An Active Collision Avoidance Method for Unmanned Vehicles Based on Forward Vehicle Trajectory Clustering Prediction

Abstract

This invention discloses an active collision avoidance method for unmanned vehicles based on clustering prediction of the preceding vehicle's trajectory. The method includes extracting a set of u-shapelets from the observed signals of the preceding vehicle's movement, calculating the distance between the u-shapelets set and the time series to obtain a distance matrix, and clustering the distance matrix to obtain clustering results. Based on the clustering results, long short-term memory network models are trained using data collected from various onboard sensors to predict the preceding vehicle's trajectory. Based on the predicted preceding vehicle trajectory and the vehicle's own dynamic stability, an active collision avoidance strategy is formulated. This invention, by formulating an active collision avoidance strategy based on the predicted preceding vehicle trajectory and the vehicle's own dynamic stability, includes warning and graded braking to ensure driving safety.

Description

Technical Field

[0001] This invention relates to the field of trajectory prediction technology for unmanned vehicles in open-pit mines, and particularly to an active collision avoidance method for unmanned vehicles based on clustering prediction of the trajectory of the vehicle in front. Background Technology

[0002] Collision avoidance technology is one of the key core technologies for autonomous vehicles. Currently, most longitudinal and lateral collision avoidance strategies assume that the trajectories of surrounding vehicles are known and accurate, rarely considering their uncertainties. Therefore, existing longitudinal and lateral collision avoidance strategies are not suitable for autonomous vehicles. Autonomous vehicles perceive external environmental information in real time through various onboard sensors such as millimeter-wave radar, lidar, and cameras, along with high-precision maps, comprehensively assessing the vehicle's safety situation in the future and activating active collision avoidance strategies in a timely manner to ensure its driving safety. Predicting the future trajectory of the vehicle in front and obtaining the dynamic safe distance between the two vehicles are prerequisites for active collision avoidance in autonomous vehicles. However, due to limitations imposed by various sensor hardware attributes, changes in driving conditions, and external factors such as road environment parameters, a single model for predicting the trajectory of the vehicle in front is insufficient to accurately predict the trajectories of vehicles with different driving characteristics.

[0003] Currently, research on methods for predicting the trajectory of the vehicle ahead includes existing literature on high-speed vehicle motion planning algorithms based on the prediction of the preceding vehicle's trajectory. These algorithms consider motion models and driving intentions when predicting trajectories, but this method is only applicable under relatively ideal conditions. Existing literature also uses unscented Kalman filters to predict the trajectories of surrounding vehicles to ensure real-time generation of safe trajectories, but this method still cannot effectively address complex situations involving the preceding vehicle's motion. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of the existing technology. To achieve the above objective, an active collision avoidance method for unmanned vehicles based on the clustering prediction of the trajectory of the vehicle in front is adopted to solve the problems mentioned in the background technology.

[0005] An active collision avoidance method for unmanned vehicles based on clustering prediction of the trajectory of the vehicle in front includes the following steps:

[0006] Step S1: Extract the u-shapelets set from the observed signal of the preceding vehicle's driving process, calculate the distance between the u-shapelets set and the time series to obtain the distance matrix, and then cluster the distance matrix to obtain the clustering results;

[0007] Step S2: Based on the clustering results, train a long short-term memory network model based on the data collected by various vehicle sensors, and make a prediction to obtain the driving trajectory of the vehicle in front.

[0008] Step S3: Based on the predicted trajectory of the preceding vehicle and the vehicle's own dynamic stability, formulate an active collision avoidance strategy.

[0009] As a further aspect of the present invention: the specific steps for extracting the u-shapelets set in step S1 include:

[0010] Data from various typical road sections were selected to construct a training set for the deep learning model; the original data of the dataset was collected synchronously by lidar, millimeter-wave radar, and cameras, and the original data was output through image processing technology.

