EIT-ICT Labs - 2012-04-23
This meeting was arranged by Christian Müller of EIT ICT Labs, together with DFKI and W3C, and took place at the EIT ICT Labs offices in Paris.
- Round table introductions
- Overview presentation by Dave
- Presentation on related research projects by Christoph
- Round table issues/topics for discussion
- Summing up by Christian
- Dave Raggett, W3C
- Stylianos Geisdakis, KTH
- Panos Papadimitratos, KTH
- Justin (Jongseon) Park, LG Electronics
- Hedimann Dominiteus, DFKI,
- Chathuranga Weeraddana, KTH,
- Christian Müller, DFKI,
- Christoph Endres, DFKI,
- Iisakki Kosonen, Aalto University, Helsinki
- Introduction by Christian Müller
- General overview of Web & automotive by Dave Raggett
- Standardized Presentation Languages and Situation-Aware Multimodal Automotive Presentation Planning by Christoph Endres
- PRESERVE project on V2X security by Panos Papadimitratos
These rough notes were taken by Dave.
- eye gaze tracking + voice interaction
- car to car, and car to road signs etc.
- DFKI working with Nuance on extending Apple's Siri service to support gaze tracking and local speech handling etc.
- working on traffic modelling and driver distraction
- in car camera as means to measure driver attention
- experiments with pedestrians and cars pulling out in front whilst the drive is also performing some other task (driver psychology).
- Planning a new project
- Concern about driver's ability to handle exceptional situations as a result of reliance on automotive support systems
- Automakers opening up in vehicle interfaces (IVI) through open standards
- IP based in-car data bus is coming on a 5-10 year plus time frame, and will replace today's CANbus
Christoph: my work on context aware situational planning: tree search (ahead and ad hoc planning) for presentation planning based upon cognitive load and situational context. Multiple modalities, multi-device, alternative presentations. Terminology: notification management as essentially same thing as presentation planning.
Dave: asks about implications when running untrusted apps, i.e. those that haven't been designed and vetted for use in cars.
Christian: there can also be competition between willing partners (i.e. car friendly apps).
Dave: how to allow apps designed for smart phones or tablets to work in cars when there aren't the corresponding apps designed for the car? This is a similar situation to tablets where you often have to run apps designed just for smart phones.
Hedimann: execution within a sandbox ...
Iisakki: driver experience, we can learn from aviation and imminent collision warnings, where the tone and volume are altered in progressive warnings to gain the pilot's attention forcefully.
Christian: driver attention can be inferred by looking at the patterns of eye movements, and whether the steering wheel is being moved smoothly or jerkily.
What feature sets can we used for detecting driver health? Opportunities for embedding machine learning in the car.
Christoph: my work is concerned with presentation planning not the applications themselves.
How critical is dependable network connectivity? We all agree that network connectivity is often poor depending on the local geography, tall buildings and narrow streets etc.
Christian: current thinking is that there should be core services that continue to work when the network becomes unavailable. There can be hybrid services that offer richer capabilities when connected, and fall back to simpler capabilities when disconnected.
Chathuranga: will car makers be able to do this by themselves?
Christian: no, they will work with key suppliers and the general trend is to greater openness.
Christoph's work focuses on what contextual cues are most important, and how to make use of them effectively. Applications shouldn't be able to game the notification system!
Hedimann: (presentation) I am leading a project on social mobility with a focus on people. This is like a social semantic web of things. UserML as a markup language for exchanging user models for personalization, based upon RDF/RDFS with metadata such as evidence and confidence in relations, ownership and privacy.
- GUMO - general user modelling ontology
- CODA - collaboration ontology for development of applications
For more details, see ubisworld.org.
Very rich user modelling approach. We started with XML but there are too many dimensions and hence we switched to using ontologies. However, there is a lack of agreement on specific ontologies.
What can we achieve via standards work? An automotive core vocabulary could have a huge impact!
I can share some thoughts. For 5 or 6 years automakers have started to respond to the rapid pace of change and pressure from companies like Apple and Google etc.
Adding new functionality is hard with monolithic systems. This is motivating a switch to more modular architectures. Web technologies are seen as the best candidate. Automotive has many operating systems, so there is a need for an OS independent solution.
For connectivity, e.g. with the user's personal devices, exploiting web technologies is a big help. There is a need for standard in vehicle interfaces (as web APIs) for the car's subsystems, including power status (car key crank position). This is needed for power management of apps when the car engine isn't running.
Christian: we should also think about electric cars and a standby mode for apps running in them.
Justin: power management will be more important in electric cars.
Panos: (presentation) I would like to describe the FP7 PRESERVE project which is focusing on V2X security and privacy. Talks through connection to Dave's opening slides.
Services based upon remote access to car's diagnostics and logs - stress on access control.
- ETSI C2C-CC IEEE 1609.2 WG
Challenge in providing transparency to users regarding their choices. The World Wide Web is today rather one sided with excess tracking of users by companies because they can! Defending against risks of theft of personal data from your car.
Christian: the FP7 Get Home Safe project treats privacy as a matter for negotiation between users and app providers. Potential for collaboration across the two projects.
Stylianos: we're working on privacy for mobile vehicular communications. Big user data footprint which provides a major challenge. What are the least privileges an app needs? What additional security mechanisms are needed for web apps in the car?
Context awareness is a huge area of work. Security and crypto - implications for power management.
- Christian would like to use the W3C Automotive Business Group to facilitate on going dialog between organizations work on on web and automotive.
- We should create a shared position paper to convince other parties to join , as there is a need for a convincing case to justify funding for participation in standards work. Likewise to attract automakers and key OEMs
Dave briefly reviews the distinctions between using a public W3C list, a W3C Community Group and a W3C Business Group. An archived mailing list and wiki offers the lowest barrier to entry with no need to talk to corporate lawyers about charters and patent policies. Community Groups and Business Groups involve the need to define a charter and IRP policy. Business Groups get more support from the W3C Team at the cost of participation fees (free for existing W3C Members).
Whilst we are at an early stage of planning for a workshop, it would make sense to have a public mailing list and a wiki that anyone who has applied for a W3C Account could update. It will be much easier to set up a Community Group, Business Group or REC track Working Group after the workshop, which is now planned for September 2012.
Dave now has a lot more contacts following the Mobile World Congress, and intends to see if he can convince them to help with the workshop planning (see overview slides).
Dave to publish the minutes on the wiki, and send a pointer to Christian to circulate amongst meeting participants. Invitation to join the email@example.com mailing list, and to help with the wiki pages. See link on public-automotive archive for a painless way to subscribe (and unsubscribe). See also: