Using Bluetooth Low Energy Technology to Perform ToF-Based Positioning
Round 1
Reviewer 1 Report
The authors present an interesting paper about a position framework based-on Bluetooth Low Energy technology and ToF-based methodologies. However you should make a few improvements, namely:
1.- the paper needs a serious review with regard on its organization, as for example, the figures do not fit in well with the text, which limits the understanding of the paper on a fluid way.
2.- From 2nd figure the paper must be reorganized.
3.- The test scenario should be more detailed. It is necessary to cleary convey to the reader what the objective is. From the figure 4 gives the idea that the authors sum up to "stop the moveing device/anchor" and just apply their method as if it were a simple triangulation problem.
4. Is the "mobile anchor" effectively mobile during the testing? What are the consequences on the accuracy and acuity in terms of mobility? How does the system behave as the anchor moves, apporaches/departs, from the reference nodes?
Finally, congratulations on your work
Author Response
Answers to Reviewers’ remarks
We would like to thank both Reviewers for their constructive criticism, that helped us gaining further insight and improving our work. In the following answers to the Reviewers’ remarks are given.
Answers to Reviewer 1 remarks
The authors present an interesting paper about a position framework based-on Bluetooth Low Energy technology and ToF-based methodologies. However you should make a few improvements, namely:
Remark 1- The paper needs a serious review with regard on its organization, as for example, the figures do not fit in well with the text, which limits the understanding of the paper on a fluid way.
Answer – The Reviewer is right; we moved figures to put them closer to the text mentioning them.
Remark 2 - From 2nd figure the paper must be reorganized.
Answer – The Reviewer is right; we checked and reorganized the paper. As already mentioned, we revised distance between figures and the text describing them. We also added a new Figure, with the aim of better describing the system behavior, and provided more insight on the system processing time, that was modeled and measured. A more detailed description of the modifications is provided in the answers to the following specific remarks.
Remark 3 - The test scenario should be more detailed. It is necessary to clearly convey to the reader what the objective is. From the figure 4 gives the idea that the authors sum up to"stop the moving device/anchor" and just apply their method as if it were a simple triangulation problem.
Remark 4 - Is the "mobile anchor" effectively mobile during the testing? What are the consequences on the accuracy and acuity in terms of mobility? How does the system behave as the anchor moves, approaches/departs, from the reference nodes?
Answer to both Remarks 3 and 4 – The Reviewer is right, even if the main aim of this paper is to demonstrate the feasibility of a BLE based fully wireless ultrasound positioning system, proving that the BLE latency is a tolerable source of uncertainty in ToF measurements, the suitability of the proposed solution for tracking purposes should be assessed. The system was initially configured to perform about 6 measurements per minute, a measurement rate suitable for portable scenarios, where the target is slowly moving. To assess the feasibility of the system for a tracking application, we explored the achievable measurement rate, tuning the BLE nodes with more aggressive setting. At the same time, we profiled the processing time of both the ToF measurement algorithm running on the BLE nodes and of the positioning algorithm running on the PC.
The BLE nodes (broadcaster and observers) were tuned by reducing the delay between consecutive measurements from 10s to about 2 s. Thus, each anchor node may perform about 30 ToF measurements per second. When reducing such delay below 2s the ToF measurement became unstable, possibly due to ToF measurement processing time. To verify this assumption, the time txcorr required to compute the cross-correlation sequence between the received chirp signal and the normalized chirp sequence stored in the anchor node was assessed using an oscilloscope. For a record of M=2000 samples we obtained txcorr@2.12 s. The processing time tHilbert required to compute the Hilbert transform and the tmax,search time required to find the maximum index corresponding to the ToF were evaluated using the same approach, obtaining tHilbert@16 ms and tmax,search@500 ms respectively. These results seemingly prove the assumption that the bottleneck limiting the measurement rate is actually the processing time tproc=txcorr+tHilbert+tmax,search@2.2s, and that tproc is dominated by txcorr. Thus, the ToF measurement rate may be improved by replacing the PSoC nodes with more powerful processors of by using more efficient algorithms. In particular, since the cross-correlation is currently computed using a time domain algorithm, replacing it with a FFT algorithm may reduce the computational complexity from O(M2)=4×106 to O(M×log2(M))@2.19×104, reducing txcorr by two orders of magnitude and reducing tproc to a few tens of milliseconds.
Conversely, the position estimation is performed by a Matlab script, running on a PC equipped with an Intel I7 3.6 GHz CPU, 8 GB RAM, and a Windows 10 Operating System. The positioning algorithm is currently a non-linear least square fitting, was profiled by estimating the completion time using the Matlab tic and toc functions over N=110 measurements, obtaining a mean time of 12 ms and a standard deviation of 0.58 ms. Since the current setup aims at characterizing the ToF measurement accuracy, data transfer between the smartphone that concentrates the ToF measurements and the PC running the positioning algorithm is done manually, by generating and transferring a text file from the smartphone to the PC. This bottleneck may be removed by running the positioning algorithm on the smartphone, and, even if it is out of the scope of this paper, is an envisioned future development. The positioning algorithm is not a bottleneck, but it may be replaced with a Best Linear Unbiased Estimator (BLUE), improving the processing time [11], or by an extended Kalman filter, that may enable sensor fusion with inertial readings.
