From Eulerian to Lagrangian Sensor Systems
This is a modal window.
Das Video konnte nicht geladen werden, da entweder ein Server- oder Netzwerkfehler auftrat oder das Format nicht unterstützt wird.
Formale Metadaten
| Titel |
| |
| Serientitel | ||
| Anzahl der Teile | 17 | |
| Autor | ||
| Lizenz | CC-Namensnennung - keine kommerzielle Nutzung - Weitergabe unter gleichen Bedingungen 3.0 Unported: Sie dürfen das Werk bzw. den Inhalt zu jedem legalen und nicht-kommerziellen Zweck nutzen, verändern und in unveränderter oder veränderter Form vervielfältigen, verbreiten und öffentlich zugänglich machen, sofern Sie den Namen des Autors/Rechteinhabers in der von ihm festgelegten Weise nennen und das Werk bzw. diesen Inhalt auch in veränderter Form nur unter den Bedingungen dieser Lizenz weitergeben | |
| Identifikatoren | 10.5446/50512 (DOI) | |
| Herausgeber | ||
| Erscheinungsjahr | ||
| Sprache |
Inhaltliche Metadaten
| Fachgebiet | ||
| Genre | ||
| Abstract |
|
5
11
12
17
00:00
SoftwareentwicklerEreignishorizontInformationSpieltheorieFreewareReibungswärmeKomplex <Algebra>Hasard <Digitaltechnik>Objekt <Kategorie>KonditionszahlWarteschlangeDatenfeldPhysikalisches SystemTransportproblemInformationstechnikSchätzfunktionTreiber <Programm>QuarkconfinementInternet der DingeQuellcodeHasard <Digitaltechnik>ReibungswärmeStatistikMathematikFlächeninhaltProjektive EbeneWorkstation <Musikinstrument>MultiplikationsoperatorSoftwarewartungWinkelverteilungPunktKollaboration <Informatik>GrundraumKonditionszahlGruppenoperationSystemzusammenbruchQuick-SortGüte der AnpassungATMDiagrammGeradeStreaming <Kommunikationstechnik>Reelle ZahlInelastischer StoßComputerspielProzess <Informatik>MaßerweiterungPunktwolkeDifferenteAnalysisSchnittmengeBitSchwebungInformationDatenmissbrauchEreignishorizontMobiles InternetBenutzerschnittstellenverwaltungssystemAutorisierungSchlussregelDatenflussProgrammierumgebungEinfach zusammenhängender RaumMessage-PassingOrdnung <Mathematik>StrömungsmechanikFlächentheorieKartesische KoordinatenDomain <Netzwerk>TelekommunikationSymplektische GeometrieEinflussgrößePartikelsystemWhiteboardShape <Informatik>VektorpotenzialArithmetisches MittelGleitendes MittelAbgeschlossene MengeMechanismus-Design-TheorieEinfache GenauigkeitWarteschlangeDienst <Informatik>DatensatzAggregatzustandWeb SitePlastikkarteCoxeter-GruppeBeobachtungsstudieSichtenkonzeptInternetworkingEndliche ModelltheorieChatten <Kommunikation>Spannweite <Stochastik>WasserdampftafelSummierbarkeitSprachsynthese
Transkript: English(automatisch erzeugt)
00:08
Hi, my name is Alna Dalsnestoszate, I work as a senior engineer within the field of intelligent transportation systems at Norwegian Public Roads Administration.
00:23
I'd like to share some thoughts with you about how I think the Internet of Things will change transportation. This may or may not be obvious from the title of my speech, the Eulerian and Lagrangian points of view, they appear in many domains.
00:41
For me, my background is in mechanical engineering and IT, and for me it is a reference to fluid mechanics. In fluid mechanics, the Eulerian approach is to watch as a flow passes through a point in time. And the Lagrangian approach is to follow a particle as it passes in the flow.
