Agenda Identification of Hazardous Road • Why is Hazardous Road Location treatment necessary? Locations on the basis of Floating Car • A need for new methods? Data • The Method briefly • Scientific background • Data source • Results and status Niels Agerholm Associate Professor Division of Transportation Engineering Dept. of Civil Engineering Aalborg University ICTCT workshop 2015 Niels Agerholm, Traffic Research Group, Aalborg University No. 1 of 20 ICTCT workshop 2015 Niels Agerholm, Traffic Research Group, Aalborg University No. 2 of 20 A definition A need for new methods? Reporting rate (%) • Hazardous Road Location (HRL) is also known as Black Spot • How comprehensive are the official Car Motor Cyclists • HRL: traffic accident databases? occupant cycle s riders A location which is more accident-prone than it it should be expected due to • Large variation from country to traffic level, road furniture and road design. country Australia 73 53 7 • Depends significant of the involved Denmark 48 31 10 road users France 63 45 11 • Significant dark figure • 33% reduction in injury accidents! Germany 52 44 22 Great Britain 68 44 66 The Handbook of Road Safety Measures - Second Edition (2009) The 63 56 24 Netherlands Norway 56 37 16 Sweden 77 67 29 USA 65 - 26 The Handbook of Road Safety Measures - Second Edition (2009) (Old results in general)!!! ICTCT workshop 2015 ICTCT workshop 2015 Niels Agerholm, Traffic Research Group, Aalborg University No. 3 of 20 Niels Agerholm, Traffic Research Group, Aalborg University No. 4 of 20 A need for new methods? A need for new methods? • An exceptional Danish challenge Small-scale Danish studies underpin the Accident reporting: Denmark • Small-scale studies underpin this need for a new/ additional method 1998 2007 2011 • Only 10-12% injury accident were No. of injured reported Police 9,660 6,897 4,158 • Especially accidents with young people have high black figures Police & Hospital 46,075 47,792 41,272 • Only 5 of 15 HRL intersection Police-reported share 21% 14% 10% were found from both Police and Hospital data Statistics Denmark 2015 • Occasionally erroneous treatment in HRLs due to low share of reported accidents (Andersen & Sørensen 2004; Jørgensen & Bach 2007; Bunton & Celis 2009) ICTCT workshop 2015 Niels Agerholm, Traffic Research Group, Aalborg University No. 5 of 20 ICTCT workshop 2015 Niels Agerholm, Traffic Research Group, Aalborg University No. 6 of 20
Hypothesis The Method briefly It is the hypothesis that concentration of registered • In principle it is an area-based conflict study technique hard decelerations on locations indicates a potential • Based on Floating Car Data (FCD) HRL • The idea is that HRLs induce more abrupt decelerations than else ( jerks ) • Too many in one location indicates a HRL Small scale studies indicate that the jerk • gives a more clear pattern (the derived of Serious Jerks the deceleration) Accidents Moderate Jerks • A Forward-looking approach (don’t wait Fatal accidents Injury accidents until the accidents appear) Vision Zero !! • Normal driving Inspired by Svensson & Hydén 2006 ICTCT workshop 2015 Niels Agerholm, Traffic Research Group, Aalborg University No. 7 of 20 ICTCT workshop 2015 Niels Agerholm, Traffic Research Group, Aalborg University No. 8 of 20 Conflict indicator? Speed, Decelerations, and Jerks – connection • Avoidance can be made in three ways: • Decelerate • Accelerate • Sideways • Decelerations (and jerks) are selected: • Intuitively • Support from literature: 72-98% of all accident avoidance activities (Horst 1984, Hydén 1987, Hantula 1994, Nygård 1999) ICTCT workshop 2015 ICTCT workshop 2015 Niels Agerholm, Traffic Research Group, Aalborg University No. 9 of 20 Niels Agerholm, Traffic Research Group, Aalborg University No. 10 of 20 Scientific background Data from ITS Platform • Nygård (1999): • Floating Car Data: GPS data from Found that serious conflicts resulted in jerks that differed significant from jerks in case of voluntary • vehicles carrying out their ‘normal’ braking purpose on the road network Found that jerks gave more clear results than decelerations • • ITS Platform: • 10 Hz Serious Jerks • Svendsen et al. (2008) Accidents • 380 cars Moderate Jerks Supported Nygård’s jerk -based approach Fatal accidents • • Acceleration data included Injury accidents • Used low frequency data (1 Hz) • ~15 Billion records on accelerations • ~14.5 million km Normal driving • Agerholm & Lahrmann (2012) • ~1.4 million trips • Simplification and selection is required • Speed changes are very sensitive to bad GPS connections Low speed: deceleration and jerks were unreliable • Jerks – and partly decelerations are highly affected by uneven surfaces (Speed bumps, pot • holes, bad managed road surface etc.) ICTCT workshop 2015 Niels Agerholm, Traffic Research Group, Aalborg University No. 11 of 20 ICTCT workshop 2015 Niels Agerholm, Traffic Research Group, Aalborg University No. 12 of 20
From raw data to relevant data Study approaches Stepwise data treatment: 9 approaches in total • 3 approaches with 3 sets of variations: • Each car: 1,000 highest decelerations and jerks in 3 months Decelerations • and similarly 1,000 jerks Negative jerks ( jerks ) • • Anonymisation Peak-2-peak jerks ( p2p jerks ) • • Selected area: Aalborg Municipality Indicator variations: • Removal of small decelerations / jerks The limiting |size| of decelerations / jerks / p2p jerks • • Removal of data on motorways • Number of registrations (high speed often make vibrations) • Removal of low speeds (<20 km/h) • Removal of data around speed bumps (~2,300 speed bump in North Jutland) ICTCT workshop 2015 Niels Agerholm, Traffic Research Group, Aalborg University No. 13 of 20 ICTCT workshop 2015 Niels Agerholm, Traffic Research Group, Aalborg University No. 14 of 20 What do the decelerations show? What do the jerks show? ICTCT workshop 2015 ICTCT workshop 2015 Niels Agerholm, Traffic Research Group, Aalborg University No. 15 of 20 Niels Agerholm, Traffic Research Group, Aalborg University No. 16 of 20 What do the p2p jerks show? Lessons learnt Data still require significant cleaning effort � • Decelerations can hardly be seen as a risk indicator � • • Jerks seems to indicate some HRLs although uneven surface is a problem • P2P jerk seems very sensible to pot holes, speed bumps and uneven road surface in general � ICTCT workshop 2015 Niels Agerholm, Traffic Research Group, Aalborg University No. 17 of 20 ICTCT workshop 2015 Niels Agerholm, Traffic Research Group, Aalborg University No. 18 of 20
Next steps • Ensure ≥ 1,000 jerks ( deceleration ?) from all cars (~380) where all data due to uneven surface are removed . • Make density analyses Thank You ☺ • Identify locations with high concentrations • Search careful for any data errors affecting the reliability. • Make inspection on the location (Street View and real inspection) • Compare 2 six-months periods • Verify/falsify the hypothesis Niels Agerholm Robin Jensen Camilla Sloth Andersen Associate Professor M.Sc. in Eng. Assistant Professor Division of Transportation COWI Ltd Division of Transportation Engineering Engineering Dept. of Civil Engineering Dept. of Civil Engineering Aalborg University Aalborg University +45 61 78 04 55 na@civil.aau.dk ICTCT workshop 2015 Niels Agerholm, Traffic Research Group, Aalborg University No. 19 of 20 ICTCT workshop 2015 Niels Agerholm, Traffic Research Group, Aalborg University No. 20 of 20
Recommend
More recommend
