Road condition recognition method based on feature recognition and neural network optimization

Abstract

The invention discloses a road condition recognition method based on feature recognition and neural network optimization. The method employs a randomized search method for selection of a feature subset, and achieves the optimization of a neural network structure and parameters and the optimization of a sampling window and an updating window. The method provides by the invention improves the classification precision of a classifier, simplifies the neural network structure and algorithm, carries out the simultaneous optimization of the sizes of the sampling and updating windows, the selection ofthe feature subset and the optimization of the neural network structure and parameters, and increases the mean contour value of the classifier.

Description

technical field

[0001] The invention belongs to the technical field of automation and relates to a road condition recognition method based on feature recognition and neural network optimization. Background technique

[0002] With the energy crisis and environmental pollution, hybrid electric vehicles are getting more and more attention. However, the driving cycle has a great impact on the vehicle's fuel economy, control performance and drivability, coupled with the uncertainty of traffic conditions and changes in road grades, etc., the identification and analysis of driving cycles is facing a huge challenge. It is necessary to identify a set of characteristic parameters to identify driving patterns, but there is no consensus among researchers on the extent to which accessibility and reliability features are used to describe driving patterns. Therefore, it is necessary to study an advanced classifier and driving characteristics to solve these problems. Contents of the inve...

Examples

Embodiment Construction

[0076] Take the real-time driving situation, consider the vehicle speed, and then select 12 statistical features of driving pattern recognition to obtain the characteristics of the driving cycle as an example to further elaborate:

[0077] 1.1 By collecting vehicle speed data in typical driving modes, the distribution of vehicle speed can be obtained. In order to further obtain statistical characteristics according to the vehicle speed distribution and to be able to analyze it quickly, the driving patterns are classified according to the real-time vehicle operation data within the sampling time. Now assume that the given input layer has n I The node of feature vector X, the hidden layer has H nodes, the output layer has a node of discriminant function, the number of training nodes is N, and the input feature vector is set as:

[0078]

[0079] in, is the input training data; nI is the number of input nodes.

[0080] 1.2 Process the output of the i-th hidden node:

[0...

technical field

[0001] The invention belongs to the technical field of automation and relates to a road condition recognition method based on feature recognition and neural network optimization. Background technique

[0002] With the energy crisis and environmental pollution, hybrid electric vehicles are getting more and more attention. However, the driving cycle has a great impact on the vehicle's fuel economy, control performance and drivability, coupled with the uncertainty of traffic conditions and changes in road grades, etc., the identification and analysis of driving cycles is facing a huge challenge. It is necessary to identify a set of characteristic parameters to identify driving patterns, but there is no consensus among researchers on the extent to which accessibility and reliability features are used to describe driving patterns. Therefore, it is necessary to study an advanced classifier and driving characteristics to solve these problems. Contents of the inve...

Examples

Embodiment Construction

[0076] Take the real-time driving situation, consider the vehicle speed, and then select 12 statistical features of driving pattern recognition to obtain the characteristics of the driving cycle as an example to further elaborate:

[0077] 1.1 By collecting vehicle speed data in typical driving modes, the distribution of vehicle speed can be obtained. In order to further obtain statistical characteristics according to the vehicle speed distribution and to be able to analyze it quickly, the driving patterns are classified according to the real-time vehicle operation data within the sampling time. Now assume that the given input layer has n I The node of feature vector X, the hidden layer has H nodes, the output layer has a node of discriminant function, the number of training nodes is N, and the input feature vector is set as:

[0078]

[0079] in, is the input training data; nI is the number of input nodes.

[0080] 1.2 Process the output of the i-th hidden node:

[0...

Images