Mobile-node to ground-node communication

[jakapoor]

Establish a communication protocol to get mobile and ground nodes to speak to one another, placing emphasis on 1) maximizing sleep time for mobile nodes and 2) keeping them rigidly scheduled to avoid overlap

[jakapoor]

Our recent meeting with Dr. Kan suggested that there may be significant advantages to a CDMA over a TDMA approach to tag transmission scheduling. An added advantage of such a communication scheme is that the signals in a CDMA approach can be ultra wideband (UWB) pseudo-random number "Gold" codes, which are extremely useful for TDOA-based localization.

The question remains whether CC1310-based tags are capable of CDMA. An on-off-keying approach to CDMA was developed by Krueger, in which he used a GPIO pin on a transceiver IC to toggle the power supply to the RF generator on the same chip at 1-2 MHz. Whether this is possible on the CC1310 is not yet known.

[ibriggs]

CDMA does seem like a promising approach to accumulating data from multiple tags at the same time instead of scheduling the tags to transmit one at a time. However, I'm not sure if the CDMA scheme is compatible with making phase-based AOA measurements. The point of using these Gold codes is that they are mutually orthogonal to each other, so it is possible to distinguish multiple bit streams from multiple tags at the same time. However, I'm not sure if the same applies to the phase from multiple tags at the same time. In other words, gold codes can be used to isolate bits for each tag, but I'm not sure if they can isolate phase from each tag. If either Hunter or Dr. Kan could clear up this point, it would be much appreciated.

The second point I'm not clear on is how exactly TDOA would be implemented. I think there has been some confusion with regards to the language, so for the purposes of this thread I will refer to a single grouping of 4 antennas for making a single AOA measurement as an "AOA device". Would TDOA be implemented within one AOA device, or between multiple AOA devices? The spacing between the antennas within one AOA device is less than a wavelength, so the time difference of arrival would be on the order of nanoseconds. Even with the ultra high BW signal, the sampling frequency of the receivers is usually on the order of kilohertz or megahertz, so measuring a difference of nanoseconds doesn't seem possible. I think Dr. Kapoor suggested using sinc-based interpolation methods to increase the temporal resolution of the signal, but every signal processing method has its limits, so I'm not sure if those interpolation methods could grant nanosecond precision. If the time difference of arrival was being measured between AOA devices, that would be a different story. If spaced far enough apart, the time difference of arrival could be significant enough to be measured, but you'd still have the issue of the resulting ranging calculation being highly imprecise.

I think the second point isn't as important as the first, since making TDOA measurements seemed like a secondary objective to making the phase based measurements. However, if we were going to commit to CDMA, I think it is important that it does not mess up the phase-based measurements for making the angle of arrival measurement, which is the whole point of the project.

Let me know what you guys think about all of this.

Ian

[jakapoor]

Hi Ian,

Excellent points. Thank you for the detailed response! And thank you for responding on the repo. I'd be interested in hearing others' thoughts on this.

Regarding point 1:

How do overlapping CDMA signals (let's assume generated from OOK modulation) affect phase measurements of signals from different transmitters? My naive perspective is that this is potentially a big problem. As you mentioned, CDMA allows one to isolate the bits of the encoded signals, but it seems unlikely that this would have any effect whatsoever on extracting and isolating the phase of the RF carrier that makes up a single bit. My only thought about how this might work is that, given a certain number of overlapping signals, not every tag will be transmitting a logic-level one bit at the same time, so it may be possible to identify largely overlapping signals where a specific bit of the transmitter of interest is the only logic one bit being transmitted. This seems like a really poor approach to me, but it's the only way I can imagine getting around the interference created by multiple tags transmitting at once. If this approach were combined with a modified TDMA schedule, however, so that only subsets of tags were transmitting at once, or even a pure TDMA approach with CDMA-like signals (i.e. Gold codes), then we might be able to capitalize on the TDOA benefits of the resulting UWB signals, while largely avoiding the phase interference from multiple tags transmitting at once. For these reasons, to me, TDMA is still an attractive proposition. Thoughts?

Regarding point 2:

Are we talking about within-array or between-array TDOA? I apologize if I've been unintentionally vague about this. The accuracy of TDOA-based localization drops off to nearly chance once you move much beyond the perimeter of your coherent antennas. For this reason (as Ian correctly points out), a TDOA-based localization approach, where we were trying to use the arrival times of a signal at multiple antennas within a single ground-node (Ian's "AOA device"), would fail miserably. Only where we were detecting arrival time differences between signals received on multiple distributed ground-nodes would a TDOA-approach have any chance of succeeding. However, since those ground-nodes are not being driven by the same clock, we would have to deal not only with bulk time offsets, but phase, frequency, and sample rate drift between ground-nodes. Luckily, Whiting and Krueger have both published papers on using known-location timing beacons to compensate for all of these errors using our proposed hardware.

Summary

So, to summarize, to me, a TDMA-CDMA hybrid approach, with non-coherent, but calibrated distributed ground-nodes ought to allow us unfettered access to good phase difference information and good time difference of arrival information. The coarse-level localization we could do with TDOA then, might greatly help us to narrow in on the fine-scale phase-based localization results. The point is well-taken though, that we cannot let the CDMA or TDOA derail the primary goals of the project if they, in fact, are threatening to do just that.

Again, open to others' thoughts on this.

Julian

[jakapoor]

From Russell's [7/11/2018] Bill of Materials document:

For the purposes of this summer, one [CC1310] will be used as a prototype radio tag. The other will send frequency hopping protocols to the prototype radio tag from one of the receiver basestations. The frequency hopping protocols at receivers and the basestation CC1310 (which will send the protocols to the transmitting CC1310) will be coordinated by NTP (Network Time Protocol) via the USB modems. This will be a foundation moving forward to a TDMA multiple tag localization system, in which multiple basestation CC1310s coordinated by NTP will be able to send the same TDMA protocol to radio tags over an area size scaled by the number of basestation CC1310s.

As illustrated by Russell's comments above, the ground-node to mobile-node communication protocol needs to accommodate the mobile-node frequency-hopping protocols as well as general tag scheduling.

Regarding the details of the use of NTP via USB modems, as in issue #18 and #19, internet access cannot be guaranteed, so we will need a way to communicate with mobile-nodes without NTP. One possibility is an approach by Ripperger et al._2016.pdf, Hierold et al._2015.pdf, and Hierold et al._2015(2).pdf in which GPS units are used to establish a precise global time to all ground-nodes. Once a common time-base is established, mobile-nodes can be scheduled to transmit with a pre-determined frequency-hopping schedule.

A remaining question is whether the RTL-SDR receivers in the ground-nodes are capable of receiving multiple transmission frequencies without the need for beacon-based recalibration between each "hopped" frequency within a single transmission.

Thoughts?