[0011] Given a subsequence length slen, obtain all subsequences of length slen from the preceding vehicle driving process monitoring signal dataset D, as a candidate set. Next, calculate the distance sdist between each candidate S and each preceding vehicle driving process monitoring signal T in the preceding vehicle driving process monitoring signal dataset D, using the following formula:

[0012] sdist(T,S)=min(dist(S,Ti));

[0013] In the formula, Ti is a subsequence of length slen in the monitoring signal T of the preceding vehicle's driving process, and i represents the starting position of the subsequence in the monitoring signal T;

[0014] Based on the preset split point d, the time series dataset D is divided into two subsets, DL and DR; the distance data is sorted, and the gap value of the dataset split is calculated based on the midpoint of every two distance data points, where the largest gap value is the optimal split point d.

[0015] As a further aspect of the present invention: the specific steps for obtaining the distance matrix and clustering the distance matrix in step S1 include:

[0016] Based on the obtained u-shapelets set, the distance sdist between each S and all process monitoring signals T in the driving process monitoring signal dataset D is extracted to obtain the distance matrix Dis;

[0017] For each vehicle in front, a subsequence with a sampling point length of 3000 (slen=3000) is extracted from the monitoring data within 3 seconds before the vehicle moves ahead. This subsequence is used as a candidate set of u-shapelets. The set is then extracted, and the distance between each u-shapelet and the process monitoring data in the dataset is calculated to obtain the distance matrix.

[0018] The distance matrix was clustered using DBSCAN to obtain the clustering results.

[0019] The clustering results are divided into several types: Type 1 is when the distance between two vehicles is less than the minimum safe lane change distance; Type 2 is when the distance between two vehicles is greater than or equal to the maximum safe lane change distance; Type 3 is when the distance between two vehicles is between the minimum safe lane change distance and the maximum safe lane change distance.

[0020] By clustering the monitoring signals of each preceding vehicle's driving process with the distance between them and the u-shapelets, the monitoring signals of different driving safety trends are clustered, that is, preceding vehicles with different safety change patterns are clustered to obtain the clustering results.

[0021] As a further aspect of the present invention, the specific steps in step S2 include:

[0022] Step S21: Create a random index for the data, using the first 80% of the data as the training set and the last 20% as the validation set. Then, normalize the training data.

[0023] Step S22: Initialize the hyperparameter set, construct four-dimensional features, use multi-dimensional features as input, use future trajectory information as output, and use various training hyperparameters to build a data-driven model.

[0024] A vehicle trajectory prediction model is built based on LSTM, with process monitoring signals as input. The model contains three LSTM layers, each with 512 nodes, and a Dropout layer is connected to prevent overfitting. After the LSTM layers, there are two fully connected layers and a RUL output layer. The two fully connected layers use the linear rectified function and the hyperbolic tangent function as activation functions, respectively, while the RUL output layer has no activation function.

[0025] Based on the constructed vehicle trajectory prediction model, the SGD optimizer is used to train it. The training loss function is defined as the mean absolute error, and the formula is:

[0026]

[0027] In the formula, This is the theoretical value of RUL; This is the predicted RUL value; The number of training samples in each batch;

[0028] Step S23: Randomly select multiple validation sets from all training data to perform cyclical accuracy validation on the trained model. Using the calculated regression error as the standard, obtain the prediction model with the best accuracy.

[0029] Step S24: Based on historical feature data, perform feature recursion, substitute it into the best accuracy model, input the process monitoring signal into the LSTM-based forward vehicle trajectory prediction model, and predict the future trajectory of the forward vehicle.

[0030] As a further aspect of the present invention, the specific steps of the active collision avoidance strategy in step S3 include:

[0031] Based on the predicted trajectory of the preceding vehicle and the vehicle's own dynamic stability;

[0032] And then, based on the longitudinal distance D measured by the vehicle's sensors between the adjacent leading vehicle in the future period and the active safety warning distance thresholds Dwmin and Dwmax, a comparison is made to determine whether there is a collision risk, and warnings and graded braking are performed;

[0033] When D ≥ Dwmax, that is, the adjacent leading vehicle does not affect the normal driving of the vehicle itself, the active collision avoidance system does not intervene;

[0034] When Dwmin < D < Dwmax, the future driving trajectory of the adjacent leading vehicle will affect the driving safety of the vehicle itself, there is a collision risk, the active collision avoidance system issues a warning, and controls its own vehicle to decelerate with the target deceleration;

[0035] When the on-vehicle sensor measures that the longitudinal distance D of the adjacent leading vehicle ≤ Dwmin, it is necessary to increase the braking intensity of its own vehicle, and at this time, the target deceleration is the maximum braking deceleration under the current road surface adhesion conditions.