Please notice that we updated the introduction, better highlighting the paper aims, and section 4.3 of the paper, including and discussing the new results.
Remark 5 - Finally, congratulations on your work.
Answer – Thank you for your good words!
Author Response File: 
Author Response.pdf
Reviewer 2 Report
In this paper, a distributed positioning systems based on ToF measurements and ultrasound signals with centimeter-level accuracy and compatible with commercial hard-349 ware platforms was developed. Time-synchronization between transmitters and receivers is achieved by using the BLE technology.
The effectiveness of the proposed scheme is discussed with experiment results.
The topic is very interesting and the contribution of this paper is solid.
However, the authors just performed experiments with relatively conditions.
As shown in the results of experiment, the distance error is increased as the distance increases, while the maximum distance was only 2 meters.
It is not clear what kinds of applications are considered with only 2m distance.
Also, 3 anchor nodes are placed in a straight line. What if the distance among the anchors is longer or shorter?
What if the different pattern for the distribution of the anchors is considered?
How many times were the experiments performed to achieve the statistical assurance?
Author Response
Answers to Reviewers’ remarks
We would like to thank both Reviewers for their constructive criticism, that helped us gaining further insight and improving our work. In the following answers to the Reviewers’ remarks are given.
Answers to Reviewer 2 remarks
In this paper, a distributed positioning systems based on ToF measurements and ultrasound signals with centimeter-level accuracy and compatible with commercial hardware platforms was developed. Time-synchronization between transmitters and receivers is achieved by using the BLE technology. The effectiveness of the proposed scheme is discussed with experiment results. The topic is very interesting and the contribution of this paper is solid.
Remark 1 - However, the authors just performed experiments with relatively conditions.
As shown in the results of experiment, the distance error is increased as the distance increases, while the maximum distance was only 2 meters. It is not clear what kinds of applications are considered with only 2m distance.
Answer - The Reviewer is right; the test conditions did not stress extensively targeted applications. Please notice that the paper aims at assessing the feasibility of a ToF based ultrasound positioning system, where ultrasound transmissions are wirelessly triggered using BLE RF signaling rather than wired or WiFi based solutions. The system operational range actually depends on both the adopted sensors’ properties and on the number and position of deployed sensors, since sensor fusion techniques may combine several noisy measurements to achieve an accurate position estimation. To this aim it is worth noting that BLE supports mesh networks, permitting the deployment of a large number of low-cost sensors, reducing the average distance between anchors and mobile nodes, and enabling increased accuracy using sensor fusion. Please notice that a positioning range of about 4m was actually reached in a 3D wired scenario, using additional anchor nodes equipped with the Murata transducers mentioned in this activity, as described in the newly mentioned reference [53]. We updated the paper introduction and section 4.3 accordingly.
Remark 2 - Also, 3 anchor nodes are placed in a straight line. What if the distance among the anchors is longer or shorter? What if the different pattern for the distribution of the anchors is considered?
Answer - As suggested by the Reviewer, the pattern distribution of the anchors indeed affects the achievable accuracy and was studied in a previous activity, using the Geometrical Dilution of Precision as a performance metric [11]. Since the main objective of the paper is to assess the impact of BLE based wireless triggering on ToF measurements, only a minimal set of anchor nodes was deployed, using a non-optimized layout, a minimum number of nodes, and the system measurement rate was not stressed. Please notice that a better accuracy may be obtained using anchor positions surrounding the region where the mobile node operates. This would require to keep into account the directivity of the ultrasound transducers. Even if this problem is mentioned in the literature, for instance in the following
- Rivard, J. Bisson, F. Michaud and D. Létourneau, “Ultrasonic Relative Positioning for Multi-Robot Systems,” Proceedings of 2008 IEEE International Conference on Robotics and Automation, Pasadena, CA, USa, May 19-23. 2008-
where a cone shaped reflector was added to the transducer to make their behavior more isotropic, we consider this specific issue to be outside the scope of the paper, scheduling it as a future development. To better clarify this issue, a few remarks were added at the beginning of Section 4.3 and the mentioned paper was added as reference [57] in the revised paper
Remark 3- How many times were the experiments performed to achieve the statistical assurance?
Answer - For each testing position, the ToF measurement and the following position estimation was repeated 18 times. This is now explicitly stated in section 4.2.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
The authors implemented the sugestions previousely indicated, thus substantially improving the quality of paper. Thus I am of the opinion that the paper meet the conditions and quality to be published.
Finally, congratulations on your work.
Reviewer 2 Report
The contribution of this paper is well explained and the revised paper can be considered for possible publications.
