Research on Short-Term Driver Following Habits Based on GA-BP Neural Network

Round 1

Reviewer 1 Report

This paper is an interesting research about driver following habits using the data of next generation simulation based on GA-BP neural network.The methodology is explained well and the structure is clear. 

Here are some of my comments:

1. The values and words in Figure 3 - Figure 5 are too small to read.

2. In line 81, there seems a typo in “headway, headway,”.

3. This paper selected main vehicle speed, main vehicle acceleration, headway, front vehicle speed and front vehicle acceleration for analysis. Did the author also try to use any other parameters that can be used to improve the model? 

4. Simulation scenario seems a very simple case. How will it perform in some edge case such as collison?

5. There are only 15 references. Does the author review enough paper to show this study is meaningful? Is there any similar or better method already studied?

My recommendation is minor revision to address the concerns above.

Author Response

 Thank you very much for your valuable advice. We are sorry that the size of the picture you mentioned in the article is not suitable, which affects reading. We have enlarged the picture to a proper size and modified the font to make it easier to read.The modified picture is as follows:

Reviewer 2 Report

1.The paper analyzes car-following habits, but the results presented in the paper do not reflect explicit content about habits. Generally speaking, habit refers to the individual differences of drivers. The data set used in this paper cannot obtain relevant information of drivers, and each person only has tens of seconds of following data. How to reflect the habit needs to be explained more clearly.

2.The vehicles described in the abstract are not clear, what are vehicle 32 and vehicle 29? No relevant content was found in the text.

3.Compared with the TTC model described in the abstract, there are two problems. The first is that TTC is not an important indicator for describing car-following behavior, and the second is that there is no content about the TTC model in the text.

Author Response

Thank you very much for your attention and comments on our paper. We have revised the manuscript according to your kind and detailed suggestions.

Point 1: The paper analyzes car-following habits, but the results presented in the paper do not reflect explicit content about habits. Generally speaking, habit refers to the individual differences of drivers. The data set used in this paper cannot obtain relevant information of drivers, and each person only has tens of seconds of following data. How to reflect the habit needs to be explained more clearly.

 

Response 1: Thank you very much for your valuable advice,and thanks to your comments, we can improve this assumption and scheme. We are sorry that we have not explained the habit of following cars in detail. We have added an accurate explanation of the habit in the text, which is as follows:

Assumption that environmental factors remain unchanged for a short time：

The definition of car following habits in this paper is somewhat special. First of all, we recognize that there may be large differences in driving habits between different drivers (for example, radical and conservative). Therefore, many papers will classify drivers based on drivers and default that the habits of a certain driver will remain unchanged[20]. In fact, it is not.

Real driving experience tells us that the driver's behavior is not only determined by the several parameters mentioned in the text, but also influenced by numerous factors. For example, when a large vehicle is running side by side in the right lane, the driver's choice will tend to accelerate to overtake the vehicle in front rather than follow as usual. In this case, the driver's driving habits will change with the environment, and the influencing factors cannot be enumerated one by one. Then this paper puts forward a hypothesis: we assume that the decision-making environment(Environmental factors) faced by the driver will not change in a

short time. As shown in the figure, the driver’s driving state will remain basically unchanged before the first emergency occurs.

This paper only studies in a short-term state (that is, the default environmental elements do not change), so an additional scheme is proposed to make the model theoretically applicable to the real scene. The scheme is as follows:

(1)Maintain data collection during the whole process of driver driving the vehicle, and this data should at least include the five main parameters mentioned in the paper;

(2)Using the latest data collected, the model is constantly updated over time (blue cube in Fig. 10);

(3)The model used in the automatic car following function is the green cube in Figure 10. This model will be updated in the following three cases:

①Automatic update with a cycle of 5 unit times；

②Driver's active intervention (automatic car following function is on)；

③When the automatic car following function is not enabled, the system will also use the model (green cube) to predict the car following status in real time and compare it with the real status of the vehicle. When the error is greater than the threshold, the model will be updated.(If the vehicle leaves the following state for 5 seconds, the system stops)

(4)The automatic updating speed of the model is determined by the computing power of the vehicle VCU

Point 2: The vehicles described in the abstract are not clear, what are vehicle 32 and vehicle 29? No relevant content was found in the text.

