Neural Network-Enhanced 'Cognitive Radio' Communicates With ISS (ieee.org) 29
IEEE Spectrum reports:
There's still plenty that can disrupt radio communications... Rather than waiting for a human on Earth to tell the radio how to adapt its systems — during which the commands may have already become outdated — a radio with a neural network can do it on the fly. Such a device is called a cognitive radio. Its neural network autonomously senses the changes in its environment, adjusts its settings accordingly — and then, most important of all, learns from the experience... Worcester Polytechnic Institute and Penn State University, in cooperation with NASA, recently tested the first cognitive radios designed to operate in space and keep missions in contact with Earth. In our tests, even the most basic cognitive radios maintained a clear signal between the International Space Station (ISS) and the ground. We believe that with further research, more advanced, more capable cognitive radios can play an integral part in successful deep-space missions in the future, where there will be no margin for error...
Our own effort to create a proof-of-concept cognitive radio for space communications was possible only because of the state-of-the-art Space Communications and Navigation (SCaN) test bed on the ISS. NASA's Glenn Research Center created the SCaN test bed specifically to study the use of software-defined radios in space. The test bed was launched by the Japan Aerospace Exploration Agency and installed on the main lattice frame of the space station in July 2012... Ours would be the first-ever cognitive radio experiments conducted in space...
During the tests, the cognitive radio clearly showed that it could learn how to maintain a communications link. The radio autonomously selected settings to avoid losing contact, and the link remained stable even as the radio adjusted itself...
Overall, the success of our tests on the SCaN test bed demonstrated that cognitive radios could be used for deep-space missions.
Our own effort to create a proof-of-concept cognitive radio for space communications was possible only because of the state-of-the-art Space Communications and Navigation (SCaN) test bed on the ISS. NASA's Glenn Research Center created the SCaN test bed specifically to study the use of software-defined radios in space. The test bed was launched by the Japan Aerospace Exploration Agency and installed on the main lattice frame of the space station in July 2012... Ours would be the first-ever cognitive radio experiments conducted in space...
During the tests, the cognitive radio clearly showed that it could learn how to maintain a communications link. The radio autonomously selected settings to avoid losing contact, and the link remained stable even as the radio adjusted itself...
Overall, the success of our tests on the SCaN test bed demonstrated that cognitive radios could be used for deep-space missions.
Simple (Score:2)
Re:Simple (Score:5, Insightful)
It sounds like a very complicated way to do something simple...
No, it sounds like a way to do things that would be fucking impossible otherwise.
Remember when you didn't notice framing issues in amateur photographs because the focus and lighting were screwed up too much to care about framing? And the people all had red eyes? Well, technology of the type these engineers are using got rid of that (for the most part), so you can now marvel at how badly framed an amateur photo is. It's the same here only with radio reception.
I don't know when Slashdot became Luddite central, but those who dismiss obviously useful technology because it uses a neural network for control functions are very misguided.
Re:Simple (Score:4, Insightful)
I don't know when Slashdot became Luddite central,
I don't think it has. What I think has happened is that many posters now an adverse reaction to certain phrases (e.g. neural network) because those terms have been associated with buzzwords or used in contexts where they don't actually apply. This is why buzzwords are a treadmill, some people recognize the bullshit and it's a turn off.
The aphorism, “Every lie we tell incurs a debt to the truth. Sooner or later that debt is paid.” comes to mind. Welcome to the payment part of the lies.
Re: Simple (Score:3)
Re: (Score:2)
Yea but can it call ''CQ TEST CQ TEST CQ TEST DE CALLSIGN... QRZ'' and log the contact so our beer drinking won't be interrupted by anything else but installing the antenna array and stuffing our faces during Field Day?
You know some club is already salivating over the possibility of top placement in ARRL published results.
Re: (Score:2)
It sounds like a very complicated way to do something simple, but at least some engineers were having fun with a somewhat new kind of ALE.
Actually it sounds like an interesting idea to implement a complex regulator that does not have to be 100% reliable. Sure, this can be done classically, but it would be a lot more expensive.
Adaptive? (Score:3)
A cool idea no doubt, but wouldnâ(TM)t the wordadaptive be a better term than cognitive? Is adaptive radio taken or something?
Re: (Score:2)
Is adaptive radio taken or something?
yes, it is: https://patents.google.com/pat... [google.com]
More neural net hype (Score:2)
You don't need a neural net to lock onto a signal with a shifting frequency. In fact you don't even need software, it can be done with basic electronics, otherwise the emergent space industry and satellite technology would have been dead in the water in the early 1960s.
Re: More neural net hype (Score:2)
Re: More neural net hype (Score:5, Insightful)
What is actually useful is people who really know the subject intimately having a look. The big problem with "people" looking at things and questioning everything is that you end up never being able to do anything, because questions.
Everything has a resource cost, including, yes, answering the varied questions that arise. This is where "burden of proof" is essential. If you're convinced that NASA are doing something wrong, then you need to know the subject matter very well, put forward a well formed and highly technical piece of evidence to say "This is why (x) could be better". Or "you're doing (y) wrong.". The burden of proof is on _you_ to provide evidence they're not working properly.
For every piece of work they do, I can assure you there are very solid business cases, with highly technical information. If you assume that they're not doing their job, you pretty much start from the "lets assume nobody is doing their job unless I check on them.". This is both paranoid, and impossible to keep check on. Unless you have a highly trained army of people with specialist knowledge.. And if they have that kind of specialist knowledge, they're usually working in the industry itself, trying to do stuff and make the world better.. Which leaves you with poorly informed people who don't know enough to know how little they really know asking questions about things they don't understand and tying up resource (or getting progressively angrier about why people aren't responding to their question which they're convinced is paramount, when it's in reality irrelevant).
Re: (Score:2)
Then
Re: (Score:2)
Re: (Score:2)
Well they can't even seem to build rockets anymore and need private industry to do it for them so who knows what else they've forgotten.
Re: More neural net hype (Score:2)
Re: (Score:2)
They can build them, just commerce can do it more effectively at scale. NASA is great for actually researching new technology and expanding the field of science. They're just expensive to push that resource at engineering well understood technologies in production.
So this is just a system working well. Research does the new (very expensive stuff), refines theory and makes something practical, then when it enters the 'tried and tested', commerce takes it up to scale up the production in a cost effective m
Re: (Score:2)
It could all be done in hardware alone, but it's not as practical as a digital solution in the 21st century.
Fact of the matter is, you could use a digital solution to develop your system, then decide if it could in fact be translated to a totally analog electronic system, design that, and the end product would in many ways be more reliable than the digital solution. But no one really does that anymore.
Re: (Score:2)
Re: (Score:2)
Little do some people know that even the most modern, most high-resolution and high sampling-rate d
Re: (Score:2)
I think the other problem is that complex analog electronics is HARD. Harder than digital eletronics (at the logic gate level anyway) and much harder than writing software, plus it requires a certain mindset who can think in parallel from the POV of the circuit as many things happen at the same time at the analog level. And I speak as someone who could never manage that :)
Re: (Score:2)
Radio isn't Trump and you aren't Mainstream Media (Score:3, Insightful)
Radio has a lot more to do with internet communications than most people realize. I don't think most of you commenting really care though. Radio "nets" have been around a lot longer than the Internet, and I'll leave it at that. Post something positive for while?
Thanks, Hedy (Score:2)
Reminds me of a story seen in a 70's/80's comic (Score:4, Interesting)
As a young comic book reader and fan of science fiction, this story reminded me of one seen in a short lived comic book series called "Space Warp" in the late 70's or early 80's.
In the story, a satellite gathered sunlight and beamed energy down via a laser to a ground station. It took precise alignment to get it right, and the deviation of the laser was catastrophic to those who were hit by the misaligned beam. The protocols in place were to shut off the laser if there was to be any chance of deviation (such as a thruster maneuver, etc).
In the story, space debris suddenly appears, and one of the main characters cannot get to the panel to shut off the laser before the impact. And from the portrayal in the comic book, it could be possibly seen like that fabulous scene in "Gravity" https://youtu.be/vKW-Gd_S_xc?t... [youtu.be] where an impact of their space craft is seen.
The character finally gets to the control panel, expecting to hear how thousands of lives were killed on earth by the wayward beam of death, and was shocked to find that the AI system on board had calculated and compensated for every impact and had kept the beam locked on to the receiving station on earth.
I object to the word 'cognitive' in this context (Score:2)
I'm glad... (Score:1)
..,that they used a specific term to describe this instead of just using the overly broad, watered down, and marketing weasel word term known as "AI".
Cognitive Radios == Software Radios that adapt. (Score:3)
To understand Cognitive Radios, you have to understand that the term Cognitive Radio came directly from Joseph Mitola's research in the mid to late 90's.
He was well known for being a software guy who got into RF. His early research was in software radios and the term "Cognitive Radio" was coined in 1999.
To understand why "cognitive radio" became a thing, one has to understand why "software radios" became a thing.
First, virtually all radio systems (even those developed now) are done by deciding the spectrum, modulation, and propagation methods up front. You do this so you can decide what the link budget (bandwidth) of your connection is, determine whether the signal is analog or digital, pulsed or CW, the modulation, demodulation, filters, mixers, noise floor, sampling, amplification, antennas, etc. This is why you typically see new wifi and cellular systems determined "by spec". The spec informs the hardware design and vice versa. This is also why you often hear about bandwidth allocation as a hot topic - not only is bandwidth fundamentally "squatters rights", so that organization and adherence to rules is important, the bandwidth allocation also fundamentally informs which physics are important and therefore the hardware of the radio system itself.
Second, the digital theory of information fundamentally transformed radio, and was in the process of transforming radio for several decades. There was a lot known about the digital equivalence of physical devices like filters, and phenomenon like noise, and ideas like channel capacity, but due to the nyquist theorem, it was actually very difficult to take advantage of any of the knowledge gained because computers were simply not fast enough to perform the necessary mathematical operations quickly to convert voltages to bits, compute, modify the bits, then convert back to voltage - especially at higher and higher frequencies. (Notice how your communications devices get faster and faster data rates over time? You can thank the fact that computing hardware is getting faster and faster.)
With the advent of fast computational devices (DSPs, mostly) it became possible to implement a lot of the theoretical advantages of digitized waveforms.
Third, RF designers started to look at the fundamental hardware building blocks of a radio, especially the filtering, modulation/demodulation. and mixing stages - which were often analog components that were "off the shelf" and fixed quantities, and created theories around how to create generalized digital representations of these devices. Once the fundamental building blocks became well understood in terms of how to build software representations of these devices, and hardware became fast enough to implement some of these software representations of once fixed hardware devices (in practice, this is still extremely primitive - even in 2020 terms), researchers began to earnestly look at how one might be able to create dynamically reconfigurable RF systems.
And the representation of these dynamically reconfigurable RF systems? Of course, it is done as a language, and therefore, programmed in software. Hence the field of "Software Radio."
Now that you realize how radio got into the software domain, you have to understand that the term "cognitive radio" is just any software radio that implements Mitola's Congitive Cycle, which is basically just a feedback loop, where a radio listens to its environment and then adapts its RF operation to best reach the receiver.
There was a program in 2012 called COMMEX (Communication through Extreme Interference) by DARPA that was ultimately won using a Cognitive Radio approach.
This is just a vastly simplified explanation. To really understand the concept of how AI can be used to steer the software radio system to receive receivers through difficult links / signal paths, you have to understand that AI is just a combination of linear algebra and statistical mathematics being used to solve objective functions. But now, maybe you have more to explore in terms of why AI and Cognitive Radio became a thing, and why it might be useful for this kind of situation.