01:02
And fluid mechanics lends itself quite readily to transportation, because we also deal with flows in transportation, traffic flows. But let me first give you one of many definitions of ITS. Intelligent transportation systems is defined by the US Department of Transportation as the
01:25
application of advanced information and communications technology to surface transportation in order to achieve enhanced safety and mobility, while reducing the environmental impact of transportation.
01:40
Needless to say, the idea that everything will be connected will change how we work with ITS. In a future society, every vehicle, every driver, and every passenger on any mode of transportation, and even the infrastructure, will be able to communicate with each other.
02:08
I made some diagrams today to try to explain how this affects my line of work, because traditionally this is how we view traffic data, it's the Eulerian approach.
02:21
And these red lines represent cross-sections where we monitor the road. And it's typically by either equipment hanging over the road, or equipment embedded in the asphalt. It could be an inductive loop counting the amount of traffic passing by, or a weather station saying something about the weather and the road conditions.
02:49
And these cross-sections are great. They provide great statistics on traffic data, it's very reliable and consistent, and it's very, very expensive. But that's not the only problem, because the other point is that they only provide
03:05
glimpses of what happens at exactly those points in the road, and we know nothing about the stretches in between. It's not true that we only have these static points. We also have a few vehicles that monitor road conditions.
03:22
We have five. We have five vehicles in the whole of Norway to monitor the road conditions. So we're trying to change this, and that's why I'm working on a project called the Road Status Information. It is a collaboration between road authorities in Norway and Sweden, Volvo Cars, and universities
03:46
in Norway and Sweden. And the idea behind Road Status Information that is at a normal car, based on its onboard sensors, can tell you whether or not the road is slippery.
04:00
And this works because the car can take the data from its onboard sensors and calculate an estimation of friction as it's driving and send this to a cloud-based back system.
04:24
The benefits from achieving this are, well, the first one is that you can warn the driver. And Volvo Cars aims to warn the driver before the driver is noticing that this is a very slippery spot, thus enabling him or her to drive more safely. But this can all be done within the confines of the car, and the real impact is when
04:43
we start to share this data. Because if one car detects a slippery road, he can be able to send this data to another car coming into this same slippery road area. And even better for us as a road authority is if we can provide this data to the winter maintenance crew so that they can come in and salt or do other things to the road
05:06
to improve friction, because we want them to be able to put in the right measures at the right place at the right time.
05:27
So what we're doing is that we're equipping a fleet of cars in both Oslo and Gothenburg area with a communications module connected to the OD port. And this enables the car to send data about the road conditions to our back system.
05:43
We aim to have about 500 vehicles equipped with this technology. But in the future, every car will have this technology already built in, and the newest cars already have. So we're trying to look into the possibilities that this gives, because connected cars fill
06:01
these gaps. They can cover extensive areas. The data is different. It's not as reliable, because cars are different, and the tires may be bad, and the drivers may be bad. There are a lot of other things that we need to think about. But the change from having a few static points to having a fleet of connected cars
06:22
is still quite revolutionary. It's a vast source of sensor-based data. I tried to really get my point across with this drawing, where I tried to compare the
06:40
old-fashioned way of looking at things through the Eulerian approach, like a fisherman standing on the shore. And you may see something in the water, and it's just like, all right, there's a fish. And in the same way, we can detect cars passing this inductive loop and say that, okay, there's a car passing. Or maybe even better, we can say that there's a set of wheels passing.
07:00
And maybe if we're lucky, we can kind of detect whether it's a car or a truck. But we can't say much more than that. But in the Lagrangian approach, the cars or fish, they can detect what's happening in the stream. They can detect things around them, like temperature and the speed of the stream and
07:20
other fish and even hazards. That's for one fish. And you can try to imagine, or a car, if we have a school of fish or a fleet of cars, how much data we will be able to collect. And if this drawing seems a bit half-hearted, it's only because nothing beats real life at trying to explain things.
07:40
And before I had time to finish this presentation, there was an event in real life that explained it a lot better than I ever could. Just outside Bergen, our second largest city, there was a huge collision. It was the biggest in my lifetime. 60 cars were involved, and luckily there were no fatalities.
