 Hello, our next speaker is Mr. Gaia, and he will talk to us about OpenADX. Yeah, hello. Good afternoon. One o'clock. So hopefully I don't make you fall asleep. Let's see. Typically I was trained to give the point of this talk in two minutes, so I have a bit more now. Let's see what is this all about. OpenADX, Open Automated Driving Accelerator. That's a great name, leveraging open collaboration and open source to accelerate development of automated driving. What's this all about? So the idea basically is, yeah, we want to have automated driving on the road, cars driving autonomously, and there is a lot to do on this way. And up to now, many companies do this on their own or in small collaborations. And the idea here is to bring this to open source in order to, yeah, decrease cost and so on. But I will go into detail. First of all, who am I? My name is Lars Gaia Blaumeiser. I'm with Bosch Software Innovations. That's a subsidiary of the Robert Bosch company. And there I'm in a division or a department that's concerned with open source services. Something about me. I'm passionate software developer. I did developing Eclipse applications for the last 10 years, mainly within Bosch for internal users, but also within another subsidiary providing tools also to the whole automotive industry. I changed the department last year and now our strategy or our mission is basically to bring in open source ideas into such a company like Bosch. So a big company typically involved in producing a lot of stuff, small devices for cars, for other stuff. And yeah, in pushing the idea into such a company, also things like creating business with open source. If you're an IP patent oriented company who sells pieces of hardware typically open up to this new world, that's a cultural change. We are looking for how to push this. So it's a big movement within Bosch, but so we concentrate on this open source activity. Another point also is that this comes up with operating systems that we develop as open source. So combining this open source development often ends with you have a piece of software that you can use but bringing this also into the company. So these are things I'm interested especially. I talked about our organization, so open source services on a very broad range coming from open source management, legal issues, infrastructure to handle this in a big organization. Ecosystem building which I will talk about now in the remaining of this talk, but also software development, project setup, project management. All this is offered by us, our group. Okay, automated driving, what's so special about it? So Bosch does software, provides business control units for cars for decades, but we have now automated driving that come a lot of new issues, new challenges that have to be looked at. So if you look at a modern car, it has ESP, it has perhaps an adaptive cruise control so there are already some autonomous things, parking assistance, where your car reacts without driver impact. All these things are already there and all of those have some perceiving so they have to recognize the situation, they have to think about what to do, do I have to brake, do I have to steer, things like that and then act. But when you look at those systems, those things are pretty small and we have this acting part that's the traditional thing which you already have in your car. So an ESP knows how to brake a certain wheel in order to keep you in the lane, stuff like that. With automated driving, we get a lot of technology on this former part so we have now perception that means we have to really understand the world. We don't have to only measure that perhaps we go out of the curve, we have to do something to stay in line but we have to understand what are the objects there, where are people, where are cars, what are they doing, where they are moving. We have to also locate ourselves pretty precisely. I don't know what your experiences are, I have a pretty old car and my navigation system sometimes thinks I'm 500 meters away when I start so it takes a while to understand that I'm really there where I currently am. This is not an option of course when you want to do automated driving. Also this whole planning, deciding what to do, I once saw a talk, there was a picture of a roundabout, a complex roundabout with perhaps five lanes and eight lanes used by people and it was a situation where someone just bounced off this roundabout. He simply as a human driver didn't get into it and this is also something that has to be solved by autonomous driving so even such complex situations where even human beings have problems getting it done have to be solved by autonomous driving. Another very nice example is of course an autonomous driving system has to stay within the rules. Now they go along a road, it's not allowed to change the lane and then there is something blocking the road, a car that has broken down. So what to do? How to teach the system that it can now break the rule by changing the lane to get around the obstacle. Such things are really new and they require hardware or the systems within the car which are not there so it's not micro-processors anymore, it's really micro-computers placed into the car which have a lot of computing power and all this stuff, all this has to be handled and of course new technologies like for example machine learning to get this perception done. So that's the problem area and that's why it's something new and building such systems, creating those, developing those systems is really a very complicated development process which includes new technologies as I said and it incorporates a lot of software tools that have to interact somehow and what typically happens is there are big departments within such companies who take those tools and integrate them because they're not used to work together and this really costs an awful lot of time. So what is the idea of OpenADX? We want to