[0036] Compared with the prior art, the present invention has the following technical effects:

[0037] By adopting the above technical solution, by extracting the u-shapelets set and calculating the distance matrix between it and the time series for clustering to identify vehicles with different driving safety trends; secondly, based on the clustering results and on-vehicle sensor data, using the long short-term memory network model to train and predict the driving trajectory of the leading vehicle; finally, according to the predicted leading vehicle trajectory and the dynamic stability of its own vehicle, an active collision avoidance strategy is formulated, including warnings and graded braking, to ensure driving safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] The following will describe in detail the specific embodiments of the present invention in conjunction with the accompanying drawings:

[0039] Figure 1 It is a schematic diagram of the steps of the active collision avoidance method for an autonomous vehicle according to the disclosed embodiment of the present application;

[0040] Figure 2 It is a flow block diagram of the active collision avoidance method for an autonomous vehicle according to the disclosed embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0041] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0042] Please refer to Figure 1 and Figure 2 In this embodiment of the invention, an active collision avoidance method for unmanned vehicles based on clustering prediction of the trajectory of the vehicle in front includes the following steps:

[0043] Step S1: Extract the u-shapelets set from the observed signal of the preceding vehicle's driving process, calculate the distance between the u-shapelets set and the time series to obtain the distance matrix, and then cluster the distance matrix to obtain the clustering results;

[0044] Step S11, the specific steps for extracting the u-shapelets set include:

[0045] Data from various typical road sections were selected to build a training set for the deep learning model; the original data of the dataset was collected synchronously by LiDAR, millimeter-wave radar, and cameras, and the original data was output through image processing technology.

[0046] Given a subsequence length slen, obtain all subsequences of length slen from the preceding vehicle driving process monitoring signal dataset D, as a candidate set. Next, calculate the distance sdist between each candidate S and each preceding vehicle driving process monitoring signal T in the preceding vehicle driving process monitoring signal dataset D, using the following formula:

[0047] sdist(T,S)=min(dist(S,Ti));

[0048] In the formula, Ti is a subsequence of length slen in the monitoring signal T of the preceding vehicle's driving process, and i represents the starting position of the subsequence in the monitoring signal T;

[0049] Based on the preset split point d, the time series dataset D is divided into two subsets, DL and DR; the distance data is sorted, and the gap value of the dataset split is calculated based on the midpoint of every two distance data points, where the largest gap value is the optimal split point d.

[0050] Step S12, obtaining the distance matrix and performing clustering on the distance matrix, includes the following specific steps:

[0051] Based on the obtained u-shapelets set, the distance sdist between each S and all process monitoring signals T in the driving process monitoring signal dataset D is extracted to obtain the distance matrix Dis;

[0052] For each vehicle in front, a subsequence with a sampling point length of 3000 (slen=3000) is extracted from the monitoring data within 3 seconds before the vehicle moves ahead. This subsequence is used as a candidate set of u-shapelets. The set is then extracted, and the distance between each u-shapelet and the process monitoring data in the dataset is calculated to obtain the distance matrix.

[0053] Step S13: Perform clustering on the distance matrix using DBSCAN to obtain the clustering results;

[0054] The clustering results are divided into several types: Type 1 is when the distance between two vehicles is less than the minimum safe lane change distance; Type 2 is when the distance between two vehicles is greater than or equal to the maximum safe lane change distance; Type 3 is when the distance between two vehicles is between the minimum safe lane change distance and the maximum safe lane change distance.

[0055] By clustering the monitoring signals of each preceding vehicle's driving process with the distance between them and the u-shapelets, the monitoring signals of different driving safety trends are clustered, that is, preceding vehicles with different safety change patterns are clustered to obtain the clustering results.

[0056] Step S2: Based on the obtained clustering results, train Long Short-Term Memory (LSTM) network models based on the data collected by various vehicle sensors to predict the trajectory of the vehicle in front. Specific steps include:

[0057] Step S21: Create a random index for the data, using the first 80% of the data as the training set and the last 20% as the validation set. Then, normalize the training data.

