Ok. I want to welcome you all for for this session.
So my name is Janskinen, and I'm heading product marketing at Mendra.
I think on the agenda it was saying that I'm heading marketing,
so I want to thank you for the promotion for for my exhibition.
So, um, quick background, Morra, so we are a software company for Finland.
We do mainly solutions for network monitoring and network management.
For the next half an hour, I'm going to be discussing about how to utilise AI and
crowdsource network information to improve reliability of critical networks.
If the clicker works.
The clicker is not working, so. OK.
So let's first establish what does crowd sourcing network information actually means.
So it basically means collecting network coverage and kind of user experience data
directly from the end user devices, and this is actually a fairly commonly used approach on
commercial networks.
And it's typically done via this OTT application.
So it basically means installing a dedicated application to the mobile phone that does the
collection. You can also do this by kind of having an SDK,
so basically a piece of software would be kind of embedded to the customers or organization's
own applications, so then the collection can have a background and no dedicated application
needs to be installed.
Or it could be done via third party applications.
So for example, there's sort of commercial application providers who does actually some
sort of network coverage monitoring, and then they kind of sell this data for operators or
other user organisations who are interested.
So The question is then why hasn't this been done on critical communication networks,
and I can think of three main reasons why it could be so.
So one is the the kind of the the PMR network with the technology,
it is actually in for example many Tetra terminals.
It is possible to collect the crowdsourced or kind of network information remotely,
but typically the APIs are kind of um proprietary, so the solutions would need to
support it. Second reason is that as the networks are
dedicated, there is kind of less users roaming around, so the kind of collection would need to
be planned carefully and maybe in some cases a radios would need to be placed on the field
which are kind of dedicated for the network collection.
And then finally is the question of the data sensitivity.
So of course as an operator, for example, it's easy to collect from your own own field team,
but then for example collecting from other organisations there definitely would need to be
an agreement in place and the data would need to be happen anonymously.
I still see that there's clear benefits for crowdsourcing network information and critical
networks as well. Uh, and I, we, I think we all know that,
uh, in many cases operators, for example, and organisations do some sort of a drive testing
already. So they are interested of coverage in different
areas, but testing can be quite kind of resource heavy,
and typically it's more of a tool to kind of investigate an area where you already know that
you have a potential problem with cloud with crowdsourcing in large scale,
we are kind of able to get this data in real time and kind of immediately so we can kind of
react even in proactively.
Um It's of course obvious that some kind of network monitoring is done on all networks,
and I think the most common one would be to have some sort of an NMS,
so network management system on your network, and these type of tools are excellent when it
when it comes to monitoring things like alarms, key performance indicators,
and equipment, but there could be problems that are not visible for NMSs.
So NMS is great for, as I mentioned, think like status monitoring or health monitoring,
basically monitoring the network from the kind of network's perspective,
whereas in crowdsourcing it's more looking at how does the end users experience the network.
What's the status of their. Services.
So we crowdsourcing we can get information such as coverage,
the serving cell. We can see what the neighbouring cells are
visible for the user and also we can, for example, the mobile network conduct speed test
so we understand the latency and the coverage in real time.
And it's these type of measurements that can be affected by these external
factors that are not visible for NMS. So let's say you have a snowstorm and your
antenna gets slightly tilted.
So what about if the actual end user device is slightly kind of broken and it's getting worse
coverage with crowdsourcing, we are able to see all kinds of problems and really understand how
how are these users experiencing the problem.
Other thing with crowdsourcing is that it's able to kind of detect very subtle changes.
The reason is that the collection happens in real time and it happens together with the
location, so we can really kind of pinpoint the problem areas and then react in real time so it
can be used also for like tactical network monitoring on incidents and um.
And it can, uh, it, it can also kind of provide information to reactive,
kind of do a proactive reaction. For example,
when the data is feeding the machine learning so we can get anomalies that we can react
proactively. I'm not saying that crowdsourcing is
necessarily a tool to replace NMS, but I think it's a very good tool to provide a completely
new angle to understand your network coverage, and together with NMS,
it can create a much more holistic view for the operators and user organisations about the
network. So let's then look at how crowdsourcing can
help on monitoring specifically for missing critical broadband.
And how we can help on this like transition period to missing critical broadband.
Um, I guess it's like safe to say that missing critical broadband and digitalization is kind
of expected to revolutionise the mission critical communication in a way.