 Response 2: We are sorry that the unclear expression here has affected your reading. The revised sentence is as follows:

Based on the data of Next Generation Simulation (NGSIM), this paper selects vehicle 32(32 represents the ID of the vehicle in the ngsim project) as the main vehicle to study short-term driving habits.

 

 

Point 3: Compared with the TTC model described in the abstract, there are two problems. The first is that TTC is not an important indicator for describing car-following behavior, and the second is that there is no content about the TTC model in the text.

 Response 3: Thank you very much for your valuable advice, Under your suggestion, we went to consult scholars in relevant directions and learned that TTC is indeed not an important indicator of car following behavior, so we made modifications,and change the comparison object from TTC model to traditional stimulus response model,and select acceleration, speed and vehicle spacing as important indicators. Content is as follows:

Because this paper is based on the stimulus response model, the following habits are defined as the driving strategy (acceleration, deceleration or maintenance) selected by a driver when facing a specific scene (event), and the intensity of acceleration and deceleration. Therefore, acceleration and speed are selected as the main indicators, and vehicle distance parameters are selected to observe the safety situation.

 

Point 4: Extensive editing of English language and style required.

Response 3: Under your suggestion, we have improved the grammar, words and punctuation of the article.

Finally, thank you for reviewing my paper in your busy schedule. I benefited a lot from your suggestion.I sincerely wish you all the best!

Author Response File: Author Response.docx

Reviewer 3 Report

The weakness of this paper is the lack of explaining the importance of this work and also the lack of reflection on the outcomes. Research needs contextualizing. The methodology is contextualized the significance is not. The process adopted is clearly explained. Also an indication of future areas of research and their methodologies would have helped. A clearer justification of the sample and its limitations would have been helpful. Do regional differences need to be considered? Are there factor of which you do not have data that you believe could have added to the insight from the study? Has this research implications beyond the specific focus of the study? 

Author Response

Thank you very much for your attention and comments on our paper. We have revised the manuscript according to your kind and detailed suggestions.

Point 1: The weakness of this paper is the lack of explaining the importance of this work and also the lack of reflection on the outcomes. Research needs contextualizing. The methodology is contextualized the significance is not.

Response 1: Thank you very much for your valuable advice.We are sorry that the introduction is not detailed and sufficient enough to explain the importance of the study. We have improved it to further improve the logic of the introduction to explain the importance of the article.

 

In order to facilitate your review, I would like to briefly describe the overall idea here:

• In the existing commercial vehicle ADAS system, most of the car following functions only have standardized functions, lacking personalized satisfaction.
• After consulting many literatures, we found that although most of them considered the driver's factors, they basically defined the human characteristics as unchanged.
• The assumption that the short-term environmental factors are unchanged is proposed, and the corresponding scheme is designed to solve its application problem.
• Based on the stimulus response model, GA-BP neural network is used to optimize it to form a new model to solve the problem of learning the driver's habit of following the car under the condition that the short-term environmental factors are unchanged.

 

The improved contents are as follows:

1. Introduction

The statistical data of road traffic accidents show that the accidents for which drivers are responsible account for a large proportion of the total number of accidents.[1]With the advancement of global intellectualization, the intellectualization of vehicles will also effectively reduce the accident rate [2]. Modern intelligent driving system obtains environmental information through advanced radar, camera and other sensors, and realizes assisted driving or replaces human driving through the artificial intelligence carried by the high-power vehicle specification intelligent driving chip. Advanced driver assistance system is developing rapidly. From AEB in the early days to auto pilot today, people are no longer satisfied with passive safety technology [3], but continue to climb the level of active safety technology. The concept of vehicle road collaboration [4] has long been proposed, but it has not been realized for a long time due to various restrictions. Now, the birth of 5g network makes it possible. Nowadays, the mass-produced advanced driver assistance system (ADAS) [5] generally does not meet the personalized needs. To meet the personalized needs of drivers, it is necessary to study their driving habits and establish driving habits model analysis rules, so as to further improve the function of ADAS.

