M a p s 2. 0 T h e c o o p era tive w ay t o ke e p m a p s u p t o d a t e O l iver Kü h n s k obble r G m b H M a i 2 0 1 3
This is a stor y about how we all get unconscious ly involved in creating digital maps and keeping them up to date 2
Digital Maps The classical way 3
The classical way From capturing to the car 12 to 24 months Well trained people are measuring the world When these data become available some of the data are already outdated 4
The classical way Evaluation Pre dictability Ve r y good Accuracy Ve r y high Atte ntion-to-de tail Limite d Turn-around-time Ve r y s low Scalability Low 5
Digital Maps The crowdsourcing approach E x a m p le C o u ntr y G e r many Each day, 7 days a week 500-600 mappers are active roughly 100.000 nodes are created almost 200 new nodes are generated per mapper People like you and me are measuring the world 6
The OpenStreetMap community grows continuously and is ver y active 2013 7
Crowdsourcing vs. Commercial maps Example Berlin Zoologischer Garten OSM Nokia 8
The crowdsourcing approach Measuring Maturity First they ignore you, then they laugh at you, then they fight you, then you win 9
Status Crowdsourcing Maps TomTom star ts fighting OpenStreetMap 10
The crowdsourcing Approach From capturing to the car Several days Editing the map in real-time T he freshness only depedents on the snapshot date from the live database 11
The crowdsourcing approach Evaluation Pre dictability Low Accuracy Varie s Atte ntion-to-de tail High Turn-around-time Ve r y f as t Scalability High 12
A new way of mapping Car Sensor Mapping A dditional computing power in the car allows decentral processing of sensor data. 13
Car Sensor Mapping Making processed(!) car sensor data centrally available 14
Car Sensor Mapping Example: Identifying road conditions 15
Car Sensor Mapping Example: Identifying free parking lots 16
Car Sensor Mapping Identifying turn restrictions, one-way-streets etc. 17
Car Sensors Mapping The challenge: processing the large amount of data Multiple reports per data point 18
Car Sensors Mapping From capturing to the car Immediately Processed sensor data are transmitted over the air The principle is known from traffic data (mainly unprocessed sensor data) 19
Car Sensor Mapping Evaluation Pre dictability Limite d Accuracy Ve r y high Atte ntion-to-de tail Limite d to s e le cte d attribute s Turn-around-time Ve r y f as t Scalability Ve r y high 20
The future Ope nstre e tmap C rowds ource d Data High Accuracy Inte llige nt Proce s s ing Low C os t Production C ar Se ns or Data Fas t Re le as e Proce s s 21
The future skobbler s Floating Car Data 22
The future skobbler s Floating Car Data 23
The future skobbler s Floating Car Data 24
The future skobbler s shor t term FCD targets Topic Description Functional clas s ing C las s if ication Re al s pe e d Turn re s trictions, one -way and mis s ing s tre e ts C onne cte d s tre e ts Road change de te ction W e will genera te functiona l roa d cla sses for routing/rendering, a s this is a ma jor disa dva nta ge of OSM compa red to commercia l ma ps (ma y be genera ted ba sed on FC D). W e will unify a nd cla ssify a ll streets into pre-defined street types (ba sed on OSM types a nd FC D da ta ). As OSM only a llows fa cts in its da ta ba se, it completely la cks rea listic/empirica lly a vera ge movement speeds on a ny pa rticula r street. We ca n a dd this via FC D. Turn restrictions a nd one-wa y streets ca n be corrected / a dded to the ma p by interpreting FC D. Missing streets ca n be identified (for ma nua l correction) or a dded a s low-level streets (without na me, etc.). W e detect wrongly connected / non-connected streets a nd correct this ba sed on FC D da ta. W e a utoma tica lly detect construction sites (e.g. on highwa y) a nd roa d closures by compa ring old a nd recent FC D da ta, highlighting spots to community/a utomated ma pping. 25
Contact me: oliver.kuehn@skobbler.com