get this development process accelerated by simply cooperation, by collaborating throughout the industry to get an ecosystem running that solves those problems for the industry, not as it is done nowadays in each company on their own. What we think, what we manage with that is really accelerate the time to market, increase the efficiency of the toolchain because if we share or spend the money together on a common solution we get more efficiency into this toolchain than by solving it on our own and getting resources free which can really concentrate on this automated driving solution and not only on the infrastructure. So yeah, we built up an ecosystem. How did we do this so far? Of course Bosch is in a big network of companies and I mean the car manufacturers are our customers. We have suppliers who deliver to us parts but also software tools, stuff like that. We are working with universities to bring up research projects to come further on in technology, stuff like that. So we identified what are potential target partners in the industry to talk to. We created a set of workshops in which we refined the concept I will talk about later and we are currently establishing an open source community in which we want to work on this and attack the problem we face here. So I talk about partners, what are those partners? What is the state we have right now? Initiated was this by Bosch and Microsoft interestingly. So not only an automotive thing. We talk to many companies and we have a lot of interested companies there be it automotive, from the IT, tool vendors, technology companies, universities, research organizations. So it's really a broad range of interested parties. We have a website and companies or organizations can enter them on the website. So we have officially listed there, Bosch and Microsoft. We have Itemis, the German Aerospace Center, Dassault Systems, MathWorks, Electrobit, Renesys, the French Nuclear Research Center, ZF, Tier 1 in Germany, and cases Deineberg, also a simulation provider. So the sequence here is in the sequence. The sequence they were entered on the website. No other meaning behind that. We have additional partners. We talk also to OEMs or car manufacturers if you miss them in the interested list. But since Bosch is a big company, so are the car manufacturers and this idea simply takes its time until it gets into the organization and really leads to a participation, an official participation listed on such a website, even putting your name there. So we talk about those, with those companies about collaborations. So we are here at an open source conference using, and so I want to talk about why we use the Eclipse Foundation as a vehicle for OpenATX. First of all, I mean, Bosch is a strategic member of the Eclipse Foundation, so it's a natural choice. As it is of now, we currently have no legal setup defined. So we work loosely coupled in workshops and setting up things, and we realize testbeds. I will come to that in a second. But it's already supported. So we have Ralf Müller being on our workshops. We have our official website is on the Eclipse Wiki, and we plan to host the stuff we do on Eclipse Labs. So why Eclipse? A company like Bosch has simply its issues with this openness, and then they need some legal backup and some trusted environment to work in such an environment. And that's exactly what the Eclipse Foundation provides. So we have this concept of industrial working groups, which is very suitable for exactly this setup that you have to want to work in a set of companies. We have the open source contribution management, the legal support of the Eclipse Foundation, which is also very helpful to convince people in a management at Bosch that this is the right way to go. The Eclipse public license, which is also very industry friendly from our point of view. And there is a lot of experience in the ecosystem setup and management where we profit from, where we get direct support by the guys from the Eclipse Foundation. And another important thing, and I hope Mike isn't in the room, the Eclipse Foundation has a good representation in Europe and North America, so not only the U.S., that's why Mike would object to do this cross-continental. So we have, if you look at the interested list, there are many companies that don't have a strong focus in Europe. I mean, there are American companies like MathWorks, but it's basically the European part. But still, this is targeted not on the European industry but throughout the industry. And we already were in talks, so parts of the companies we talked to are based in the U.S. And we also had talks with companies from Japan and other Asian countries. So what happened so far and what's our current plan? The initiative started roughly in quarter two last year. There was a first workshop held in Redmond with Microsoft and there were basically American companies involved. In early August last year, we had then the IAA, the Automotive Trade Fair in Frankfurt, where we pushed the concept and explained it to a lot of potential Bosch customers. So we were at the Bosch site there and discussed with our customers about this. We had then a first workshop on doing something concrete on the EclipseCon Europe last year and we are working on those topics still. We have now, end of February, the Bosch connected world, where we want to push the topic further where also there will be some announcements officially done by Bosch and partners. We will also do a hackathon there which shows some technologies involved in the initiative where people can experiment with those technologies. And we expect to have the first demonstrator soon. So we had recently a workshop where we planned what to do and I will show you this later in the talk where we built up a demonstrator in a dedicated topic we have here. So I think I explained why it's interesting for developers of automated driving systems, car manufacturers, OEMs, automotive