Experiences with a vehicular cloud computing platform Jason

  • Slides: 13
Download presentation
Experiences with a vehicular cloud computing platform Jason Flinn, T. J. Giuli, and Brian

Experiences with a vehicular cloud computing platform Jason Flinn, T. J. Giuli, and Brian Noble University of Michigan and Ford Motor Co.

The Goal A senior-level course around emerging technologies: • Vehicular sensors and interfaces •

The Goal A senior-level course around emerging technologies: • Vehicular sensors and interfaces • Cloud storage and computation • Widespread wireless connectivity • Social networking platforms • Partnership between Ford and Michigan • Class called “Cloud Computing in the Commute” Jason Flinn

The Hardware Platform Ford • • research hacked up a Fiesta to include: 10.

The Hardware Platform Ford • • research hacked up a Fiesta to include: 10. 4” LCD color touch-screen x 86 PC running Windows 7 Wireless connectivity via 3 G mobile broadband Access via CAN bus to telematics, GPS, infotainment Jason Flinn

The software platform Integrated sensors via Microsoft Robotics Studio Also provided: • Network access

The software platform Integrated sensors via Microsoft Robotics Studio Also provided: • Network access services • Services for interfacing with social-networking sites • Voice recognition, text-to-speech Deployed software via a virtual machine • Allowed development without access to vehicle • Included APIs for testing with location and sensor traces • Minimized student time spent with configuration • Porting to vehicle environment was surprisingly easy Jason Flinn

Green Ride Challenge Promotes ride sharing to save fuel, costs • Structured as a

Green Ride Challenge Promotes ride sharing to save fuel, costs • Structured as a game - earn points for transporting friends • Real-time updates of who needs rides where • Agreement protocol for arranging ride-share • UI runs in browser • Connectivity issues • Back end is cloud DB Jason Flinn

Fuel. Tracker Measure instantaneous fuel economy: • Compare to previous economy at same location

Fuel. Tracker Measure instantaneous fuel economy: • Compare to previous economy at same location • Compare to economy of others driving same route • Leverage social aspects to share green-driving tips • Paul Coldren, Clayton Willey, Petch Wannisorn, Amy Kuo Substantial back-end computation Cloud DB Web interface for complex stats Jason Flinn

Caravan. Track Helps co-ordinate a multi-vehicle road trip • For each vehicle, shows location,

Caravan. Track Helps co-ordinate a multi-vehicle road trip • For each vehicle, shows location, fuel, status, etc. • Later modified to use voice, steering wheel controls • Collin Hockey, Sangmi Park, Joe Phillips, John Ciccone Simple structure • DB backend Jason Flinn

Challenge: App participation required Caravan. Track aggregates sensing data from all vehicles • Speed

Challenge: App participation required Caravan. Track aggregates sensing data from all vehicles • Speed • Fuel • Stop preferences Jason Flinn

The Road Trip Caravan. Track team tested their application on a road trip •

The Road Trip Caravan. Track team tested their application on a road trip • Journey from MI to Maker Faire in CA. • Four Fiestas – two with integrated platforms • Sponsored by Ford “American Journey 2. 0” • All ran Caravan. Track • One vehicle (“AJ the Fiesta”) also ran social apps • Connectivity was generally quite good (except in Nevada) • Some unexpected benefits of Caravan. Track Jason Flinn

Lessons Learned: Network Connectivity Hosting interactions among vehicles in the cloud works well •

Lessons Learned: Network Connectivity Hosting interactions among vehicles in the cloud works well • Can easily handle human speeds for social apps • Connectivity usually sufficient on major roads • However, bandwidth can be a critical issue • Good performance needed for short, interactive msgs. For next time: • Would like to bring Intentional Network to platform • Allow background transfers w/o affecting interactive apps • Utilize networks other than cellular when available Jason Flinn

Lessons Learned: Sensing Many interesting apps used vehicle as sensor platform • Mostly limited

Lessons Learned: Sensing Many interesting apps used vehicle as sensor platform • Mostly limited to basic sensing: fuel, location, speed, etc. • Sensor data often shared among vehicles • Driver is also an excellent sensor However, substantially more sensors available: • Steering, breaks, temperature, seat position, cell phone presence, engine performance data, cameras (internal and external), … Vehicle is a great platform for participatory sensing: • Next time: add and enable more sensors Jason Flinn

Lessons Learned: UI still a Problem Designing for vehicular UI very challenging: • Must

Lessons Learned: UI still a Problem Designing for vehicular UI very challenging: • Must minimize distraction by keeping driver’s eyes on road • No dialogs with timeouts • All actions restartable • Only one button per screen region (four corners) • Use alternative modalities (voice, steering wheel controls) • Helped a lot to have design student(s) on each team Many successful apps included a Web component: • Complicated set-up done at a computer • Simpler interface used in the vehicle • More detailed information available later from PC. Jason Flinn

Thanks to… Timur Alperovich T. J. Giuli Brett Higgins Brian Noble Venkatesh Prasad Azarias

Thanks to… Timur Alperovich T. J. Giuli Brett Higgins Brian Noble Venkatesh Prasad Azarias Reda David Watson Jason Flinn