08:06
And this is where the accident happened. And that's where we have these two static points saying something about the road conditions. And these points, along with the weather forecast and the maintenance crew's own experience,
08:24
is what they have to base their job on. And what happened was that at 3 a.m., the maintenance group passed this accident area, and they salted the road. They said, okay, it's slippery, we'll put on salt. And at 6 a.m., they passed another time.
08:40
And this time, they gathered in their subjective experience that nothing needed to be done. And at 6.31, the accident was a fact. And the accident was, of course, reported through traditional means, typically by phone. And even when we knew about the accident, we could close off the road leading into the accident,
09:01
but cars continued to crash into the rear of this crash site, amounting up to 60 cars. And even the conditions were so bad that even the emergency vehicles couldn't avoid hitting each other. So we have to think about how we can improve the data that we give to our maintenance crew,
09:22
so they can perform a better job, so we avoid these sort of accidents. And I told you we had five trucks. I'll just try to get across how likely it would have been for one of these five trucks to be in the area to provide more data. We have about 104,000 kilometers of public roads.
09:41
We have five cars. And they have to drive quite slowly to be able to measure the conditions. So it would have taken them about 322 hours to cover all this road. So we can't use them for anything even near live data. They're good for statistics for providing us information about how the road changes over time,
10:03
but it's useless for anything else. And the other thing is that they don't really broadcast this data back to our systems directly. They have to be often uploaded manually later on, so it's really, really tough.
10:21
So the question is, of course, how could the technology that we're using in the road status information project have avoided this collision? Now the first one was what I told you to warn the driver. If the driver was given an advanced warning, maybe the first car wouldn't have crashed at all. But even if the first car had crashed, it could still have warned the other cars coming into the area that you have to watch out,
10:46
because this is really, really slippery. But as a road authority, we want to avoid these conditions at all. We don't want conditions to be as bad as this. So the best thing would be if we could give this information to the maintenance crew in time for them to make the road safe for everyone.
11:07
So I think the real benefit from having everything connected is when we can harvest the data and try to analyze them in a manner so that we can make forecasts. So we can say that, okay, if the temperature is like this and the precipitation is like this, we have to go and solve it in advance.
11:24
Then maybe we can avoid these sorts of situations. Now we may have started at the wrong end, because friction is one of the hardest things we could have done based on sensor data from the car. They produce a lot of data, and some come from the car itself, like if you press the hazard warning,
11:45
or if you brake really hard, those can be converted into signals that we can use. And you have data based on how the car perceives the world around it, like road signage, road markings, wear and tear, conditions, temperature, or even cues.
12:03
You can say something about the speed that the cars are traveling in. And we see this as a huge potential for us. We see this as the future of how we have to work with traffic data, and we realize that we really need to change. We need to stop being the fishermen on shore, and we need to get in and try to see how the fish are working.
12:22
And I wanted to say that as a world authority, our policy is that we try to provide all the data we have free of charge. So I hope that by releasing our data, making good APIs, and harvesting this new source of sensory information,
12:40
we can make real changes to how we view traffic, and to really enhance safety and mobility on our roads. That's it. Thank you.
13:25
Privacy has been really important for us. My project was initially shut down because our boss didn't think it was following the privacy rules to a high enough extent. Usually the car manufacturer and the car owner enters an agreement where the car manufacturer can log the data from the owner.
13:47
That's how every car manufacturer does it. So we have an agreement with Volvo that we only get anonymized data. So we should never have data about a single person or a single vehicle, ever. So that's our policy, and you're quite right. There's a very big difference between how we work and how Google and other people can work.
14:03
And we also have written consent from all people participating in the project in the first place, just to be super safe. But we want to try to not have to have the personal data at all. We want to just see more clouds of data and try to make good analysis from having data that can't be linked to a person or a vehicle.
14:27
That answers your question. Other questions? Thank you.