Unmanned vehicle reliability early warning method and device and storage medium

Abstract

The invention discloses an unmanned vehicle reliability early warning method and device, and a storage medium. The method comprises the following steps: initializing feature importance; a feature importance set is obtained through calculation based on a random forest algorithm, and correlation analysis is conducted on the nonlinear relation between the features in combination with the maximum mutual information coefficient; sorting the feature importance set to generate a priority sequence; deleting the feature with the lowest priority from the priority sequence, and recalculating the accuracy of the reliability early warning model; if the accuracy is improved, returning to the previous step to continue iteration; if the accuracy is not improved, the deleted features are recovered, and the priorities of the deleted features are improved until all the features are traversed and the priorities are adjusted; and outputting a feature combination with an optimal early warning result, wherein the feature combination is used for constructing a real-time reliability early warning model. The method improves the negative feature deletion efficiency through the priority, does not delete the front operation features, and improves the reliability early warning precision of the unmanned vehicle.

Description

Technical Field

[0001] The present invention relates to the field of unmanned driving technology, and in particular to an unmanned driving reliability early warning method, device and storage medium. Background Art

[0002] Autonomous vehicle reliability warnings monitor and analyze various data and indicators during autonomous vehicle operation, issuing timely alerts before a failure occurs, helping personnel take measures to avoid or minimize the impact of vehicle failures. Clearly, both excessively low and high warning levels hinder the full utilization of autonomous vehicle systems.

[0003] Due to the complexity of the hardware accessed in unmanned driving systems and the frequent topology changes between devices, the reliability topology identification of unmanned driving systems has problems such as many measurement features and low identification accuracy.

[0004] It can be seen that exploring the problem of selecting input features for the reliability warning model of unmanned vehicles, finding a universal feature selection method, and building an operation reliability warning model for unmanned vehicles to improve the accuracy and robustness of the reliability warning of unmanned vehicles are topics worthy of study for technical personnel in this field. Summary of the Invention

[0005] In a first aspect, the present invention provides a reliability warning method for an unmanned vehicle to solve the above-mentioned technical problems. The method comprises: Initialize the feature importance used to build the vehicle operation reliability warning model, where the current operation feature is set to the highest priority and the priorities of other features are initialized to 0; calculate the feature importance set based on the random forest algorithm , and combined with the maximum mutual information coefficient to perform correlation analysis on the nonlinear relationship between features; Importance set of the features Sorting to generate a priority sequence of features; Delete the feature with the lowest priority from the priority sequence and recalculate the accuracy of the reliability warning model; if the accuracy of the reliability warning model improves, return to the previous step and continue iterating; if the accuracy of the reliability warning model does not improve, restore the deleted feature and manually increase the priority of the deleted feature until all features have been traversed and the priority adjustment is completed; The vehicle operation reliability feature combination with the best warning result is output to build the real-time reliability warning model.

[0006] Furthermore, the feature importance set is calculated based on the random forest algorithm , satisfying the formula:

[0007] Where, and Respectively represent the out-of-bag data errors before and after adding noise interference, is the number of decision trees in the hypothetical random forest.

[0008] Furthermore, the calculation formula of the maximum mutual information coefficient is:

[0009] Where, , are two randomly selected running features, , are the measured values of the two operating characteristics, is the preset upper limit of the value. for and The joint distribution probability of , are the marginal distribution probabilities of x and y respectively.

[0010] Furthermore, if the accuracy of the reliability warning model is not improved, the deleted features are restored and the priority of the deleted features is manually increased until all features are traversed and the priority is adjusted, further comprising: The preset threshold of the priority of the deleted feature is increased to ensure that the key reliability features are retained preferentially in subsequent iterations.

[0011] Furthermore, the method of deleting the lowest priority feature from the priority sequence and recalculating the accuracy of the reliability warning model also includes: adopting a backward step-by-step deletion strategy, deleting only one lowest priority feature at a time, and iteratively optimizing the feature combination based on a dynamic priority update mechanism.

[0012] Furthermore, the initialization is used to construct the feature importance of the vehicle operation reliability early warning model, and further includes: Constructing initialization feature variables based on the change values of all vehicle operation data; By using the random forest algorithm and the maximum mutual information coefficient, key reliability features are selected to identify the topological structure of vehicle reliability warning.