tier ones, that's first suppliers, first level suppliers like Bosch, they develop really automated driving and they need this tool chain, they need this development environment why it's interesting for them, but I mean a good question is in this market typically we have big tool vendors like mathworks, like vector informatics and so on. Why should they be interested in such an initiative? And we believe they also have a lot of, we discussed with them and what they see is that it's also helping them because what happens now is every company has its tool chain and they have to integrate in an awful lot of tool chain spending a lot of effort into integration often don't want to do this, so there is always a discussion ongoing. So also for such companies it's easier if they have a central tool chain definition to say hey we integrate in this and you can use it out of the box using those technologies so it's also for them interesting to be part of the initiative. Also semiconductor suppliers who have often also tool chains which they can simply plug in and can provide to many companies throughout the industry. Also for research organizations and for engineering service providers it's interesting to have such a backbone because it's easier to bring technologies into your system and it offers for engineering service providers for example the possibility to realize stuff as a service in throughout the industry paid by OpenADX. So if you're watching the scene there are quite a lot of collaborations close to often but also for example you perhaps you have heard of Baidu they announced Apollo framework for automated driving as a solution. Why do we think something new something special is needed and I think we have a unique selling point here. All the initiatives I know are pretty much ok we provide the solution something also we know of tool chains which are integrated but from a certain vendor and you can use it as is you can perhaps plug in some tools but it's not an open approach to integrate whatever is needed. If you look at Bosch when you have three departments developing similar systems for different customers you typically have three to five different tools used for similar tasks. So it's even within Bosch it's really an issue to integrate all the different tools used and have a tool chain that runs smoothly for the developers. And in collaboration with customers we often develop our systems in collaboration with a customer he brings it in his car it's even more difficult they have their tool chain we have our tool chain we have to bring those things together so that there is a smooth and efficient flow of data through the tool chain. So that's really what we target at so we don't say we will not point out which tools are used in this tool chain but our purpose is really to be integrative and allow all the tools used to be part of this tool chain to have a backbone that's running and you can plug in the tools you want so we will concentrate of course on those tools we use and our partners use but in general the idea is to have a really open framework that solves issues we currently have so it's not targeted at this is the final tool chain and we work towards this we target at providing a backbone solve issues we have in certain situations and grow this towards really something bigger a whole tool chain in the end not but never in a way that we say this is the way to go this tool cannot be integrated but always open for anyone to participate Okay so I want to look a bit into detail so what you see here and you saw it on one of the slides earlier is basically the tool categories used for current development of automated driving solutions not for the fully automated driving solutions but for example for a highway pilot or support in traffic jam stuff like that so what you see is I mean you need some development some structuring that's architecture definition we have a lot of data to handle we get a lot of sensor data that has to be stored and managed throughout the process of course especially in the perception area deep learning is the way to go machine learning we have to simulate, test the stuff integrate build it here often we have the situation that the system is not on one ECU as it is used or was in the past at Bosch so we had divisions and they provided one ECU with their stuff and there were some network interaction but here we now have a system that is really distributed over several ECUs which have to be synchronized and work together we have then of course the validation of the whole car so the simulation based validation is then targeted at the whole car also test driving also the connectivity has to be tested and the idea really here is to bring this together as a tool chain a reference architecture and allow tools to integrate so as I said it's not about replacing tools it's not about defining tools but we really want to make a map of what tools are available what tools are relevant for the current situation for the partners who work on this and then provide an integration for these tools so that people can use them easily in one picture we want to basically understand which tools are used and what are the tool chains and how is the data flowing through those tool chains we want to describe that define the interfaces data structures which flow through the tools that's often the point where people have issues or companies have issues because one tool delivers a data format that has to be converted to be used by the next tool so a lot of time is lost in those inter-tool communications so we want to standardize at least on data structures that flow through the tool chain and yeah so that's the big picture we want to start small we want to solve first problems and grow into open source projects and come up to this basically vision so that's really the way to go since many of those companies are not used to use open source or work with open