An Autonomous Tracking and Landing Method for Unmanned Aerial Vehicles Based on Visual Navigation
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsAll comments in the attached pdf.
Comments for author File: 
Comments.pdf
All comments in the attached pdf.
Author Response
Please see the attachment
Author Response File: Author Response.pdf
Reviewer 2 Report
Comments and Suggestions for AuthorsThe authors propose a visual tracking algorithm based on fiducial markers and experimentally validate the results. An UGV is used to move an ArUco marker and the UAV is intended to follow the trajectory of the UGV.
Line 71: Please explain: Why are ArUco marker more suitable than e.g. april_tags or ar_tags?
The authors do not describe in detail what kind of controller they use. In line 87 a fuzzy PID controller is mentioned but not explained. In addition it seems that another controller is used for position control in an "outer loop". Is it a cascaded controller? The authors should clarify the complete controller setup. How is the transform - generated by the OpenCV algorithms - fed into the controller? Fig. 1 only provides a kind of state machine but not the real implementation with the used controllers.
Line 152: Where are these cameras mounted? What resolution do they provide? Details of the used hardware are quite important to get an idea about marker size related to pixel resolution.
Some of the data in Table 1 is misleading. The "deviation" seems to be the difference btw. ground truth - wherever it comes from - and the average of the measured values and not (!) the standard deviation. So why is it mentioned in the table?
The experiments carried out by the authors are not referenced to any "ground truth" data. How is e.g. the position or velocity measured? Is there another optical tracking system involved? Is it just based on odometry data of the UGV wheel encoders (Fig. 7a red line)?
Line 221: Is there an influence of the controller architecture? The fluctuation could be an indicator of a not optimized PID controller. The authors should consider the influence of the controller parameters.
With Fig. 9 the assumption of a - not well tuned - PID controller is obvious: the PT1 behaviour is clearly visible and the overshoot. The authors should clarify this and tune the controller accordingly.
Line 228: This conclusion is not supporting the fact from line 221. Please quantify it - what does "well tracking" mean?
Line 314: The authors should investigate the final landing procedure by the flight controller which is influencing the positional uncertainty of the complete procedure. When switching to a "standard" landing procedure in a distance of 0.3m on a moving target, the positional uncertainty of course increases.
The authors should investigate the influence of different markers at the same setup. That would at least generate a practical impact when choosing visual markers. In the current form the paper have a rather low scientific contribution.
Comments on the Quality of English LanguageEnglish language is understandable but a couple of expressions are misleading:
Line 17: "The inner and outer nested markers aid in the independent monitoring and touchdown of UAV." Sounds strange to me.
Line 33: "It is also [...]" Is this sentence necessary?
Line 66: "[...] potential for loss [...]" Do you mean risk?
Line 68: "[...] steer [...]" It is more a guidance or control as mentioned in the next sentence.
Please check the complete text again.
Author Response
Please see the attachment
Author Response File: Author Response.pdf
Reviewer 3 Report
Comments and Suggestions for AuthorsDear authors,
Your paper addresses an important topic in the field of UAV flight some tracking and landing methods based on visual navigation,relying solely on vision and landmarks and brings interesting approaches. It is of good quality, well and correctly grounded.
Author Response
Please see the attachment
Author Response File: Author Response.pdf
Round 2
Reviewer 2 Report
Comments and Suggestions for AuthorsThe 2nd version of the paper provides additional and necessary information in order to understand the authors approach.
Line 69 to 73: The explanation is not detailed enough for preferring ArUco markers. Please provide information about the mentioned detection errors comparing different markers e.g. by citing a valid reference.
The explanation of the controller is quite detailed and the simulation results are a valuable input for the latter interpretation of the results. But it is still unclear why the velocity controller has a latency of ~5s for vx as depicted in fig. 13. None of the simulation values provide a valid explanation for this behaviour. The authors should identify the source of this influence.
Line 258 to 259: This is not a valid scientific statement. The authors should identify the reason behind this behaviour. What is generating the overshoot? Is it the controller?
Line 293 to 297: Here the effect is described but the source of this observation is not identified nor quantified.
Line 384 to 386: The explanation of the landing procedure is not providing enough information. It is clear that the motor control PWM signal is used. The question is about the way the flight controller is recognizing the landing e.g. waiting for a mechanical impact of the IMU or using an optical distance sensor. This procedure influences the uncertainty of the results in table 10.
In addition, the odometry data of the UGV is not sufficient for ground truth. Especially when a circular motion is applied, odometry will never give precise results and the positional uncertainty increases with each turn. This data shouldn't be used as a reference. The authors should use external tracking devices or other measurement systems to generate a valid reference, so reliable ground truth data.
The results presented are all influenced by the insufficient tuned PID controller. It is not clear, whether e.g. the deviation in fig. 17 to 18 is generated by the tracking algorithm or the controller. The source of the deviation can not be clearly identified, so the results are difficult to consider in a scientific setup. The authors should clearly identify the sources of these deviation, quantify them and generate a valid scientific statement.
Comments on the Quality of English LanguageSome wording is still unclear:
Line 94 to 98: "[...] sends horizontal channel commands [...]" What are "channel commands"? RC channels?
Please clearly explain your approach by using appropriate technical terms.
Author Response
Please see the attachment
Author Response File: Author Response.pdf