[0013] According to a second aspect of the present invention, an electronic device is provided, comprising a memory, a processor, and a computer program stored in the memory, wherein the processor implements the unmanned vehicle reliability warning method when executing the program.

[0014] According to a third aspect of the present invention, a computer-readable storage medium is provided, on which a computer program is stored. When the program is executed by a processor, the unmanned vehicle reliability warning method is implemented.

[0015] Compared with the prior art, the embodiments of the present invention have the following beneficial effects: The present invention studies the selection of input features of the unmanned vehicle reliability warning model, strengthens the operating features that have a positive impact on the vehicle reliability warning, and weakens the operating features that have a negative impact on the vehicle reliability warning; at the same time, adopts a backward step-by-step deletion method to avoid destroying the feature combination; conducts multiple rounds of tentative selection, improves the efficiency of deleting negative features by prioritizing, and does not delete positive operating features, thereby improving the accuracy and robustness of the unmanned vehicle reliability warning. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0017] Figure 1 This is a flowchart of the present invention. DETAILED DESCRIPTION

[0018] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0019] This paper aims to explore the problem of selecting input features for the reliability warning model of unmanned vehicles, in order to study a universal feature selection method and build an operation reliability warning model for unmanned vehicles.

[0020] See also Figure 1 ,A reliability early warning method for an unmanned vehicle ,mainly includes the following steps: Step 1: Initialize the feature importance for building the vehicle operation reliability warning model, where the current operation feature is set to the highest priority and the priorities of other features are initialized to 0; calculate the feature importance set based on the random forest algorithm , and combined with the maximum mutual information coefficient to perform correlation analysis on the nonlinear relationship between features.

[0021] Those skilled in the art further explain that the reliability feature selection of unmanned vehicles is to screen out meaningful and non-redundant feature information from a large number of features, thereby optimizing the real-time warning solution for vehicle operation reliability.

[0022] Due to the performance limitations of on-board sensors and measuring instruments, in order to solve the problems of missing and error in operating data under actual conditions, the random forest algorithm and maximum mutual information coefficient are used to select vehicle reliability features.

[0023] First, construct the initialization feature variables based on the change values of all vehicle operation data, and use all operation attributes as the initial feature set , where the currently running feature is set to the highest level and the priorities of other features are set to 0.

[0024] Then, the random forest method can solve the multi-category classification problem of reliability features, so in all sample features Add noise to calculate the out-of-bag data error. According to the obtained feature importance set Initialize the priority.

[0025] Assume that there is a random forest Decision tree, operation reliability characteristics Importance set Satisfies the formula:

[0026] Where, and They represent the out-of-bag data errors before and after adding noise interference.

[0027] Furthermore, a correlation analysis is performed on the vehicle reliability characteristics to screen out the key features that are important and non-redundant for system reliability identification. Therefore, the maximum mutual information coefficient is introduced to measure the degree of correlation between two operating characteristics and determine whether there is a linear or nonlinear relationship. It has high accuracy. The calculation formula of the maximum mutual information coefficient is:

[0028] Where, , are two randomly selected running features, , are the measured values of the two operating characteristics, is the preset upper limit of the value. for and The joint distribution probability of , are the marginal distribution probabilities of x and y respectively.

[0029] Then, the key reliability features are selected through the random forest algorithm and the maximum mutual information coefficient to identify the topological structure of vehicle reliability warning.

[0030] Step 2: According to the value of feature importance, the feature importance set Sorting is performed to generate a priority sequence of features.

[0031] Step 3: Delete the feature with the lowest priority from the priority sequence and recalculate the accuracy of the reliability warning model. If the accuracy of the reliability warning model is improved, return to the previous step and continue iterating. If the accuracy of the reliability warning model is not improved, restore the deleted feature and manually increase the priority of the deleted feature until all features are traversed and the priority is adjusted.

[0032] In a further solution, the initial feature set is deleted The lowest priority feature , recalculate the accuracy of the vehicle operation reliability warning model. If the accuracy of the reliability warning model is improved, return to step 2; otherwise, withdraw this round of features At the same time, manually add features When all reliability features have been traversed and all feature priorities have been increased, proceed to the next step, otherwise return to step 2.

[0033] Those skilled in the art further explained that, considering that multiple operating characteristics will jointly have a positive impact on the final result of vehicle reliability, while a single operating characteristic will bring a negative impact, a backward step-by-step deletion strategy is adopted, deleting only one lowest priority feature at a time, and iteratively optimizing the feature combination based on a dynamic priority update mechanism.