source we want to experiment first and that's really the idea of creating testbeds in which we concentrate on one defined use case and create a demonstrator create something usable which can then already be used in a certain company for special problems out of those testbeds we expect that we will grow new open source projects which will solve the certain issue for long term so that this can really be used throughout the industry in a general way on the path we will see ok most of the time it will be the clue between the tools the data structures the communication infrastructure the middleware that allows to share the data but sometimes we also expect that we will have new solutions to be developed on this way if you look at how the industry today works it typically starts with standardization so big example the autosaw initiative where a lot of big companies came together make an even closed IP solution that is then used throughout the industry more or less successful because the nice thing with standards is it's written on paper and you can interpret it so there are a lot of issues with exchanging data and understanding the data in the same way as the previous tool does so for us standardization really comes at the end so if we have the open source projects running, if we have the solutions in place we intend by doing this to get a kind of industry standard which is used in multiple companies throughout the industry so it's really not the goal to say ok let's standardize first but let's code first ok so what we are currently doing I talked about the testbeds what turns out to be two major issues in the industry in automated driving is one of our first of all simulations so there are a lot of great simulation tools and simulation mechanisms a lot of technologies things that need to be simulated and we also talk about an enormous amount of simulation needed to validate the system so one metaphor I heard from colleagues is to validate an autonomous driving system it is expected to drive one time to the sun and back it's really an awful lot of validation needed and driving around needed and the problem also is when you do this in reality out on the streets it's pretty hard to really create all the situations you need to validate the system correctly so it's really simulation needed at that point to ensure that the system works as expected in all the situations that are to be expected in reality so that's why we concentrated on simulation as one of the first areas to experiment and the second topic we look at is really capturing and managing large amounts of data as you can expect typical sensors are video sensors, radar leader sensors especially the video data is an awful lot of amount of data and cars drive around on a daily basis gathering sensor data to be used throughout development so we talk about this year of petabytes of data that have to come in and have to be processed within the organization to develop the systems so this is also a topic where we identified where we need solutions now and where we want to work on so being one level more concrete so now comes the powerpoint architecture part of the talk I said we have basically two initiatives ongoing where we really produce results one is the Bosch connected world we have the hackathon there where we want to show yeah where we want to give a playground for experimenting with automated driving it's a bit complicated but that's the setup basically so we have there are three different kinds of hardware some small vehicles which you can really steer within the context of the hackathon two kinds of rover one coming from research project we are also involved in the abstract project which is about connectivity connected services in the car and we use a demonstrator there where you can onboard and off-board control the movement of the rover and also get data of the sensors while the rover is moving there's a second kind of rover which is already ROS-based so ROS basically is one of the technologies we identified is used in different companies perhaps even throughout the industry I would say so that's a second hardware device we have and we are cooperating with a cool thing from universities the formula student driverless they have a race series universities organized teams where people can run race cars construct race cars and run them in a competition already since last year driverless so it's autonomous and we will have a sensor set from those cars available which can be used to experiment in the area of perception of SLAM and stuff like that we also will have provide simulation means to also test then the outcomes of such developments in an offline fashion we also added from Microsoft and simulator ASIM where you can also simulate in a way that ASIM provides a wheel or close to reality camera view of a scenery and you can use also ASIM to simulate AD perception as stuff in such environment data analytics who will look into what's coming in and want to visualize things and finally we will also include Eclipse sumo for those who don't know this this is in traffic simulation so you can simulate the traffic within a city with sumo and you can combine this with the other simulations to work on this so you see a lot of simulation in there which is basically representing our testbed on simulation and you have this big data lake here this is our first work towards handling big amounts of data so we're not talking about pit we're talking about doing the two days of the hackathon it will still be gigabyte but nonetheless we planned to use the technologies which can then be scaled up and here we are cooperating with Microsoft to get at least some initial setup towards data management, data ingestion to have this hackathon it will be on 21st and 22nd of February it's a big event under the hood of the Bosch connected world the trade fair of Bosch in the IoT sector there will be 5 hack challenges this is one of them