[0058] Step S22: Initialize the hyperparameter set, construct four-dimensional features, use multi-dimensional features as input, use future trajectory information as output, and use various training hyperparameters to build a data-driven model.

[0059] A vehicle trajectory prediction model is built based on LSTM, with process monitoring signals as input. The model contains three LSTM layers, each with 512 nodes, and a Dropout layer is connected to prevent overfitting. After the LSTM layers, there are two fully connected layers and a RUL output layer. The two fully connected layers use the linear rectified function and the hyperbolic tangent function as activation functions, respectively, while the RUL output layer has no activation function.

[0060] Based on the constructed vehicle trajectory prediction model, the SGD optimizer is used to train it. The training loss function is defined as the mean absolute error, and the formula is:

[0061]

[0062] In the formula, This is the theoretical value of RUL; This is the predicted RUL value; The number of training samples in each batch;

[0063] Step S23: Randomly select multiple validation sets from all training data to perform cyclical accuracy validation on the trained model. Using the calculated regression error as the standard, obtain the prediction model with the best accuracy.

[0064] Step S24: Based on historical feature data, perform feature recursion, substitute the best precision model, and input the process monitoring signal into the front vehicle driving trajectory prediction model based on LSTM to predict the future driving trajectory of the front vehicle.

[0065] Step S3: According to the obtained predicted trajectory of the front vehicle and the dynamic stability of its own vehicle, formulate an active collision avoidance strategy.

[0066] Among them, the specific steps of the active collision avoidance strategy include:

[0067] According to the obtained predicted trajectory of the front vehicle and the dynamic stability of its own vehicle;

[0068] And then, according to the longitudinal distance D between the adjacent front vehicle measured by the vehicle's sensor in the future period and the active safety warning distance thresholds Dwmin and Dwmax, determine whether there is a collision risk, and perform warning and graded braking.

[0069] When D≥Dwmax, that is, the adjacent front vehicle does not affect the normal driving of its own vehicle, the active collision avoidance system does not intervene.

[0070] When Dwmin<D<Dwmax, the future driving trajectory of the adjacent front vehicle will affect the driving safety of its own vehicle, there is a collision risk, the active collision avoidance system gives a warning, and controls its own vehicle to decelerate with the target deceleration.

[0071] When the vehicle-mounted sensor measures that the longitudinal distance D of the adjacent front vehicle ≤ Dwmin, it is necessary to increase the braking intensity of its own vehicle, and the target deceleration is the maximum braking deceleration under the current road surface adhesion condition.

[0072] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents and should be included within the protection scope of the present invention.

Claims

1. A method for active collision avoidance of unmanned vehicles based on clustering prediction of the trajectory of the vehicle in front, characterized in that, Includes the following steps: Step S1: Extract the u-shapelets set from the observed signal of the preceding vehicle's driving process, calculate the distance between the u-shapelets set and the time series to obtain the distance matrix, and then cluster the distance matrix to obtain the clustering results; The specific steps for extracting the u-shapelets set in step S1 include: Data from various typical road sections were selected to build a training set for the deep learning model; the original data in the dataset was collected simultaneously by LiDAR, millimeter-wave radar, and cameras, and the original data was output through image processing technology. Given a subsequence length slen, obtain all subsequences of length slen from the preceding vehicle driving process monitoring signal dataset D, as a candidate set. Next, calculate the distance sdist between each candidate S and each preceding vehicle driving process monitoring signal T in the preceding vehicle driving process monitoring signal dataset D, using the following formula: sdist(T,S)=min(dist(S,Ti)); In the formula, Ti is a subsequence of length slen in the monitoring signal T of the preceding vehicle's driving process, and i represents the starting position of the subsequence in the monitoring signal T; Based on the preset split point d, the time series dataset D is divided into two subsets, DL and DR; the distance data is sorted, and the gap value of the dataset split is calculated based on the midpoint of every two distance data points, where the largest gap value is the optimal split point d; The specific steps for obtaining the distance matrix in step S1 and performing clustering on the distance matrix include: Based on the obtained u-shapelets set, the distance sdist between each S and all process monitoring signals T in the driving process monitoring signal dataset D is extracted to obtain the distance matrix Dis; For each vehicle in front, a subsequence with a sampling point length of 3000 (slen=3000) is extracted from the monitoring data within 3 seconds before the vehicle moves ahead. This subsequence is used as a candidate set of u-shapelets. The set is then extracted, and the distance between each u-shapelet and the process monitoring data in the dataset is calculated to obtain the distance matrix. The distance matrix was clustered using DBSCAN to obtain the clustering results. The clustering results are divided into several types: Type 1 is when the distance between two vehicles is less than the minimum safe lane change distance; Type 2 is when the distance between two vehicles is greater than or equal to the maximum safe lane change distance; Type 3 is when the distance between two vehicles is between the minimum safe lane change distance and the maximum safe lane change distance. By clustering the monitoring signals of each preceding vehicle's driving process with the distance between them and the u-shapelets, the monitoring signals of different driving safety trends are clustered, that is, preceding vehicles with different safety change patterns are clustered to obtain the clustering results; Step S2: Based on the clustering results, train a long short-term memory network model based on the data collected by various vehicle sensors, and make a prediction to obtain the driving trajectory of the vehicle in front. Step S3: Based on the predicted trajectory of the preceding vehicle and the vehicle's own dynamic stability, formulate an active collision avoidance strategy.