So there's a lot of new technologies being taken and use like mobile phones,
cameras, sensors, and, and all of these all of these devices can generate a lot
of data. And of course having such important services on
top of these critical networks will also increase the demand for the network robustness,
the network coverage, and the network availability.
So these things are expected to be available all the time to kind of.
Be able to do this kind of missions successfully.
At the same time with the introduction of kind of missing critical broadband,
a lot of the organisations will kind of move into this model of hybrid network operations.
So there will be organisations will be using, for example,
Tetra and LD at the same time. So understanding what is the actual coverage
will actually become more complex, and I think crowd sourcing can be a very cool tool for
these types of use cases because it kind of cuts through the complexity.
So if you look at the three scenarios which we are clearly seeing a lot of the countries being
utilising on this kind of moving the misal broadband, the first one is that for those
countries who have the budget and the willingness and the frequencies available are
looking into maybe building a dedicated network, and I think crowd sourcing here can be a great
tool. For the operator to kind of audit the network
in larger scale to understand is the networking, is the network matching on what has been
planned on the planning tool and kind of do the adjustments while kind of building the network
and and kind of tuning it.
Another model is where kind of commercial radio networks are planned to be used for critical
communication purposes, and in these cases, many times the existing kind of a tetra
operators, um, the public safety tetra operators, they,
they kind of become like an envy and nose on this network where the radio network actually
then comes from the actual commercial operator.
And crowdsourcing can be a very good tool for these operators,
for example, to audit the network to investigate, to see where are the possible
problem areas, maybe use that data then on conversation with the commercial operator to
improve the network while the kind of process of kind of deploying goes on.
And with crowdsourcing, of course, the good thing is that you don't really need access to
the back end infrastructure. You can actually source the data simply by
having a SIM cards available to access the network.
And then final model is kind of this roaming, roaming across networks so you can have a SIM
card that kind of utilises multiple networks.
Here, of course, the problem can be that you could have multiple network monitoring tools,
each kind of monitoring individual networks, whereas with crowdsourcing we are more looking
at the problem or the coverage from the end user's perspective collected from the end
user's devices, which means that we are kind of really able to cut through some of the
complexity. So, who can we then source and how?
As I'm going to briefly mention, sourcing your own field force,
let's say as an operator you have a field that is kind of managing the network or user
organisation, you have your own field, of course you understand their needs,
you understand what services they are using.
It's, it's easy. Also, you don't have the problem of necessarily
the data sensitivity and deploying the actual applications or configuring the devices is
also kind of possible.
The second, of course, is then use other organisations on,
on the same network.
And here. It becomes a little bit more complex,
but for example, we are already seeing some of the operators now putting agreements in place
with the user organisations to allow collection also from their devices,
and this is typically happening with, for example, the mutual agreement.
In some cases the operator then can also provide some of the coverage data for the user
organisation as part of their control room if if their operation is highly critical.
Of course the collection can happen kind of very anonymously,
and there's multiple ways to kind of guarantee anonymity.
So you could, for example, do the collection only of course first of all,
you don't store the information of the person you are collecting on the device.
You could only do collection on a certain time frame.
You can use geo fences to only allow collection in certain locations,
etc. So there's multiple ways to tackle this issue.
And then the final is actually the public. So now if you think about starting to kind of
using this commercial network, we can actually purchase this data or kind of crowdsource the
public, and that data is as, as, as, as kind of equivalent as as any because they are kind of
using the same network.
So how the collection can then be done?
Well, from Tetras side, um, it's actually fairly straightforward.
Tetra, a lot of fun example of PMR, let's say a lot of the Tetra devices,
for example, do support throughout this kind of a remote collection already.
It's just a matter of having those proprietary interfaces available on your solution.
It's a matter of configuring the device to allow the collection or to enable the
collection, and then maybe in some manufacturers you might need to purchase a
specific licence to allow the collection, but it is possible,
and like I said, you can already have the infrastructure available to do so.
From the mobile, the mobile phone side, I think this has at least from our experience been
quite straightforward because it's about typically deploying an application and in many
cases there is some sort of an MDM in place, so the application can be simply deployed via the
MDM App Store, for example, all devices all selected devices on the.
OK, so now we have established How to collect data, from whom to collect data,
and what are the benefits. So now you're getting this data in.
What next? Of course there are probably multiple models to
do this, but this is a 11 way of kind of kind of reading the data and transferring that data
to actual meaningful insights.
