameter synchronization among all functional domains. In the operation process of the vehicle, passengers get on or off the vehicle or load and unload cargoes can cause the load of the vehicle to change, new dynamic parameters (including the mass and the mass center position of the vehicle) need to be estimated online at the moment, and updated parameters are synchronized to each functional domain through a gateway so as to ensure that a decision domain and a chassis domain use consistent parameter references when track planning and control are performed. The prior art has the technical problem that after the load of the vehicle changes, the updated dynamic parameters need to be synchronized to each functional domain through a gateway. If the gateway fails or performance is degraded in the parameter synchronization process, a part of functional domains may acquire new parameters while other functional domains still use old parameters, so as to form a state with inconsistent parameter versions. Such inconsistencies may cause the decision domain to program the trajectory based on the new quality parameters, while the chassis domain performs control based on the old parameters, resulting in systematic program-execution bias. In the prior art, gateway fault processing and dynamic parameter updating are treated as independent problems, and the cooperative consideration of gateway fault time and parameter synchronization process is lacked, so that potential threat to parameter synchronization cannot be predicted in the gateway performance degradation stage. Disclosure of Invention The invention provides an automatic driving decision planning cooperative method and system based on fault prediction, which solve the technical problem that the gateway fault time is uncontrollable in the related technology. The invention provides an automatic driving decision planning cooperative method based on fault prediction, which comprises the following steps: acquiring running state data of the vehicle-mounted gateway, analyzing time sequence changes of health indexes by using an abnormal trend detection algorithm, and generating a gateway fault probability prediction curve; acquiring a load change event trigger signal and vehicle dynamic response data, and estimating current quality parameters of the vehicle on line by using a recursive least square algorithm; acquiring suspension displacement sensor data, calculating a centroid position estimated value by using a statics balance equation, and combining a mass parameter and a centroid position to generate a dynamic parameter update package; the method comprises the steps of obtaining parameter version identifiers of controllers of all domains, comparing a predicted value of gateway fault probability with a threshold value, and judging a gateway risk state; under the gateway low risk state, a dynamic parameter update package is sent to each domain controller through the gateway to generate a cross-domain parameter consistency check code; based on the updated dynamic parameters, recalculating a dynamic equation coefficient matrix, and outputting the dynamic equation coefficient matrix to a decision domain and a chassis domain; When the gateway fails, each domain independently operates based on the local parameter snapshot, and the parameter consistency is checked by using the version identification comparison algorithm, so as to output the collaborative state identification. Further, the acquiring the running state data of the vehicle-mounted gateway, analyzing the time sequence change of the health index by using an abnormal trend detection algorithm, and generating a gateway fault probability prediction curve comprises the following steps: Acquiring processing delay data, packet loss rate data and temperature index data of a vehicle-mounted gateway in a current time window; respectively extracting trend characteristics from the processing delay data, the packet loss rate data and the temperature index data; weighting, fusing and calculating trend characteristics of each health index to generate a comprehensive fault probability prediction value; and arranging the comprehensive fault probability prediction values according to a time sequence to form a gateway fault probability prediction curve for representing the probability distribution of