Robust Path Tracking Control for Autonomous Vehicle Based on a Novel Fault Tolerant Adaptive Model Predictive Control Algorithm

Round 1

Reviewer 1 Report

The authors have presented a very interesting research topic which has a real-world application for autonomous vehicles. The proposed methodology is also very sound and the experiments carried out and reported on are all valid to the best of my knowledge.

 

Please my comments/queries are detailed as follows:

 

(1)

This manuscript has several grammatical and typographical errors. The authors are advised to thoroughly proofread and reword the manuscript as far as possible to make it more meaningful and easy to interpret by the potential English speaking readers.

E.g.,

(a)

Authors need to clarify and/or reword lines 21 to 23 in the abstract for meaningfulness:

"In addition, this developed fault tolerant path tracking control algorithm performers robustness and effectiveness when sensor failure occurs in the double lane change lane condition."

(b)

Abstract (Lines 24 to 25) - missing definite article and use of singular noun instead of plural noun.

that this research will contribute to the development of safer and more intelligent autonomous driving systems

(c)

Others

(2)

Authors need to clarify lines 45 to 47 in the Introduction

Due to the inevitable variation of vehicle longitudinal velocity, the previous researches on a fixed constant face a significant limitation in real driving conditions.

What fixed constant is being referred to?

 

(3)

Authors need to write the full-meaning of the following acronyms (and possibly others) in at least one instance before reusing them in the paper:

LIDAR

RADAR

GPS

 

(4)

In lines 77 to 78 in the Introduction, the authors need to clarify the difficulty attributed to data collection. To the best of my knowledge, many sensors, sensing units and sensing paradigms either have embedded data logging features or can easily be synchronised to data loggers. So, why is data collection difficult in this case?

 

(5)

If the authors are going to include this paragraph (i.e., lines 95 to 100) in the Introduction, then they must clearly state which sections of the paper the stated contents can be found in. E.g., The vehicle dynamic model is established and model linearization process is described in section 2, and so on.

 

(6)

Figure 1 can be made clearer and much easier on the eyes by simply highlighting the key components or features (i.e., Nonlinear vehicle dynamic equations, linearization, multi-sensor data fusion and fault isolation, fault sensor signal detection, and adaptive model predictive controller design) of the proposed method in the flow diagram and providing fuller details (explanations and equations) in-text (or in the text body). 

 

(7)

In section 2.1, line 112, what is the implication of the statement below compared to the objectives/goals/aims already stated in the Introduction?

"The main goal of this work is to enable the vehicle to track the target path quickly and steadily."

 

(8)

In section 2.1, lines 117 to 123, the authors need to clarify the rationale behind the idealized assumptions proposed. 

 

(9)

Even though I am quite confident about the validity of the experiments and results in this paper, the authors need to provide more information to support the reliability of the results presented. Such additional information may include details about the system or platform setup for the simulations, repeatability and/or reproducibility of experimental results (given the same conditions).

Author Response

Response to Reviewer 1 Comments The authors have presented a very interesting research topic which has a real-world application for autonomous vehicles. The proposed methodology is also very sound and the experiments carried out and reported on are all valid to the best of my knowledge. Please my comments/queries are detailed as follows: Point 1: This manuscript has several grammatical and typographical errors. The authors are advised to thoroughly proofread and reword the manuscript as far as possible to make it more meaningful and easy to interpret by the potential English speaking readers. E.g., (a) Authors need to clarify and/or reword lines 21 to 23 in the abstract for meaningfulness: "In addition, this developed fault tolerant path tracking control algorithm performers robustness and effectiveness when sensor failure occurs in the double lane change lane condition." (b) Abstract (Lines 24 to 25) - missing definite article and use of singular noun instead of plural noun. That this research will contribute to the development of safer and more intelligent autonomous driving systems (c) Others. Response 1: Dear professor, first of all, thank you very much for your comments! According to your suggestions, we made the following changes: (a) We reword the mentioned sentence as: “In the double line change path tracking control experiment, when the sensors failure occurs, the proposed method shows better robustness and effectiveness than the traditional methods.” (b) “It is foreseeable that this research will contribute to the development of safer and more intelligent autonomous driving system, which in turn will promote the industrial development of intelligent transportation system.” Point 2: Authors need to clarify lines 45 to 47 in the Introduction “Due to the inevitable variation of vehicle longitudinal velocity, the previous researches on a fixed constant face a significant limitation in real driving conditions”. What fixed constant is being referred to? Response 2: Thanks for your comments. The fixed constant is being referred to the longitudinal velocity of vehicle. “Due to the inevitable variation of vehicle longitudinal velocity, the previous researches on a fixed constant longitudinal velocity face a significant limitation in real driving conditions.” Point 3: Authors need to write the full-meaning of the following acronyms (and possibly others) in at least one instance before reusing them in the paper: LIDAR, RADAR, GPS. Response 3: Thanks again for your comments. LIDAR and RADAR are the names of the sensor types, not acronyms, as described in many related papers. GPS is the acronym of “Global Positioning System”, which has been given in the article. Point 4: In lines 77 to 78 in the Introduction, the authors need to clarify the difficulty attributed to data collection. To the best of my knowledge, many sensors, sensing units and sensing paradigms either have embedded data logging features or can easily be synchronised to data loggers. So, why is data collection difficult in this case? Response 4: Thanks again for your comments. I very much agree with your point that the recording of sensor data is very not difficult in most instances. However, a large number of manual annotations on the collected data are required to form a labelled dataset before it can be used for training and testing of deep learning neural networks, which means a large manual effort is required. To the best of our knowledge, there is still no open source dataset can be used for solving the relevant problems described in our paper. We changed the mentioned sentence as: “Even though the deep learning neural network can deal with the raw data without manually feature extraction, however, the deep learning based methods require a large scale of dataset with manual annotations, which is difficult to collect.” Point 5: If the authors are going to include this paragraph (i.e., lines 95 to 100) in the Introduction, then they must clearly state which sections of the paper the stated contents can be found in. E.g., The vehicle dynamic model is established and model linearization process is described in section 2, and so on. Response 5: Thanks again for your comments. We made the following changes: “The rest of this paper is organized as follows. The vehicle dynamic model is established and model linearization process is described in section 2.1 and 2.2. The constraints and objective functions are constructed from section 2.3 to 2.6, and the path tracking control algorithm based on model predictive control is designed in section 2.7. The multi-sensor information fusion algorithm and fault signal detection and isolation algorithm are described in section 2.8 and 2.9, respectively. In section 3, the simulation and experimental verification using double lane changing path tracking scenario is presented. Finally, the conclusions are given.” Point 6: Figure 1 can be made clearer and much easier on the eyes by simply highlighting the key components or features (i.e., Nonlinear vehicle dynamic equations, linearization, multi-sensor data fusion and fault isolation, fault sensor signal detection, and adaptive model predictive controller design) of the proposed method in the flow diagram and providing fuller details (explanations and equations) in-text (or in the text body). Response 6: Thanks again for your comments. The explanations and equations have been given in details in our paper, and we improved the figure 1 as follow: Figure 1. The flowchart of the proposed method Point 7: In section 2.1, line 112, what is the implication of the statement below compared to the objectives/goals/aims already stated in the Introduction? "The main goal of this work is to enable the vehicle to track the target path quickly and steadily." Response 7: Thanks again for your comments. In Section 2.1, we mainly established vehicle dynamics equations. In fact, it is necessary to establish much more complex vehicle dynamics equations for specific problems, such as analysing the coupling relationship between vehicles and roads. However, the main goal of this work is to enable the vehicle to track the target path quickly and steadily. Therefore, we established a relatively simplified vehicle dynamics model by ignoring road fluctuations, suspension motion, the mechanical effects of steering system, etc. Point 8: In section 2.1, lines 117 to 123, the authors need to clarify the rationale behind the idealized assumptions proposed. Response 8: Thanks again for your comments. The monorail vehicle dynamics model e is used to verify the lateral motion control during the path tracking process. The vertical motion is mainly related to the vehicle suspension system and load movement, and has little effect on the lateral motion. In addition, in the process of lateral motion control, the steering systems often have delays and other mechanical effects, which can be compensated by parameter correction. Thus, road fluctuations and suspension motion, the load movement, the vertical and horizontal coupling relationships and the mechanical effects of steering system are ignored in our paper. Point 9: Even though I am quite confident about the validity of the experiments and results in this paper, the authors need to provide more information to support the reliability of the results presented. Such additional information may include details about the system or platform setup for the simulations, repeatability and/or reproducibility of experimental results (given the same conditions). Response 8: Thanks again for your comments. The details about the system and vehicle platform setup for the simulations follows the configurations of our own developed autonomous vehicle, as shown in Table 1 in section 3.2. We also conducted a path tracking control experiment using self-developed autonomous vehicle. The details about the sensor system and autonomous vehicle platform setup are shown in Table 1. Table 1. Configurations of autonomous vehicle. Hardware Property Laser LIDAR: HESAI Pandar 40 Lines:40; Range: 200m; Angular resolution: 0.1°; Updating: 20Hz; Accuracy:±2cm Radar: Delphi ESR Range:100m; Viewing field:±10deg; Updating: 20Hz; Accuracy:±5cm, ±0.12m/s, ±0.5° Navigation system: NovAtel SPAN-CPT Accuracy: ±1cm, ±0.02m/s, ±0.05°(Pitch/Roll),0.1°(Azimuth); Updating:10Hz Camera:SY8031 Resolution: 3264×2448; FPS:15;Viewing field:65°(vertical), 50°(horizontal) Image processer: NVIDIA JTX 2 CPU:ARM Cortex-A57(Quad-core,2GHz); GPU: Pascal TM(256 core,1300MHz); RAM:LPDDR4(8G, 1866MHz, 58.3GB/s) Controller:ARK-3520P CPU: Intel Core i5-6440EQ (Quad-core,2.8GHz); RAM:LPDDR4(32G, 2133MHz, 100GB/s) Finally, thank you very much for your comments, which are very valuable for the improvement of our paper.

