Red Hat OpenShift Automotive 2022 OpenShift Commons Gathering, 6 Apr 2022

Previous Meeting

⏯

youtube image

►

From YouTube: Safer Driver Human Driver Perception Platform Bojan Mrazovac (NTT Data) Ramakrishna Yekulla Red Hat

Description

Safer Driver: Human Driver Perception Platform That Prevents Accidents, Before They Happen
Bojan Mrazovac (NTT Data)
Ramakrishna Yekulla (Red Hat)

OpenShift Commons Gathering on Automotive
April 6th 2022

full agenda here: https://commons.openshift.org/gatherings/OpenShift_Commons_Gathering_on_Automotive.html

A

Okay, so let me introduce myself first as being the guest here, so my name is blam brazos, I'm responsible for research and development department at entity date, romania. uh Basically, if you go to the next slide, my company is a partner with redhead and we are belonging to entity. Group entity group is one of the largest uh service and it providing company from japan. It's a global corporate and basically companies.

A

Let's say a sub company from the group from where I coming is ntt data, so basically identity data we are focused mainly to service providing and some of our key focus areas are actually the automotive uh what we are going to cover today as well: uh ai blockchain cybersecurity uh data and intelligence and basically uh other emerging technologies that can generate a big technical footprint at today's market.

A

So ranki back to you.

B

So we've been partnering uh with ntt as a system integrator over the past 18 months and uh specifically in the newer areas which red hat has come into. uh uh That would be the edge and openshift and kubernetes being rapidly adopted into these spaces uh entity uh we've.

B

We found a perfect partner in entity where there are lots of industry ledge cases in the automotive use cases uh which can be hosted on our red hat openshift container platform and uh it being a true hybrid solution, and uh it being uh uh uh it I mean taking your taking your workloads to a multi-cloud and also the hybrid cloud way uh is the way to go forward and uh over the past uh 18 months, we've been working with various projects which were traditionally written for the embedded site are to be onboarded onto the I.t side, as most of the previous speakers were speaking about how to integrate uh developers at uh uh and uh car or uh or automotive oem onto reddit platform.

B

So uh we kind of use various uh accelerators in order to adopt these uh cloud and these practices uh these solutions onto the openshift cloud.

B

So one of the ways in which we uh started integrating is we have these validated patterns, uh because there are lots of ways in which red hat has built expertise over the past one decade about linux containers, uh devops practices, how to how to write enterprise solutions uh on an enterprise grade.

B

Kubernetes uh distribution like openshift container platform, that there were many uh aspects of various industry verticals which needed to be adopted, uh uh are adopted in an opinionated way in terms of what would be the right way of getting these solutions on top of openshift container platforms.

B

uh So uh working with the various engineering and the field solution, architects, they came up. They they came out with various validated patterns in which you can you can use the open source software which is built into these validated patterns, as as code like in the words of liner store walls, it's more of a talk. It's talk is cheap.

B

Show me the code and many developers wanted to look at how to integrate their solutions and how to integrate their edge solutions on openshift and uh uh uh in general, in the cooperators platform, and many of the developers are eager to get their workloads predominantly, which were in the automotive in the industrial use cases to use a modern distributed software like kubernetes in their production environments.

B

So these validated patterns and the accelerators which power these validated patterns could be one of the first steps in which both the partner developers and the customer developers can start accelerating, uh can start adopting their uh code uh to be uh to be from staging to uh production environments pretty soon and on boarding. Developers is easy uh with these validated architectures uh and onboarding uh projects and getting your projects.

B

uh Two products uh is also uh is very helpful with these and uh all the software which we are talking about like, for example, here the accelerator which we used for this particular product was the red hat bobby carts. uh So uh this is a vehicle uh accelerator. It's a vehicle simulator in which uh there are various red hat products, not just the linux containers. uh Also, various red hat run times various uh various kubernetes operators, which are used to get uh to get your solutions to uh to staging and production.

B

So uh it's beyond just the just the vehicle workload. It's also how? How are you going to take it to production across multiple clouds? So your your your customers could be using aws, they could be using uh the google cloud or they could be using uh azure or they could be having their own private data centers. uh So getting these applications and processing these workloads at any uh at any of these endpoints. uh uh These accelerators make it possible, and also another thing, is, uh we believe in doing uh doing things in the open.

B

All source code of this is available on github and uh and on. Most of these are reproducible to handle most of the uh automotive and the industrial use cases, and we are also a part of various uh collaboration like elisa, is one of the one of the foundations where we uh actively participate, and we also collaborate with many uh customers and partners like in this case, uh ndt raters, helping us onboard their automotive expertise and their work, automotive workloads on the openshift container platform.

B

uh So this is one of the uh one of the ways in which we use bobby car to accelerate uh an entity in car solution uh for human driving, uh human driver perception platform and one of the gateways which it uses called the multi-sensory analytics platform gateway onto openshift uh all the code. What we're seeing here is is is is workable and uh the integration platforms all of these source code is available on github. We will share the links later uh so uh bobby.

B

Car uh itself is an accelerator, and uh we we use this because uh you once you're adopting a enterprise distribution like kubernetes uh uh uh like openshift you're, trying to integrate various redhat components, not just the operating system. uh You have messaging queues, you have time series databases, uh you have the authentication mechanism. Again, you need to build your ai ai ml uh pipelines on top of your on top of your existing cloud and platform solutions.

B

So uh what bobbycard did is was it was opinated architecture of how how these uh uh in-car solutions can be onboarded and what are the? uh What are the various software components? You would need from the vast cloud native compute foundations landscape, uh because there are quite a few. uh There are quite a few software which you need to choose, uh so what uh bobby car does is it accelerates the process of you, focusing on your own automotive code and red hat, taking the workload and distributing it across multiple clouds.