[0034] Those skilled in the art further explain that increasing the preset threshold of the priority of the deleted feature is used to ensure that key reliability features are preferentially retained in subsequent iterations.

[0035] Step 4: Output the vehicle operation reliability feature combination with the best warning result, which is used to build a real-time reliability warning model to improve the accuracy and robustness of the unmanned vehicle reliability warning model.

[0036] The present invention addresses the problems of excessive measurement features and low recognition accuracy in reliability topology identification of unmanned driving systems. The invention strengthens the operating features that have a positive impact on vehicle reliability warnings and weakens the operating features that have a negative impact on vehicle reliability warnings. Taking into account the impact, a single operating feature will bring negative impacts, so a backward step-by-step deletion method is required to avoid destroying the feature combination. At the same time, when the feature combination contains multiple negative features, deleting a single negative feature may produce a temporary positive impact, so multiple rounds of tentative selection are required to improve the efficiency of negative feature deletion through priority, and do not delete positive operating features.

[0037] In addition, the present invention also protects an electronic device, including a memory, a processor, and a computer program stored in the memory, and when the processor executes the program, the above-mentioned unmanned vehicle reliability warning method is implemented.

[0038] The present invention also protects a computer-readable storage medium having a computer program stored thereon, which implements the above-mentioned unmanned vehicle reliability warning method when the program is executed by a processor.

[0039] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A reliability warning method for unmanned vehicles, characterized in that: The method comprises: Initialize the feature importance used to build the vehicle operation reliability warning model, where the current operation feature is set to the highest priority and the priorities of other features are initialized to 0; calculate the feature importance set based on the random forest algorithm , and combined with the maximum mutual information coefficient to perform correlation analysis on the nonlinear relationship between features; Importance set of the features Sorting to generate a priority sequence of features; Delete the feature with the lowest priority from the priority sequence and recalculate the accuracy of the reliability warning model; if the accuracy of the reliability warning model improves, return to the previous step and continue iterating; if the accuracy of the reliability warning model does not improve, restore the deleted feature and manually increase the priority of the deleted feature until all features have been traversed and the priority adjustment is completed; The vehicle operation reliability feature combination with the best warning result is output to build the real-time reliability warning model.

2. The unmanned vehicle reliability warning method according to claim 1, characterized in that: The feature importance set is calculated based on the random forest algorithm , satisfying the formula: Where, and Respectively represent the out-of-bag data errors before and after adding noise interference, is the number of decision trees in the hypothetical random forest.

3. The unmanned vehicle reliability warning method according to claim 1, characterized in that: The calculation formula of the maximum mutual information coefficient is: Where, , are two randomly selected running features, , are the measured values of the two operating characteristics, is the preset upper limit of the value. for and The joint distribution probability of , are the marginal distribution probabilities of x and y respectively.

4. The unmanned vehicle reliability warning method according to claim 1, characterized in that: If the accuracy of the reliability warning model is not improved, the deleted features are restored and the priorities of the deleted features are manually increased until all features are traversed and the priorities are adjusted, further comprising: The preset threshold of the priority of the deleted feature is increased to ensure that the key reliability features are retained preferentially in subsequent iterations.

5. The unmanned vehicle reliability warning method according to claim 1 or 4, characterized in that: The step of deleting the feature with the lowest priority from the priority sequence and recalculating the accuracy of the reliability early warning model further includes: A backward step-by-step deletion strategy is adopted to delete only one lowest priority feature at a time, and the feature combination is iteratively optimized based on a dynamic priority update mechanism.

6. The unmanned vehicle reliability warning method according to claim 1, characterized in that: The initialization is used to construct the feature importance of the vehicle operation reliability early warning model, and further includes: Constructing initialization feature variables based on the change values of all vehicle operation data; By using the random forest algorithm and the maximum mutual information coefficient, key reliability features are selected to identify the topological structure of vehicle reliability warning.

7. An electronic device comprising a memory, a processor, and a computer program stored in the memory, characterized in that: When the processor executes the program, the unmanned vehicle reliability warning method according to any one of claims 1 to 6 is implemented.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the unmanned vehicle reliability warning method according to any one of claims 1 to 6 is implemented.

Images