So first thing that typically has been done is to establish some sort of continuous collection,
and this is important so you can start creating a baseline of the network.
This is also the basis of of kind of been doing large scale to start building that kind
of understanding that is the network matching what has been planned on the planning tools.
And kind of identifying the problem areas.
Next we're gonna kind of feed in this data to this match machine learning algorithm which
kind of starts to build a view of what the normal looks like.
We typically call this like creating a fingerprint of the network.
And once you then have the fingerprint available, you can start understanding if
there's any kind of anomalies.
For example, anomalies that can be easily found is unexpected good service,
unexpected bad service, unexpected handovers.
You can also detect things like sort camping, cell tracking.
There are also some security benefits with crowdsourcing.
You can actually fairly easily detect things like jamming or for example,
MCs and fake base stations. So we can also improve the kind of the network
security. And the final, final point where we can use
this crowdsourcing is kind of this tactical network monitoring.
So let's say you have a larger incident or natural disaster,
you can actually easily crowdsource and see how the network is performing in a very specific
area or or for the specific fleet.
Um, this type of tactical use case we have seen, for example,
by some operators to use when they have larger events.
So they are bringing in, for example, mobile base stations.
So with crowdsourcing you can very quickly understand how the network has changed,
how it's now kind of serving the becoming events, or possibly even do this kind of
monitoring throughout the event. So you can,
for example, crowdsource all the event radios or mobile phones during the event and
constantly see how the network is performing as the event unfolds.
Then let's look at some of the challenges.
I think we have discussed many of these, but this is maybe kind of big to recap and also
maybe overcome some of the problems.
So like I said, I think the major one is agreeing the data saving.
So having that agreement in place together with the operator and user organisations to allow
collection. Understanding how to do this anonymously,
like I mentioned, you can do it in time frame, you can do it with the geo fences,
or you can have this like a mixed model which we have seen for some operators where they
mostly crowdsource their own field force.
Um, then they place some radios in strategic places to get the coverage data on those areas
and actually for the mobile application they are simply providing that for the user
organisations so that they can actually create a ticket from the app.
And once they create a ticket only then we crowdsource for like the next 5 to 10 minutes,
so we immediately get the data from the ticket area that how was the coverage to to to kind of
to be combined with the ticket information.
Next is the deployment of those applications like I mentioned with Tetra.
Uh, the challenge is to support the proprietary interfaces.
Also, in some cases, the problem has been that the radios are already on the field,
and if you need to do the configuration, so how do you kind of recall some of the radios for
configuration purposes and like I said, then it's a matter of of licencing to be in place.
For mobile application side, we haven't really seen much bigger problems as with MDM it's
typically fairly easy to deploy the crowdsourcing application.
Then it's about tuning the algorithm.
With crowdsourcing, you can actually get a lot of data in a very short amount of time.
So you really need to kind of make sure that your algorithm is tuned in the right way so
that you don't end up in a system which looks like a Christmas tree and it's full of alarms
and all the imported information get crowded or buried and kind of beneath the the real kind of
problems. And the final one is kind of having the
organization's own processes in place.
So whenever you introduce a new system, it's very important to train your people to
understand the data and plan the actions, what to do with the data,
also how to kind of look at the network holistically with crowdsource data and then for
example NMS together, um, and maybe one higher level problem we have seen is that.
In many organisations now kind of the LT team might be separate from the Tetra team,
but from end user's perspective, they just want to have a good service and their kind of
services available.
So how to work within these teams to kind of guarantee the service when multiple network
technologies are being used.
So, to summarise, Uh, the demand.
For kind of The demand for service availability will simply going to be increasing as new
technologies and digitalization is being introduced.
This will put a lot of pressure for the tech for the network,
as everybody's expecting a robust network that will have high availability and very good
coverage. So ability to detect issues in real time
will become more important than ever to be able to deal with issues before they become urgent.
So we think that with crowdsourcing, it can truly help as the information is received real
time. We can detect a lot of the problems proactively
and also because it can help a lot in this transition period to missing critical broadband,
building a more holistic view to the network.
So, I want to thank you for, for participating. I think now would be time if you any of you
have any questions.
If you want to talk more privately, uh, um, you can visit us in our booth,
A 99. Mentora group, so we have some of the
crowdsourcing solutions available there as well to, to see.
And discuss about the topic.
I want to thank you all for joining. Thank you.