B

uh We also use various various operators uh from the open data hub and uh and various kubernetes operators in order to facilitate or onboarding of this process. um So I will stop over here and go to the next slide uh boyan over to you.

A

Okay, thank you ranki. So, during the following slides, I will try to explain this concept of human driving perception why we are doing it, what actually we are doing and why we think it is important.

A

Basically, if we take a look at connected vehicles, market trends, which was published by gartner for the last year, we can see that a lot of topics around metaverse and human-centric design has been raised nowadays, so autonomous vehicles market is a little bit stunning. This is something that many other analysts has also reported, and according to that, we need to find a new way.

A

How to approach the clients, and one of the most important parts is definitely to understand the clients needs and to understand them in real time, while the driver and passengers are inside of vehicles.

A

uh If you go to the next slide, uh basically not only from the trends on the market, but also the regulations that we've seen uh have been issued by european union, but also by the u.s nate they're all requesting to have this kind of mandatory safety features which needs to be implemented and which needs to be supported by the vehicles.

A

So if you take a look at the list of such features, you will see that basically uh different solutions which are supporting for a driver to prevent from accidents or at least to recognize early signs of possible accident and to try to improve the behavior of vehicle which will be correlated with the perception of humans. This is something that definitely we want to have basically at our next slide.

A

This is something where we want to show exactly the concept so ramke, if you can just uh move to the next slide- and this is exactly the concept of human centered approach that we are proposing here- what we want to do, we want basically to give more feedback from the client, so we know the famous hazel formula, exposure, controllability and severity, so these are three main factors that we are calculating today uh to understand the ac level of certain models which are developed according to functional safety regulation.

A

But what is missing here is human sentiment and perception of certain features. You can imagine situation that emergency braking will do its job, so the vehicle will break on time, but how smooth it will be. Will it stress people inside of a vehicle uh how it will uh cope with their emotions and level of stress? This is something that, unfortunately, at this moment, we do not have many possibilities to understand, also lane assistance, warning and similar operations and similar features by the vehicle.

A

We can just measure the behavior of a vehicle, but we cannot measure the human feedback for that, and this is very important because at the end of the drive at the end of the test drive, you can ask people to fill the questionnaire, but in most cases it will not represent uh the exact moment in time when certain manner happened and when the corrective operation needs to be taken. If you go around to the next slide, this is exactly uh the solution that we are proposing.

A

We are taking several different parameters all together and we are applying it to cohort of vehicles, meaning that we want to understand the signals coming from the canvas we want to understand the human behavior and facial recognition to understand the emotional level. We want to classify the objects in front of the vehicle.

A

We want to know the exact context in terms of weather and in terms of the road, uh through real-time navigation, to understand exactly what is happening and why this situation happened that basically uh produced certain behavior from a driver or other passengers, and this is exactly if you go to the next slide shown in this short sequence, how we are collecting these data and how we are trying to understand the vehicle to collision risk, which is happening, for instance, during this specific manoeuvre.

A

I must mention here- and I would like to thank to avl uh company, which is also one of our partners and which uh supported us for collecting this data uh during the large data collection and the large experiment that we uh did in the previous period.

A

So if you put this all together and if we put it in the context of openshift and red hat solutions, we can apply it to the large cohort of vehicles, and this is exactly what we tried to achieve through this concept. Ranki, if you go to the next slide, please.

A

Basically, uh the whole solution is very probabilistic and we are trying to understand. Basically, our machine learning model has been trained to understand different signals and their change in time. So just from these signals, you cannot understand if you take one signal outside of other. You cannot understand the exact situation and you cannot predict a certain manner or event, but if you put them all together and if you train the model in a way that, by monitoring in our case it's around 14 signals at the same time, what is their distribution through time?

A

We can exactly recognize the patterns that in the next 500 milliseconds to one second, a certain situation.

A

A certain maneuver will happen during the drive, which will give us the possibility to check what the drivers are doing and, if necessary, we can do additional corrective operation uh through the monitoring of vehicle, but also we can understand how people in general behave in this situation. This is very important because not only to do corrective operations which are correlated with human perception, we can also do the benchmarking with certain features, and this is what we are trying to achieve.

A

So, as our future cases, if you go rank into the next slide, please uh this is the goal, and this is something that we are achieving all together with our colleagues from redhead.

A

We are trying to apply it to a large number of vehicles, so not only to monitor one vehicle but to apply uh the joint understanding, the federated learning, which will help us to improve the general model and, of course it will keep safety and security of the information's exchanged between a vehicle and between the exact uh device, which is in vehicle collecting information in real time and the global knowledge which is stored on a cloud. And this is something that we believe could help to prevent.

A

Even the hazards and uh accidents in future, because we will try to understand, we will try to put the human in the center of happenings in the center of drive and that way we will try to make much better interaction between a machine which is, in this case, of course, vehicle and the human who is driving or being driven in autonomous vehicle.

A

So that would be all ranking back to you.

B

And if you're a customer or a or a partner or oem, uh we are happy to demo uh demo. uh The human driver perception platform rtu also here so uh it is. It has been hosted uh on the github project of of the bobby car project and then uh you could you could if you already have a openshift cluster, you can just uh run this project and get there uh and and try it for yourself.

C

Thank you so much um for joining us today and ramakrishna this. This is great. I am um almost afraid to buy a new car until all of this stuff is gets ready um because it's moving so fast. I just don't know um what I should do next.

D