Figure 1. Development history of car following model[8].

As one of the most common scenes in the process of vehicle driving, domestic and foreign scholars have done a lot of research on it. Car following model has always been the research hotspot of traffic flow theory. With the development of assisted driving and autonomous vehicle, car following model, as one of driver models, has been widely used in the field of intelligent vehicles.The concept of car following model was first proposed by foreign scholar reuschel. He believes that the expected distance between cars behind is linear with the speed of cars behind [6][7]. Literature [8] divides car following models into five categories: data-driven models, safe distance models, stimulus response models, physiological and psychological models, and cellular automata models. As shown in Figure 1.

BP (backpropagation) neural network is a concept proposed by scientists led by Rumelhart and McClelland in 1986. It is widely used in data-driven models.It is a multilayer feedforward neural network trained according to the error back propagation algorithm, and it is one of the most widely used neural network models. It can be used as a mapping equation [9] without confirming the relationship between input and output in advance. Only by learning the law of a given training set, the output is close to the target value when a given input value is achieved. Its basic principle includes two parts: the forward propagation of the signal, the input signal is finally output through each node of each hidden layer; The backward propagation of error is gradually adjusted from the back end to the front end by comparing the error between the output value and the expected value, and the weight of each layer is adjusted through the error of each layer, which is reciprocating, so that the error decreases in a gradient way, which is most in line with the requirements [10].

As early as 1953, the Pipes model provided a model to maintain a dynamic safe following distance with the vehicle in front [15]. With the proposal of GM model and Herman model, the car following model theory has been further improved. GM model takes into account the influence of relative speed on the acceleration of the rear vehicle, while Herman model takes into account the common influence of multiple vehicles in front.

The safe distance car following model simulates the driving characteristics of the driver to a certain extent, with strong logic and explanation, but it does not consider the structured mathematical model, which may cause large errors in describing the driving behavior and is difficult to simulate the nonlinearity and uncertainty of the driving process[11]; The model based on data-driven can simulate different driving habits and predict different drivers' car following behavior under different environments by fitting the input characteristic parameters, but it is easy to ignore the state relationship between internal characteristic parameters[12][13], and the key feature mining is not perfect. At the same time, a priori theory related to driver characteristics is required, which is prone to over fitting problems.Therefore, this paper proposes a short-term assumption that environmental factors are invariant, which is used to solve the problem that environmental factors are not exhaustive. At the same time, the physical model and the data-driven model are integrated, that is, based on the stimulus response model, the BP neural network is the core, and the GA algorithm is further used to optimize the network. Finally, the model is validated by ngsim data.

Based on the stimulus response model, BP neural network is used to learn driving habits in order to optimize the traditional stimulus response car following model and increase the personalized driving needs of drivers. As a data-driven model, BP neural network needs a large amount of data as support. In this paper, ngsim data is used for research to analyze drivers' driving habits and their influencing factors. Based on BP neural network optimized by genetic algorithm, a car following model with drivers' driving habits is established. Compared with the traditional safe distance model, the model has better habit learning and prediction ability through experiments,and it can greatly enhance the human-computer interaction experience of the automatic vehicle following function. 

Point 2: Also an indication of future areas of research and their methodologies would have helped.

Response 2: Thank you very much for your valuable advice.After reflection, we find that the article really lacks descriptions of the help in this research field in the future.Therefore, we added the help of this study for future research in this field to the conclusion.

The improved contents are as follows:

 

5. conclusion

In this paper, a short-term environmental factor invariant assumption is proposed and an application scheme is designed.Based on the ngsim data, through the correlation analysis of Pearson correlation coefficient, Kendall correlation coefficient and Spearman c-correlation coefficient of the selected data segment, the parameters related to the car following strategy are obtained, and the stimulus response model with the highest matching degree with the driver is selected for research. The model is trained and adjusted by ngsim data, and an improved car following model is established by using BP neural network based on genetic algorithm. The results show that the improved stimulus response model can truly reflect the driver's intention and ensure safety.