2. The active collision avoidance method for unmanned vehicles based on clustering prediction of the trajectory of the vehicle in front, as described in claim 1, is characterized in that... The specific steps in step S2 include: Step S21: Create a random data index, use the first 80% of the data as the training set, and the last 20% of the data as the validation set; then, normalize the training data. Step S22: Initialize the hyperparameter set, construct four-dimensional features, use the multi-dimensional features as the input, and the future trajectory information as the output, and establish a data-driven model using a variety of training hyperparameters. Build a front vehicle driving trajectory prediction model based on LSTM, use the process monitoring signal as the input, the model includes three LSTM layers, with 512 nodes in each layer, and then connect a Dropout layer to prevent the model from overfitting; after the LSTM layer, there are two fully connected layers and a RUL output layer. The two fully connected layers use the rectified linear unit function and the hyperbolic tangent function as the activation functions respectively, and the RUL output layer has no activation function. According to the constructed front vehicle driving trajectory prediction model, use the SGD optimizer to train it, and the loss function of the training is defined as the mean absolute error, and the formula is: In the formula, This is the theoretical value of RUL; This is the predicted RUL value; The number of training samples in each batch; Step S23: Randomly select multiple groups of validation sets from all the training data, and perform cyclic accuracy verification on the training model; take the calculated regression error as the standard to obtain the prediction model with the optimal accuracy. Step S24: Perform feature recursion based on the historical feature data, substitute the model with the best accuracy, input the process monitoring signal into the front vehicle driving trajectory prediction model based on LSTM, and predict the future driving trajectory of the front vehicle.

3. The active collision avoidance method for unmanned vehicles based on clustering prediction of the trajectory of the vehicle in front, as described in claim 1, is characterized in that... The specific steps of the active collision avoidance strategy in step S3 include: According to the predicted trajectory of the front vehicle obtained and the dynamic stability of its own vehicle; and Then, compare the longitudinal distance D between the adjacent front vehicle in the future period measured by the vehicle's sensor with the active safety warning distance thresholds Dwmin and Dwmax to determine whether there is a collision risk, and perform warning and hierarchical braking. When D≥Dwmax, that is, the adjacent front vehicle does not affect the normal driving of its own vehicle, the active collision avoidance system does not intervene. When Dwmin<D<Dwmax, the future driving trajectory of the adjacent front vehicle will affect the driving safety of its own vehicle, there is a collision risk, the active collision avoidance system issues a warning, and controls its own vehicle to decelerate with the target deceleration. When the on-vehicle sensor measures that the longitudinal distance D of the adjacent front vehicle ≤ Dwmin, it is necessary to increase the braking intensity of its own vehicle, and the target deceleration at this time is the maximum braking deceleration under the current road surface adhesion conditions.

Images