if you're interested you're invited just search for Bosch connected world and you get to the web page and you can register for the hackathon as I said there are 5 different ones so if you're interested have a look okay so one activity where we show what we plan with open 80x a second one is really to put up a demonstrator as I said and I said ROS is used with ROS 2 you get typically you have DDS data distribution service technology as the middleware to share the information between the participants it's a published subscribe framework standardized by OMG and used in quite some companies I know so we said okay let's base our demonstrator on DDS since this has relevance in the industry and there are of course open source implementations we can simply use in our setup what we plan to do is an emergency braking system so nothing fancy at least but a good way to show how our partners here and it's quite a number as you can see a lot of different technology providers also competing technology providers can show in this scenario how they can bring them in and how we can connect all those tools together to show one use case so we have of course things like sensor simulation so that we can run something without a car the sensor data processing the control algorithm we have some visualization and we have a lot of simulations of one of our first of all the vehicle so the behavior how the car behaves in a certain environment we have the environment simulation and the concrete scenario so all these are typically provided by different tools which are connected then with the DDS backbone and the idea really is to build up a demonstrator to use standardized formats we already have so for example we plan to use ROS image to get the data to into the system there is there was a research project which came up with a standardization of simulation interfaces in this context so define some interfaces which are open source you can download them simply from a github account so we will use also these they use a different mechanism for communication so we will do a mapping towards DDS in this case but that's basically it if we can use them so this will come up hopefully in the second quarter of this year so that we can show how we can really integrate such tools in such an environment yeah that's OpenATX so we really want to create an industry-wide initiative and I think we are on a good way with the first other tier one ZF on our interested parties list participating in all the stuff we are in talks with car manufacturers we will build this ecosystem up I'm pretty sure we will find others on this list soon the way we start the interaction is this testbed so as the demonstrator I just showed they come up with such solutions and yeah you will see something from us this year I'm pretty sure and as I said it's Open it's really an open initiative so if you are working for a company who is working in this area you are all invited to join to help us getting things done and also to present what we have in this environment so that people can see it it's also a good marketing instrument to show things in a real context in a relevant context yeah so website as I said on wiki eclipse org and there is a mailing list in which you can register if you are interested in getting some information of course you can always contact me yeah we have some others since I'm not really in the part of Bosch but we are supporting the autonomous driving division there so of course there are interested people who organize this from their side that's it thank you very much questions anyone somebody is pointing put your hand up please Pablo see one question concerning this openness and so open source I guess it wouldn't be able that when I had a car equipped with the stuff that I just download the source code and change some parameters and do my own stuff I guess it won't be possible so how is your standpoint here first of all we target now at the development environment there is for example Apollo who really also targets at this stuff and we think about some parts to be open source especially infrastructure operating system stuff there is no gain in doing this proprietary downloading and it's like this PC you buy this or you buy a smartphone and the first thing you read is if you open it you lose all the guarantees and to be honest I know how difficult it is to calibrate those systems and also some legal regulations on this so I would not expect this to happen soon let's say Bosch and Daimler announced that they are working on a big automated driving project and is your toolchain used there or is it planned to be used we are in talks of them so in intensive talks it seems that they are simply a bit too far ahead so the problem here is of course we have some lag time to get things onto the ground and they already doing this stuff so we will stay in touch if they can use things we provide they are open to that but as of now they need solutions which we are about to create and this is a timing issue thank you how do you plan to align your plans your very top down plans and approaches with the way the open source community which is usually more like bottom up approach to align so I mean you are talking about creating a foundation but I haven't been able to see a single line of code in a github repository that's about to come that's exactly the demonstrate that we are working on it's kind of bottom up in a way that the people we often talk in companies are not the lead management guys it's really interested parties for example I talked to a guy from a german car manufacturer and he simply heard of it and was interested because they have a component they think they can share it's now still very interesting industry based so it's we need some flash basically to be interesting also for other interested parties so hopefully this will come but we are at the beginning so anyone else don't be shy no thank you I guess thank you