Summary of the Web and Automotive Workshop
On the 14 and 15 November 2012, W3C (the World Wide Web Consortium) held a workshop on Web and Automotive (Shift into High Gear on the Web). The discussions focused on opportunities and challenges for exploiting Web technologies within the car, and what kinds of standards work may be needed to realise the potential.
We live in a connected world, and people will increasingly expect to have access to applications and services whilst on the road. Safety is critical to realizing the potential, and there is an opportunity to enable people to be better and safer drivers through increased situational awareness. Web technologies are strategically important, and W3C is seeking to launch standards work to support the adoption of Web technologies in automotive contexts.
The Web and Automotive Workshop was a first step, bringing together a wide range of stakeholders to share ideas and discuss the potential. The next steps are likely to include further outreach with the aim of launching a standards activity. The Workshop prioritized work on use cases and requirements, security/policy mechanisms, a user centric vehicle API and a reference model. The W3C staff will work with stakeholders to identify opportunities for launching work in support of standards for Web and Automotive.
The World Wide Web originated in a proposal at CERN for a lightweight solution for information sharing across heterogenous systems coupled by the Internet. The early Web was dominated by the PC, but the focus has now moved to mobile devices along with connected TVs. More recently still, we are starting to see the emergence of Web technologies in the connected car. The Web and Automotive Workshop was organized to bring together a wide range of stakeholders to assess whether it is timely to address standardization as a means to avoid fragmentation and encourage strong growth of automotive applications and services.
The top 5 topics in the call for papers were:
- Putting safety first
- Relationship between the car and the mobile phone
- Creating markets for automotive applications
- Improving quality and reducing costs through Web technologies
- Unique opportunities for the automotive user experience
The Workshop participants reached a broad consensus that HTML5 is a compelling basis for automotive, but further work is needed, and that it is now timely to launch standardization of user centric vehicle APIs in order to avoid the risk of fragmentation from competing approaches, and the consequent slow growth in the provision of applications and services. There was strong support for work on use cases and requirements, security/policy mechanisms, a reference model and other topics. The full list is given in the summary of the wrap up session.
Main workshop discussions
Following an introduction by the Workshop Co-Chairs, Nils Opperman (Audi) presented an OEM perspective. This was followed by presentations from QNX and Harman and then after coffee, ACCESS, Ericsson and Gemalto.
The Audi presentation noted that standards were key to reducing application development costs and contrasted the number of new car sales versus the number of daily activations of Android smart phones. Of interest would be a standard API for access to the infotainment system. Some challenges include authenticating users when several people share a car. This is important for connecting people to their personal services (e.g. Facebook and Twitter). Audi have also explored the use of W3C's SCXML and Multimodal Interaction (MMI) Architecture.
Developers need a way to customize applications for a particular brand of car. One partial solution is a means to exploit native user interface controls. HTML templates provide another solution with a combination of markup, scripting and style sheets. Developers would like to leverage their effort across many brands of cars. This would greatly expand the number of applications available by reducing the development cost per brand of car.
Andy Gryc (QNX) identified several areas where W3C could help: application packaging, native access APIs, and integration of the car with mobile devices. One possibility is having a server on the phone that the car can access, but the other way around is also possible. Avoiding driver distraction and OEM skinning were called out along with application development guidelines. Locally installed HTML applications are appealing to some vendors, whilst others prefer server hosted applications.
Roman Steiger (Harman) indicated that they were interested in contributing to work on standardized interfaces to car subsystems, based upon their experience in GENIVI. Examples include control over the radio and heating systems. An open question is where the applications are executed: in the car's head-unit, a connected smart phone/tablet, or in the cloud. The look and feel may differ, but the control logic would remain the same.
Marcin Hanclik (ACCESS) described the role of a broker ("nBox") as a means for the web run-time to access the car's subsystems and for remote services hosted in the cloud. This provides extensibility without the need to modify the web run-time itself, furthermore services can be transparently provided by the car or by the cloud. This was described as HTML5 as a Platform (HaaP). The similarity to Web Intents was noted.
Magnus Olsson (Ericsson) described requirements for leveraging the open web platform: adminstrative features (configurability, discoverability, capability, and upgradeability), connectivity "drivers" (protocols such as HTTP and Web Sockets, and communications technologies such as WiFi, 3G/4G, and Bluetooth), responsive user interfaces (multi-screen, abstract UI events, intuitive security and non-intrusive safety measures), and finally the need to be future proof with a level playing field that enables plug-able business models. He mentioned connected ecosystems (walled gardens, app stores, bundled service, cloud services and business to business).
Automotive is an untapped source of interesting sensors and data. There needs to be agreements on how this data can be shared and consumed, along with security and privacy concerns. Lots of opportunities for developers, e.g. “crowd sourced” automotive speed sensors, and temperature readings, etc. Avoiding driver distraction means avoiding intrusive application notifications, and this can be handled according to the context and user preferences. Magnus tantalised us with his vision of a Web based head up display that combines sensor data with maps and social apps.
Pierre Girard (Gemalto) talked about the car as a programming platform and the challenge of avoiding the native application fragmentation problem. How to protect safety, privacy and security. What are the threat models given the greater complexity of the car use case and lifecycle compared to electronic appliances? He advocated work on designing a safety/security/privacy model that is permission and role based, along with a flexibile authentication framework. This should couple to existing work at W3C and elsewhere.
After lunch we focused on the role of vehicle APIs with presentations from LGE, BMW, Intel and AKQA. This seems like a very promising area for standardization.
Justin (JongSeon) Park (LG Electronics) talked about the need for standardized vehicle APIs and work being done at GENIVI. There is a limited window of opportunity to develop standards as many OEMs are now working on APIs in their own way. This makes it time critical to launch new standards work for vehicle APIs.
Simon Isenberg (BMW) described work on enabling rich web applications for in vehicle infortainment. BMW has developed and implemented vehicle APIs as part of the European webinos project. He outlined a vehicle API covering such things as brand, model, year, fuel, hybrid, steering and transmission. The navigation API provides a simple means for applications to interact with the car's navigation software, including the means to query for nearby points of interest. The webinos demonstrator is based upon a Pandaboard and the BMW iDrive controller, together with a screen and keyboard. The vehicle data is currently read-only, and he noted that security and safety have to be addressed properly along with user interface constraints.
Simon added that the webinos API was a starting point for the work in GENIVI. Webinos supports the notion of personal zones as a means for users to manage their devices and services. This is based upon local servers that act as proxies for the zone in conjuction with scripts injected into the web run-time.
Mikko Ylinen (Intel Open Source Technology Center) presented ideas for a lightweight vehicle API and its integration with the Tizen OS. A demo system was available for workshop participants to explore. The approach is based upon proxying events via Web Sockets (similar to webinos) and copes with a high throughput at low CPU loads.
Alex Ajao and Paul Slattery (AKQA) described AKQA's vision for the connected car. Your personal profile is downloaded from the cloud based upon your unique contact-less car key. AKQA is looking for a standardized API for accessing car subsystems via the CAN bus. It is important to come at the right level of abstraction for the API. They look forward to working with W3C on this.
The second day started with an invited presentaton by Roger Lanctot (Strategy Analytics) on automotive market trends. This was followed by presentations from Vodafone, KDDI, Visteon, Renault, KIT, DFKI, Intel, Igalia and CSC.
Roger described the car as a "persistent inquiry" with constant contextual awareness of the vehicle, driver and external conditions. Icons as used on the desktop, smart phones and tablets aren't appropriate for automotive user interfaces. What's the alternative? Proximity sensors and gestures are difficult to get right. Roger sees natural language as key to connected services. He asked how will connectivity be monetized and suggested that cost savings through collection of data is likely to be the most important factor.
There are lots of opportunities for improving on the current experience of using smart phones in cars. Deciding whether to answer an incoming call or to look at an incoming message, difficulties in interacting with phone-based navigation software, and awkward to control streaming media applications. The ultimate objective — millions of dollars in cost savings, e.g. from avoiding recalls, plus superior customer/vehicle relationship management.
Diana Cheng (Vodafone) started by observing that the automotive technology ecosystem is fragmented. The Web is rapidly evolving with new web platforms pushing the boundaries, e.g. Firefox OS, Chrome OS and Tizen. The gap with native apps is closing. Diana then took us on a review of the Web platform and what is still needed, for instance, a consensus on offline storage for large amounts of binary data, further work on Web Notifications, better implementation support for Web Sockets and Server Sent events, automotic update of web widgets, integration with local and remote speech processing, and advanced touch interactions.
Shigeyuki Sakazawa (KDDI corporation) described KKDI's vision of a seamless personalized services that are available anytime, anywhere, via any device. People want to focus on the service and not on the device, thus there should be seamless handovers as users move from home to the car, to the cafe, the supermarket and back again. This implies smart hand overs across networking technologies and smart prefetching and offline/online transitions. KDDI propose a hybrid of smart phone and car head-unit, with the smart phone as a web server, and apps executing in web run-time on the head-unit. They are interested in W3C work on service discovery and application authentication.
Jean-Marc Temmos (Visteon) noted that Visteon has a great deal of experience with creating human machine interfaces with a wide variety of technologies. Key considerations include customization, multi-modal interaction, context awareness, and an open and dynamic architecture. For personalization, one approach is to hold user settings and preferences on the smart phone. Seamless synchronization is needed between the car and other devices (mobile, cloud). The potential for machine to machine communication and for swarm intelligence. Jean-Marc described the E-Bee concept car featuring the UI.Cockpit framework. This centralizes the application logic and distributes the user interface across various HTML5 client displays via the model view controller pattern and Web Sockets.
Ludovic Alidra (Renault) started by observing that car displays are special due to the many purposes they can be put to. On safety, Ludovic noted that apps should avoid distracting drivers, or requiring driver action when driving. Apps should be validated before deployment on the app store. He has hopes for the future of more dynamic run-time enforcement of design rules. He listed challenges for functional API definitions including the need to be user, trip and car centric. What is needed to monetize in-vehicle services, and collected vehicle data?
This was followed by two presentations of research work. Sebastian Speiser (KIT) talked about speech based driver assistants embodying integration of multiple data sources on the Web, logical inference and statistical learning. Michael Feld (DFKI) talked about approaches for assessing and adapting cognitive load, along with ideas for improving situational awareness.
After lunch we had a presentation by Tsuguo Nobe (Intel) on the challenges posed by the differences in product life cycles for cars and electronic appliances. This can be addressed by the means to provide updates throughout the car's life cycle, and the role of cloud based resources for handling content and services. He believes that applications and services execution will migrate from the phone to the car's head unit, driven by evolution of big data and cloud based services. HTML5 will be a “glue” which connects cars with Cloud-based services via Smartphones, and will pave the way for new service & business development.
Alejandro Piñeiro Iglesias (Igalia) talked about the role of accessibility based solutions to improve safety whilst driving. The user interface on smart phones is distracting due to the need to concentrate on the display whilst using touch gestures. This is bad for safety! Speech based interfaces can reduce the level of driver distraction, and this can be enabled via accessibility APIs that allow scripts to interact with applications programmatically on behalf of the end user.
Chris Delaney (CSC) talked about changing business models for automotive, for instance, the role of in vehicle apps for improved customer relationship management with new opportunities for car manufacturers to streamline the sales process and engage with customers for the long term. He illustrated this in terms of an online approach to selecting and personalizing a new car, and sharing the information with friends.
This session was chaired by Philipp Hoschka [slides]. He noted that Web technology is strategic for automotive and people will increasingly expect to be online and connected whilst on the road. What is needed to build a developer ecosystem, to remove potential roadblocks and achieve critical mass?
He then described the motivation for W3C and the need to ensure that all stakeholders are represented to ensure success. Speed depends on the participants. You start with presentations and analyses of current approaches, drafting of specifications and responding to feedback based upon prototype implementations. This process iterates until mature specifications emerge as the basis for standards.
A brainstorming sesssion was conducted to collect and prioritize potential work items for consideration as part of a standardization activity:
Ideas and counts in descending order
- Use cases and requirements (19)
- Security/policy mechanisms (17)
- User centric vehicle API (12)
- Includes entertainment API
- Reference Model (10)
- Gap analysis (7)
- Context Management (6)
- Device centric API, e.g. phone (6)
- Define ecosystem (6)
- Best Practices (5)
- HMI integration (4)
- Native UI, customization with templates/CSS
- High level APIs(4)
- Update/sync of apps and data (1)
- Low level APIs (1)
- Tooling, e.g. for developers (0)
How to get started? Philipp outlined the role of W3C Working Groups and Business Groups. A Working Group develops standards, interacts with other W3C Groups, liaises with external groups, and is W3C Member only, but often works in public. A W3C Business Group doesn't work on standards, but rather explores requirements and use cases that feed into standardization work. Non W3C Members can join for a small fee. Existing W3C Business Groups have been formed for Web Signage, Oil+Gas and Broadcasting. A Business Group could operate in parallel with a Working Group. The work could be kick started by member submissions.
The Workshop concluded with thanks to Intel for hosting, our sponsors QNX and webinos, the programme committee members and the administrative staff for logistical arrangements.