Smart cars of tomorrow: real-time driving patterns
summary
Prototype streams car data from a Bluetooth OBD II dongle through an iOS app Herbie into a Spring XD pipeline, stores raw traces in Hadoop, caches live state in GemFire, and retrains models via PySpark driving scikit-learn. The claim is real‑time destination prediction and range estimation using k‑means clustering plus random forests with per‑vehicle fingerprints; architecture reportedly scales to 100,000 devices, millions of messages per minute and 100 GB, the demo has flaky connectivity but models correctly converge to the true destination in recorded runs.
AI-written summary · about
chapters (11)
- 0:00Connected car prototype intro
- 1:13Project origin and tech stack
- 2:58Live demo setup
- 8:30Dashboard and real-time predictions
- 12:27Technical architecture
- 19:01Destination prediction science
- 26:37Model training with Spark
- 30:34Dashboard scalability
- 31:39IoT challenges
- 32:45Future plans
- 34:34Q&A
related talks
38mSampling and Modelling the World in Real-Time for Mobility
23mMaking Sense of Streaming Sensor Data: How Uber Detects On-trip Car Crashes
20mInnovating at speed
08mdata science in automotive
46mConnected Vehicle is far more than just the car
41mMobileye: Smarter, safer autonomous solutions with Habana & Granulate (PRT279)
22mBuilding a movement data and analytics platform to de-stress customers & save lives
34mData Drives Business