3D Distance Filter for the Autonomous Navigation of UAVs in Agricultural Scenarios

Round 1

Reviewer 1 Report

Review of paper "3D distance filter for the autonomous navigation of UAVs in agricultural scenarios"

The paper proposes an extension of a 2D ellipsoid-based filtering localization method to the 3D case for real-time navigation of the UAV within the vine rows. The paper is highly technical, and the computational details of the methodology will probably be comprehensible only to a selected group of readers familiar with cited previous work, to which the paper refers. The conclusions seem valid with respect to the reported results, which demonstrate the benefit of using of using the proposed 3D distance filter. Although expected, there is a value in experimental validation of accuracy improvement of UAV position estimation when using the LCG maps and additional ultra-sound sensors. However, the present methodology description is quite difficult to digest, and I believe there are possible improvements in making it more accessible.

Introduction:
In the last paragraph on page 3, the main contributions of the paper are given. Should the first contribution be "the distance from the UAV Center-of-Mass (CoM) and the projected LCG map" be "the distance between the UAV ..."?

Section 2.2:
Figure 2 and its caption are not very informative. Why is the terrain plane used to derive the roto-translated plane - is the UAV orientation given with respect to the terrain? The reason why the vertices of 2D slices are using only x and y coordinate becomes clear only in Section 3.1, it should be explained here which 2D reference coordinate system is used.

Section 3:
Some descriptions here are particularly convoluted. Since Figure 4 is at the heart of the proposed method, I suggest that the authors first provide an overview of the procedure, e.g. describe the meaning of all variables (input, output and intermediate) and the data flow in Figure 4. This should be followed by clearly separated explanation of individual steps (2d map update, row selection, distance filter update) with references to corresponding parts of Figure 4. It should also be explicitly labelled which part in the Figure 4 is the 3D distance filter.
Some values in Figure 4 are not used anywhere in the text (e.g. p_k, could be a typo?).
The way the relevant part of the 2D slice is selected (lines 220-222 in Section 3.1) should be explained in more detail.
In the same way, it should be better explained what it means that "a segment complies with the corresponding d". Why is only one segment used in this step?
It is not clear where the equations 16a, 16b, and 16c come from? In case they are derived in some related work, a corresponding reference should be provided.
It is not clear where Eq. (23) applies in the Figure, because the notation does not match; I assume the quantities computed by Eq. (23) are denoted by x_{k|k-1} and P_{k|k-1} in the Figure. Further, what are the values of x and delta with superscript D in Figure 4, and how are they computed?
Just below eq. (18), u_k is used to denote process noise, it should probably be w_k.

Section 4:
The reference to Figure 10 is missing in the paragraph before Table 3. What do the values on the vertical scale of Figure 10 represent?

Section 5:
In the last sentence, "real scenarios" should probably be "real-time scenarios".

There are quite a few spelling errors, so I would also suggest a careful spell check.

Author Response

We  thank the Reviewer for his/her detailed comments.. Our replies are reported in the attached file on a point-to-point basis. Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

The manuscript entitled: „3D distance filter for the autonomous navigation of UAVs in agricultural scenarios” presents improved UAV localisation system to allow the autonomous navigation of rotary-wing UAVs within vine rows. The quality of the manuscript is good and the study is rather interesting. However, there are some drawbacks which should be improved. The topic of the study seems to be rather technical and not very scientific. Study can be extended to more comprehensive, i.e. more replication in the study should be performed.

Why the values in the Fig. 7 are not from 0 but from higher values, eg. from 374 to 380. The same case is for Fig. 9.

Figures and tables should be self-explanatory, i.e. clear enough without reading all the manuscript. Some of them are not clear, e.g in the Fig. 8 is not clear where is “apex i” and what does it mean time on horizontal axis.

Discussion is without any comparison to similar studies. There was no similar studies on such topic? It should be explained.

In conclusions should be explained what should be the next research step. It is too general.

Author Response

We  thank the Reviewer for his/her detailed comments. Our replies are reported in the attached file on a point-to-point basis. Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

Lines 91-93: “On the other hand, it is challenging to retrieve accurate 3D maps first and then to exploit them in real-time for guidance and navigation tasks, because of their high data storage demand with respect to the computational capabilities of the UAV hardware.” should be better rephrased into something like: “On the other hand, it is challenging to retrieve accurate 3D maps first and then to exploit them in real-time for guidance and navigation tasks, because of their high data storage and computational demand with respect to the commonly available UAV hardware.”

Self-citation at line 114 appears to be inappropriate.

Lines 292-294 contain a self-evident statement. It should be expunged or rephrased.

Has been the proposed novel approach tested only in a simulation environment or also on a real mobile configuration?

Are there any suitable hardware platforms on the market being able to host the proposed solution and to comply with the strict weight limitations that characterize UAVs?

Author Response

We  thank the Reviewer for his/her detailed comments. Our replies are reported in the attached file on a point-to-point basis. Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

The manuscript was improved according all suggestions and current version of the manuscript can be published.