However, this paper still has some shortcomings.The model data in this paper comes from NGSIM, so the data duration of a driver is short, often only a few tens of seconds, which makes the simulation of the application scheme impossible. In the follow-up study of this topic, we may try to use real vehicles for testing.In addition, ngsim data are all from the United States, and whether there are obvious differences between American drivers and Chinese drivers needs further research and discussion, which should also be considered in the follow-up real vehicle experiments.Although the assumptions and schemes proposed in this paper have not been verified in the real vehicle verification, they can still give other researchers some new ideas. At the same time, it solves the problem that the driver characteristics change with time in the car following problem to a certain extent

Point 3: A clearer justification of the sample and its limitations would have been helpful. Do regional differences need to be considered?

 

Response 3: Thank you very much for your valuable advice.Under your guidance, we found that this article really lacks a description of the limitations.As for regional differences, because this paper is a pure theoretical study and does not involve the experimental steps that need to invite Chinese drivers, regional differences can be ignored for the time being.Therefore, we improved the conclusion and added the content about the limitations of this article.The improved contents are as follows:

However, this paper still has some shortcomings.The model data in this paper comes from NGSIM, so the data duration of a driver is short, often only a few tens of seconds, which makes the simulation of the application scheme impossible. In the follow-up study of this topic, we may try to use real vehicles for testing.In addition, ngsim data are all from the United States, and whether there are  differences between American drivers and Chinese drivers in this paper can be ignored for the time being, which should  be considered in the follow-up real vehicle experiments.Although the assumptions and schemes proposed in this paper have not been verified in the real vehicle verification, they can still give other researchers some new ideas. At the same time, it solves the problem that the driver characteristics change with time in the car following problem to a certain extent.

Point 4: Are there factor of which you do not have data that you believe could have added to the insight from the study? Has this research implications beyond the specific focus of the study?

Response 3: We are sorry that you have this question due to our unclear expression.In fact, in addition to this article, we have conducted preliminary experimental screening for many parameters. Because the experiment is too repetitive and boring, we think that writing it into the paper will lead to lengthy length. However, your questions has given us a hint that we should describe this process in short sentences, so that readers can understand that we have conducted preliminary screening before selecting these parameters, so as to avoid such problems from happening again.

All the research in this paper is based on the assumption that the short-term environmental elements are unchanged. The impact of this assumption on such research may even be greater than the experiment itself in this paper. However, due to the restrictions of conditions, it is temporarily impossible to carry out the real vehicle test. However, we are also willing to carry out the next real vehicle verification in future papers, hoping to prove that my scheme is effective.

 

The improved contents are as follows:

2.3 Correlation Analysis

The actual ngsim data set contains many parameters, but for the experiment, it is necessary to select the parameters with greater influence. Therefore, the author conducted preliminary experiments on many parameters before. Because of the large amount of experiments and the steps are too repetitive, the experimental process is not shown here, and six parameters with greater influence are selected.

In this paper, Pearson correlation coefficient, Kendall correlation coefficient and Spearman correlation coefficient are mainly used for correlation analysis [16][17]. Pearson correlation coefficient, as a coefficient reflecting the linear relationship between two variables, is mainly used in this paper to analyze whether there is a linear relationship between the six variables screened through empirical analysis. If there is a strong correlation between them, it will reduce the training effect and efficiency of subsequent neural networks. Therefore, it is necessary to compare the Pearson correlation coefficient to reveal the variables with strong correlation. Since the data operation in this paper is based on the MATLAB platform, the Pearson coefficient correlation function provided by Matlab is also used for analysis, and the results shown in Figure 3 below are obtained.

Finally, thank you for reviewing my paper in your busy schedule. I benefited a lot from your suggestion.I sincerely wish you all the best!

Author Response File: Author Response.docx

Round 2

Reviewer 2 Report

The data unit in the paper is recommended to be unified.

It is recommended to refer to other papers for the writing of NG data names, and the names in all lowercase have not been seen.

Author Response

Point 1: The data unit in the paper is recommended to be unified.

Response 1: We are very sorry to have caused you confusion about the disunity of the unit, and we will explain it to you in detail here.In fact, the calculation data involved in this paper are all SI units, but there is no clear explanation. Under your guidance, we have added the unit description to the paper.The reason why the units in Table 1 of the paper are not SI units is that the data interpretation table is the original version provided by NGSIM, and the change of the units will affect the readers' interpretation of the source data. We have added the explanation of units in front of the table to prevent readers from misunderstanding.

The improved contents are as follows:

Since the units in the table are not convenient for calculation and observation, all the data in this paper have been converted into international system units. The time unit is seconds, the distance unit is meters, the speed unit is m/s, and the acceleration unit is m/s2.However, it needs to be clarified here that the data unit in the source table is the unit shown in the table. All the data in this paper have been converted before calculation.

 

Point 2: It is recommended to refer to other papers for the writing of NG data names, and the names in all lowercase have not been seen.

 

Response 2: Thank you very much for your valuable advice,Under your suggestion, we consulted similar literatures and drew lessons from their writing of NG data units.We have standardized and unified all charts and words involved in the lower case description of NG data in the text. The contents and methods of the specification are as follows:

Based on Preliminary test results, this paper preliminarily selects six data of v_Vel, v_Acc, Time_Headway, Space_Headway, Preceding_Vel and Preceding_Acc for subsequent analysis.

Column Name	Description	Unit
Vehicle_ID	Vehicle identification number.	-
Frame_ID	The number of frames of the data at a certain time.	0.1s
Total_Frame	The total number of frames of this vehicle in this dataset.	0.1s
Global Time	Time stamp.	ms
Local_X	Abscissa of the center of the front of the vehicle.	Feet
Local_Y	Ordinate of the center of the front of the vehicle.	Feet
Global_X	Abscissa of the vehicle in the global coordinate system.	Feet
Global_Y	Ordinate of vehicle in global coordinate system.	Feet
v_length	Vehicle length.	Feet
v_width	Vehicle width.	Feet
v_Class	Vehicle class:1-Motorcycle；2-Light-Duty Vehicle；3-Large Vehicle.	-
v_Vel	Instantaneous speed of vehicle.	Feet/s
v_Acc	Instantaneous acceleration of vehicle.	Feet/s2
Lane_ID	Current lane position of the vehicle.	-
Preceding	Vehicle ID of the vehicle in front of the same lane.	-
Following	Vehicle ID of the vehicle following the vehicle in the same lane.	-
Space_Headway	Distance from the front center of the vehicle to the front center of the front vehicle.	Feet
Time_Headway	Time required to travel from the front center of the vehicle to the front center of the vehicle.	s

By observing the Kendall correlation coefficient analysis results and Spearman correlation coefficient analysis results, it can be seen that the v_Vel has a large correlation with Space_Headway and Preceding_Vel, and it can be preliminarily concluded that:

v_Vel expected by the driver is greatly affected by the Space_Headway and the Preceding_Vel, and it is found from the data changes that the greater the Space_Headway, the greater the expected v_Vel, and the greater the Preceding_Vel, the greater the expected v_Vel.

The expected v_Acc of the driver is greatly affected by the Preceding_Acc, and it is found from the data change that the greater Preceding_Acc, the greater the expected acceleration.

The driver's desired Space_Headway is greatly affected by the Preceding_Vel and the v_Vel. From the data changes, it is found that the larger the Preceding_Vel and the smaller the v_Vel, the greater the desired Space_Headway,The above three conclusions are fully consistent with the stimulus response model. The driver's response behavior (v_Acc) is mainly affected by the Preceding_Vel (excitation). The v_Vel is mainly balanced with the Preceding_Vel, and the Space_Headway is related to the speed of the two cars.

 

As there are many and scattered modifications, only some of them are shown here. Please review the paper for details

 

 Finally, thank you for reviewing my paper in your busy schedule. I benefited a lot from your suggestion.I sincerely wish you all the best!

Author Response File: Author Response.docx