Author Response File: Author Response.docx

Reviewer 2 Report

In this study, the fault signals of sensors are detected and isolated in real-time to enhance the reliability of fused data. The simulation and experimental results of double lane changing show the effectiveness of the proposed method in this work. However, there are some comments as follows.

1. If multiple sensors are used to measure the same parameter of autonomous vehicle, the authors merge the outputs of all sensors using the weight assignment method, Eq.(22), for data fusion purpose. While the fault signal is detected, e.g. i_g, the corresponding element, i.e. m_g, of the fault detection matrix (M) is set as 0 (m_g=0) in Eq.(21). Although Eq.(22) can guarantee ∑w_j = 1, the zero element (m_g) multiplies the corresponding weight (w_g) will lead to an incorrect result of data fusion (O) in Eq.(21). Is that correct?
2. In Eq.(23), N is the number measurements from each sensor. How to determine a proper N?
3. The typos and grammatical errors (such as, in lines 23, 225, 329, 330, … etc.) should be correct.

Author Response

Response to Reviewer 2 Comments

 

In this study, the fault signals of sensors are detected and isolated in real-time to enhance the reliability of fused data. The simulation and experimental results of double lane changing show the effectiveness of the proposed method in this work. However, there are some comments as follows.

 

Point 1: If multiple sensors are used to measure the same parameter of autonomous vehicle, the authors merge the outputs of all sensors using the weight assignment method, Eq.(22), for data fusion purpose. While the fault signal is detected, e.g. ig, the corresponding element, i.e. mg, of the fault detection matrix (M) is set as 0 (mg=0) in Eq.(21). Although Eq.(22) can guarantee ∑wj = 1, the zero element (mg) multiplies the corresponding weight (wg) will lead to an incorrect result of data fusion (O) in Eq.(21). Is that correct?

 

Response 1: Dear professor, first of all, thank you very much for your comments!  We all agree your arguments and we have modified Eq. (21) and Eq. (22) as follow:

                                                (21)

                                                             (22)

The Eq. (22) still can guarantee .

In this paper the weights for each sensor system can be write as:

The subscript symbols g, l and v represent the GPS, LIDAR and vision systems. We can see that, when sensor failure occurs to GPS system,  and . Thus, we can isolate the fault information while ensuring that the data fusion results are correct.

 

Point 2: In Eq.(23), N is the number measurements from each sensor. How to determine a proper N?

 

Response 2: Thanks for your comments.  The output frequencies of GPS, LIDAR and vision systems are 10Hz, 10Hz and 30Hz respectively. According to the Nyquist sampling theorem, the sampling frequency in our work is 75Hz, which means that the number of samples per second is 75 times and N can be obtained by multiplying the measure time. We hope that we have correctly understood and answered your question.

  

Point 3: The typos and grammatical errors (such as, in lines 23, 225, 329, 330, … etc.) should be correct.

 

Response 3: Thanks again for your comments. We have correct the typos and grammatical errors you mentioned. Each author of this work has read the paper carefully again and corrected some of the grammatical errors, but since English is not our native language, there may be some sentences that still can be improved. We will strengthen our writing skills in English in the future.

 

Finally, thank you very much for your comments, which are very valuable for the improvement of our paper.

Author Response File: Author Response.docx