Internet of Cars

The HERE Automotive Cloud can improve driver safety. Find out how.

Transcript

1. Using the Cloud to Make Driving Safer Richard Cross ACS DevOps 2016.05.10 HERE Automotive Cloud Services
2. 2 • What is HERE, Who am I, Why am I at Geekfest? • The Future of Driving • Sensors • Digital Transportation Infrastructure – Ingesting Data – Mapping Events – Alerting • Safety at Scale – The DevOps Approach • Working in the Cloud – Challenges – Benefits Overview
3. 3 • Founded in 1985 – Navigation Technologies Corporation (NAVTEQ) – Bought by Nokia in 2008 • Purchased by The Consortium in 2015: – Audi (subsidiary of Volkswagen) – BMW – Daimler (Mercedes) • HERE makes digital maps – For Tizen, Bing, Windows Phone, Kindle (Consumer) – Garmin, Facebook (Enterprise) – Sell 4 out of every 5 car onboard navigation systems What is HERE & Why Cars?
4. 4 • My story – PhD in Physics from UW-Madison - Discovered interest in enterprise computing – MBA University of Chicago - Studied Commercializing Innovation – 14 years enterprise computing at Bank of America - Stumbled into DevOps - Automation, testing, continuous delivery • Past 2 years at HERE – In Automotive Cloud Services – Run a DevOps team – Goal: - Create an infrastructure that can provide services for 10 Million cars by 2020 Richard.Cross@HERE.com
5. 5 • We want you to use our maps – HERE Maps are embedded in many applications – Real time layers such as traffic • Free as in Libre – If you use our maps, you own the data – We don’t mine your data stream • Our owners want us to be an open platform for automotive applications – An OS for cars? We want to be part of that We have APIs and SDKs for Location Services!
6. Self-Driving Cars? The HERE Vision for Transportation Making driving smarter, safer and more enjoyable Maps & Automotive Services Fresh, dynamic, global availability, high performance Intelligent Car Computational, automated predictive, real-time Smart Guidance Contextual, personalized intelligent, cross platform Digital Transportation Infrastructure Connecting vehicles and travelers to smart infrastructure Location Cloud © 2016 HERE
7. 7 • 10 Million self-driving cars on the road by 2020 – OK, fully self-driving cars ~2019* – 90 million connected cars with internet connectivity on the road now Self-Driving Cars: The Future is (almost) HERE *Assumes regulatory issues are solved
8. 8 • Improved Productivity • More Efficient • Much Safer! Why Buy a Self-Driving Car?
9. 9 • We are nervous about giving up the wheel • We don’t entirely trust computers • Or car makers Safer, Really?
10. 10 • Even a well programmed car will have accidents • Must avoid the blue screen of death • Infosecurity must be solid Press Paranoia
11. 11 • It’s easy to give into our fears • Regulations are a big factor in when these cars will be allowed on the road Will We Let Our Fears Stand in the Way of Progress?
12. 12 • 1.25 Million driving fatalities in 2013 • Americans spent 6.9 Billion hours stuck in traffic in 2014 There Are Problems with How We Drive Now: http://www.who.int/gho/road_safety/mortality/traffic_deaths_number/en/
13. 13 • Pros: –People feel more in control –The fun of driving • Cons: –People have a false sense of security –More expensive Industry Debate: Should Self-Driving Cars Have Steering Wheels? The Ultimate Driving Machine
14. 14 • No Auto Deaths by 2020! –How? Volvo Believes They Can Deliver Perfect Safety http://money.cnn.com/2016/01/20/luxury/volvo-no-death-crash-cars-2020/
15. 15 • Faster Computers • Security • Better Sensors • Better Maps • Cloud Services HERE has a Vision to Improve Safety
16. 16 • Passenger selects destination • Several sensors gather data on local objects • Car CPU processes images • GPS used to find car’s global position • Car CPU evaluates local and global data to pick best route Sensors and Computers Make Self-Driving Cars Work
17. 17 • Today Backup Cameras • Tomorrow 360° Coverage Better Sensors: Cameras
18. 18 • Better Image Recognition Better Sensors: Image Recognition
19. 19 • Lidar uses emitted light to measure distance • Lidar good at identifying lanes • Expensive • Short Range –Problems with Fog, Rain and Snow Better Sensors: Lidar
20. 20 • GPS gives uses satellites to find your location • Inaccurate indoors and near buildings • Cheap, global coverage Better Sensors: GPS
21. 21 • HERE HD maps are built by sending out mapping cars that use lidar and high precision GPS Create Base HERE HD Map with Lidar and GPS 1 LIDAR capture from True vehicles for 3-D rendering 10 cm accuracy for attributes including lane width and height of objects
22. 22 • Driver gives car destination • Car downloads most recent live maps –Car uses GPS and sensor data to find precise location and lane by matching to HD Map Localization Layer –Cloud provides suggested route with projected traffic and real time events HD Maps for Routing
23. 23 • We have published a Sensor Data Ingestion Interface (SDII) specification: – Vehicles driving on the road are equipped with a magnitude of sensors. These Sensor Data may be transferred over any kind of technology from the vehicle to an Analytic Processing Backend. Between individual vehicles and the Analytic Processing Backend, an OEM-, or System Vendor-Backend may be located as a proxy. The Sensor Data Interface Specification defines the content of Sensor Data Messages and their encoding format as they are submitted to the Analytic Processing Backend. However, the specification may be used between other components as well. – Licensed under Creative Commons 4.0 Put Real Time Events from Connected Cars on the HD Map https://www.iex.nl/Forum/Upload/2015/8777932.pdf
24. 24 Sensor Ingestion Interface Specification
25. 25 • We are creating a communication channel between transportation authorities and drivers via intelligent connectivity over existing wireless networks • Vehicles tell the cloud where they are (with high precision) – GPS – Sensors & HD Maps • A Digital Transportation Infrastructure sends relevant events to a car (Weather, Road Conditions, Construction, Obstructions, Moose, etc…) • Using the existing wireless network can reduce the cost by 95% relative to building an entirely new wireless network HERE Vision for Digital Transportation Infrastructure (DTI)
26. 26 • How do we get the data to the cloud? HERE is network agnostic • There are also proposals for car-to-car communications (DSRC/WAVE) • And infrastructure to car communications (say car to traffic light) • There are a lot of Internet of Things proposals for transportation improvements that require building out a new network • We are piloting our next generation system using 4G in Finland today Industry Debate: Use Cell Networks or Dedicated Network
27. We Have a Live Pilot on Standard 4G Road and Network layer Cloud layer Outbound message Inbound message Traffic Management Center / other 3rd party clouds (optional) HERE Location Cloud Data Processing and Analytics Sensor Data Ingestion Intelligent Data Store Finnish pilot sends alerts about weather conditions and large animals
28. Finnish authorities broadcast targeted alerts such as inclement weather Users can self-report events such as poor visibility The standard is forward compatible with ingesting car sensor data Finnish Pilot Uses Mobile Phones, Cloud and Traffic Management Center Implements Decentralized Environmental Notification Message (DENM) standard which is written to support infrastructure to car and car to infrastructure alerting
29. Goal of Fell-Lapland pilot is to reduce reindeer accidents by 50% by 2020 App will send reindeer proximity alerts Pilot set to launch in August Future Pilot Focused on Reindeer Alerting
30. Working with automakers on traffic signal alerting: • Green Light Optimal Speed Advisory • Time To Green • Fuel consumption reduction around 22% DTI: Traffic Light Alerting
31. If cars and traffic infrastructure are connected, we can use map data and predictive analytics to optimize traffic signal timing Your car approaches an intersection, there are no other cars approaching… • Traffic light tells car, I’ll stay green for you • Car maintains speed, does not have to slow down at intersection Communication handled on 4G LTE mediated by cloud or directly from vehicle to signal? DTI Vision: Optimized Traffic Signals
32. If we can do all this now over LTE, why wait to build a dedicated communications infrastructure? DTI Value to Transportation Departments
33. Insurance premiums past driving history, age and zip code-> Insurance premiums based on location and behavior DTI and Car Safety Technology will Have a Big Impact on Insurance • The combined financial impact of all these safety systems is projected to be massive • HERE is working with insurance companies to reassess their premiums and develop new insurance models for vehicles with advanced safety features and driverless modes • Location aware big data solutions can generate a more sensible pricing model
34. 34 • We have years of historic data on driving patterns • We can help insurance companies write better policies using data • Data can improve traffic predictions: – A car that left at 3:30 took 30 minutes to get to O’Hare – It is 4 PM now, but cars that leave at 4 PM on a Saturday usually take 1 hour - Historically a lot of cars have gotten on the road at the same time - We run over historic data and use that to improve our projected driving time • We also analyze driving data to determine historic speed profiles over a stretch of road – These profiles are attached as a layer to our HD Map – Self-driving car can drive to match economical, comfortable, or sporty profiles (Adaptive Cruise Control) • We mine the data using batch processes which saves money Driving Analytics in the Cloud
35. 35 • Speed profiles are an additional layer on top of our HD map How Fast Should the Car Drive You? Median Speed Faster Than Speed Limit
36. 36 • Amazon Web Services and Google Cloud have temporary computers available for a (generally) lower price • AWS lets you bid for “spot instances” that cost less than standard instances, so we can use them to run batch jobs – Downside, you get kicked off the servers when the price spikes • Google Cloud Platform has preemptible machines which are ~similar to spot instances • If you’re looking at a cloud platform and you need batch processing take this into account. – AWS spot instances used with an Elastic Map Reduce job could take a $30,000 run over a data set to $3,000 Public Cloud has made Data Analytics Cheaper AWS EC2 spot instance prices are generally ~10% of on demand instances, but sometimes people screw up their bidding code, leading to odd price spikes.
37. 37 • Deployment automation – Repeatable Tests - Executed when code is committed • Scalability – Environment needs to handle spikes in traffic caused by spikes in traffic • DevOps toolset helps manage change at scale • Resiliency testing – Chaos Monkey and the Simian Army DevOps Helps Support 10 Million Cars by 2020
38. 38 • Deploying a production-like environment is scripted (we let any developer do it) • This means more reliable tests • Unfortunately scripting deployments so they are easy has unplanned side effects: DevOps, Automate Everything
39. 39 Cost Control: Nightly Shutdown and Autotermination • Cloud hosting + deployment automation makes it easy for a developer to start an environment • We have a lot of servers, so we shut development hosts off at night – Nightly Shutdown is on by default, but many developers turn it off – We bribe them with candy and kittens – And automate termination of unused instances © 2016 HERE | DevOps
40. 40 • We’re in Chicago –We’re hiring – https://www.linkedin.com/company/heremaps/careers – https://company.here.com/careers/ • Detroit is also hiring –We need good data from safe cars! How Can You Help? http://www.builtinchicago.org/2015/10/14/2015-top-100-digital- companies-chi
41. 41 • GM bought Cruise for $1 Billion –Y Combinator startup –Pre-revenue • BMW, Mercedes, and Audi bought HERE for $3.1 Billion –We make digital maps for self-driving cars A New Gold Rush?