Maneuver-Based Objectification of User Comfort Affecting Aspects of Driving Style of Autonomous Vehicle Concepts

Round 1

Reviewer 1 Report

The topic is interesting, timely, and of importance in autonomous vehicle design and development. The paper presents a framework detailing procedure beginning with a variety of driving styles  to technical specification  to realization and testing. Furthermore, a combination of quantitative and qualitative evaluation is performed.

I noticed that citations are missing in the body of the manuscript, which should be fixed.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

English

1. Thorough editing is needed. The paper, in its current form, is challenging to read.
2. Throughout the paper, references are replaced by “Error! Reference source not found”.
3. There are too many non-English references. As this is an international journal aimed at all readers, the authors should remove non-English texts.

State of the art

1. The main idea in the introduction is that AV should allow multiple driving styles so passengers will feel comfortable and adopt the technology. Therefore, sections 2.1-2.3 are not needed and distracting.
2. Finally – The authors characterized maneuvers based on comfort and on travel time -the introduction provides information about comfort but says little about the travel time.
3. Line 216 – missing explanation on what are “conventional driving maneuvers” and what are “described maneuvers for investigations”.
4. Lines 216-224 it is not clear why a “one-dimensional optimization” is suitable for a conventional vehicle, and a “multidimensional problem” is ideal for an AV.
5. Line 225, what are “one-dimensional maneuvers”?
6. Line 232 should the word “quality” be replaced with “comfort”?

Methodology

1. The term “functionality of the maneuver” should be explained.
2. Line 239, what are “comfort-related variables.”
3. Line 241 what are the “meaningful boundary conditions”?
4. Figure 4 is not mentioned in the text.
5. Lines 247-250 - I did not manage to understand the message in this paragraph.
6. Lines 259-269 what are the “seven defined driving maneuvers”?
7. Lines 261-262, who were the “test drivers” and “test supervisor”? How could the test supervisor know what are the levels of comfort? How could the authors understand that the test driver approach defined comfort levels reliably? Was there some objective measure for that?
8. Line 296 how “Experimental runs” different from test drives?
9. Line 325: The model in equation 1. should consist of two random effects: (1) one term that accounts for the repeated measurements of the same test person. (2) one term that accounts for the specific time point of measurements (for example, in maneuver B there are two turns and several accelerations and decelerations)). Also, the authors should explain (1) the differences between the original model and their modified model. (2) the meaning of the coefficients and the |X-X0| term (what is X and what is X0).
10. The authors did not use an automated vehicle for their testing, the level of comfort (stress/trust) may be significantly different between a regular vehicle driven by a test driver and an AV. Keeping the discussion on AVs in the introduction, not using an AV in the study, and specify this as a limitation (as the authors did ) is not the right way to present the context of this work.
11. Line 357-361: The authors should provide a detailed description of the questionnaire - especially the second block. The description should include the exact wording of the questions. Also, the authors should spec
12. Pilot study: Two maneuvers were selected; the authors should also describe the maneuvers that were not selected and their corresponding results. Also, the authors should explain how common these maneuvers (and their comfort levels) in real driving.
13. Figure 5: what is the meaning of the term “t_max”? Why there are points between the arrows? What do they signify?
14. Line 393 – some descriptive statistics for the “subjective comfort score” should be provided.
15. Regression analysis – lateral/longitudinal could be taken at any time point throughout the maneuvers A and B during these maneuvers, several accelerations, decelerations, and turnings right/left occurred. Which time points were used to collect the data, and why?
16. Figures 7 and 8 – I believe that the authors used the prediction band rather than the confidence level.
17. Subjective data: Please describe H0.1…H0,4.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

The paper shows the results of research into providing a driving style of autonomous vehicles, which could as well be a major decision point for purchase in the future. Very well written and interesting research 

Author Response

Thank you.

Reviewer 4 Report

This paper deals with the problem of comfort related to driving styles of autonomous vehicles. It proposes an interesting methodology for describing driving maneuvers and relate the proposed parameters to perceived comfort. The set of parameters include different aspects of acceleration, braking, jerk, steering, etc.

Authors report on a user study in which participants subjectively evaluated the driving style and comfort. The reported comfort level was then the basis for calculating thresholds of objectively measured driving parameters with nonlinear regression analysis.

This paper is in my opinion very well written and clearly describes an interesting approach to controlling and manipulating comfort of autonomous vehicles. The most important message of the authors is that the perceived comfort changes significantly when the role of the driver changes to the role of passenger and this needs to be considered by car manufacturers when designing and parametrizing and vehicle dynamics.

I have some question related to the manuscript:

287: Authors mention that the maneuvers must have a factor of passage time > 1.  What exactly is meant by the “passage time”? Is that the total duration in seconds of the maneuver in each measurement group?

320: How is this averaging performed? Provide more details.

332: How is the boundary X0 defined?

335-336: A R-squared is a typical measure of goodness-of-fit for LINEAR regression. Scientific literature suggests that it should not be used with non-linear regression and it’s results are not reliable:

https://blog.minitab.com/blog/adventures-in-statistics-2/why-is-there-no-r-squared-for-nonlinear-regression

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2892436/

I don't have experiences with Stevens’ power law regression model but it seem like a very typical non-linear model. Authors should provide very clear justification of why R-squared is a reliable measure of goodness-of-fit in their case.

This last question is in my opinion the most important concern as it questions the validity of the results and the entire experiment.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf