Keynote Speech-1: SPECTRUM SHARING IN THE 5G & BEYOND ERA
- Finally, many people in the wireless industry are talking about using Artificial Intelligence (AI) in communications systems because with the new generations (5G and beyond) there are many parameters and these systems are too complicated to be manually configured.
- Domain expertise is the key to utilize AI in communications.
- spectrum sharing will still be a problem in 5G, 6G and beyond but 5G opened new bands that can facilitate the problem.
- Now there are licensed, unlicensed, and shared spectrums that AI is trying to use all in one system (cognitive Radio, Software Defined Radio, …etc.).
- 3GPP standard for the 5G which is called New Radio (NR) is immature about Over The Air (OTA) testing and especially about the beam steering.
- OTA of the mmWave proved that mmWave has sparse channel (only few multipaths between Tx and Rx).
Panel-1: WHAT 5G CANNOT SOLVE AND THE NEED FOR 6G
- 5G will be a big disappointment!.
- But it is a necessary evil to understand what we need to fix in 6G.
- we need to focus more on the core network more and not just the RF design.
- 5G false promises include:
- – Remote surgery with (1msec) delay: a lie!
- – Five nines (99.999%) reliability: a lie!
- – 100s of Mbps for all devices anywhere with mmWave: this is true but it will be too expensive that there will be no profit for the companies to deploy (so far)!.
- DSA: Dynamic Spectrum Access.
- We have to distinguish spectrum sharing from coexistence (google it!).
- WiFi and Cellular coexistence possibility?
- what can be achieved from 5G promises includes the high bandwidth or the eMBB which is the easiest to achieve. But the low latency (or URLLC) and the high density (or mMTC) are still not sure to be achievable with 5G.
- 100GHz+ can not travel for miles. So, one possible usage is sensing.
- mmWave have been used for sensing before communications.
- The main benefit of 5G so far is the reduction of the cost per bit.
- 6G will be about unifying our experiment of the physical, digital, virtual, and biological worlds (realities).
- 5G requires (700,000) small cells to be deployed in the US for full coverage.
- Should we think about possible applications of the 5G and beyond?
the answer is NO! because the wisdom says “Just build it and they will come”. - Fully phase shift arrays (fully digital beamforming) that are currently developed by the DoD consumes from 40 to 145 Watt for each antenna for only one phase of 3 phase process for the fully digital beamforming and large amount of heat emissions.
Nokia Keynote Speech: 5G E2E NETWORK FOR THE FUTURE OF INDUSTRIAL AUTOMATION
- As Nokia did not realize that having (43) phones with different operating system for each of them was the end of Nokia handheld devices, Also, Apple does not realize that the shift in interest into the Augmented Reality (AR), edge computing, and cloud computing, is the iPhone!
- Wireless packet drop rate is (2/1000) so, the reliability is (99.8%) and the need of the IoT reliability to be 99.999% or even 99.9999% is very hard.
- We need new architectures for End to End (E2E) security, E2E slicing, E2E 5G, …etc.
- There will be 6 dimensions to go:
spectrum.
data
efficiency
computing
space
energy. - 6G network will be a network with a 6th sense. A network that can read your mind and guess what you are about to do (telepathy).
- Information and Communication Technology (ICT) is going in the direction of creating a digital twin of yourself.
- your medical record is a primary version of that.
- Google is trying to create a version or twin of you by studying your behavior to estimate what you like, want, or do next.
Panel-2: TRUSTING MACHINE LEARNING AND AI IN WIRELESS
- Can we trust AI and Machine Learning (ML) in the wireless industry? if we know that it can suffer from:
– Bias.
– lack of explainability.
– Privacy and Security issues.
– Robustness.
Compliance with regulations. - Biggest Wireless Infrastructure companies in the world now (from the larger to the smaller) are:
1- Huawei
2- Ericsson
3- Nokia
4- ZTE (Chinese).
5- Samsung. - This company is great startup idea https://www.deepsig.io/
- Supervised learning will never give us a trustworthy ML/AI.
- Why ML/AI is still suffering in the wireless industry? because we are following the wrong people (computer vision and statisticians). Instead, we must go back to our mathematical roots (domain expertise) to get better models and better trustworthy ML/AL.
- Data bias vs model bias? google it.
- No reproducibility=== no trust.
- Any trust needs oversight from a third party. So, who is that party in Wireless industry.
- How can FCC (or any other party) approve any device that uses ML/AI that change modulation or any other aspect of the systems while working to improve the performance?
General Notes:
Listening to people from different backgrounds and focuses is always useful in broadened your view about any issues. They rarely agree on anything but always enrich the topic with different insights.
When working for an organization or when you borrow some of their contents in your slides, it is necessary to mention clearly that what you say is your opinion and does not necessarily reflect theirs.
PLA: Priority Access License.
NoC: Network on Chip.
Short story: the students of the Software Defined Radio graduate level class in one of the Chinese universities has to hack the DoD to get (A) in that class! this was the case in 1993.
Instructions Set Architectures (ISA) and the registers reuse is the main reason for increasing hacks.
Pure dataflow machines is the answer.
FPGA can do 3 times as a CPU and consume only one third (1/3) of the power and that is why most data centers have FPGA more than CPUs.
Federated Wireless–> google it.
RF NOC –> google it.
5G current spectrum: 3.1-3.55GHz, 3.7-4.2GHz (c-band), 6GHz (1200 MHz for unlicensed use), 7GHz (1275MHz for federal spectrum), 24, 28, 39GHz.
Pure dataflow does not reuse registers.
Rainbow tables (brute force tools) + big data = breaking the code eventually.
More info. about the dataflow is available here: http://get.directstream.com/
Modulation classification and identification is a hot topic to prevent civic/military interference, different applications interference, shared spectrum applications coexistence.
LAA is the solution to allow WiFi and LTE to coexist (listen before talk).
SRD is the future of wireless research and systems.
ROF: Radio over Fiber.
Many countries are doing great research, but it is who build the products that will win.
More details about the Dyspan 2019 can be found in this playlist:
