Webinar Series TMIP VISION

Transcription

1 Webnar Seres TMIP VISION TMIP provdes techncal support and promotes knowledge and nformaton exchange n the transportaton plannng and modelng communty.

2 DISCLAIMER The vews and opnons expressed durng ths webnar are those of the presenters and do not represent the offcal polcy or poston of FHWA and do not consttute an endorsement, recommendaton or specfcaton by FHWA. The webnar s based solely on the professonal opnons and experence of the presenters and s made avalable for nformaton and experence sharng purposes only.

3 The most mportant part of a speech s the openng lne. When tme s not a factor, I lke to try out 3 or 4 dfferent ones. Steve Carell (portrayng Mchael Scott n The Offce ) The plural of anecdote s data Raymond Wolfnger

4 The need for speed n travel modelng Data sources for road segment speeds (floatng car, spot speeds, GPS (va data vendors or local nfrastructure), other) Thngs to watch for n the data Items to thnk about The need for data on varablty/relablty Data sources on varablty n speeds/travel tmes (n addton to average values)

5 because our customers want to know:

6 Where: (Ideally) all routes where one has some jursdcton (wth approprate segmentaton) When/whch condtons: free flow, average, medan, and crtcal plannng stuatons (comparable to the desgn hour ) Other tems to thnk about (next sx sldes):

7 For perspectve, 95 th Percentle tme corresponds to the 438 th hghest hour of year, project desgn typcally uses 30 th or 100 th hghest hourly volume n year. 95 th (or even 85 th ) hghest travel tme *mght* correspond to a bottleneck condton wth low volume throughput. Level Of Servce crtera sometmes a functon of travel tme and sometmes a functon of volume

8 Varablty n travel tme (for plannng trps or tmng an arrval), or The percentage of the tme that the trp took the expected amount of tme

9 The average travel speed durng low-volume condtons? (typcally used n travel models) A certan percentle value of speed across the entre tme perod of nterest? (hstorcally, 85% - from desgn speed/speed zone studes) A percentle value of hourly average values ( reference speed)? Can be crtcal n ntal development of model network speeds

10

11 GPS readngs on vehcles not very revealng so far on the value of tme vs dstance vs other. Are the choces avalable on the nternet revealng n any way?

12 Wthn samplng error of studes (.e. wthn 95% confdence nterval) rule from smulaton studes (85% of the tme the modeled travel tme should be wthn 15% (or 1 mnute) of the feld measured travel tme)

13 CMS / project beneft / sgnal system upgrades Used for MPO travel model valdaton snce 1990 s to focus on depcton of congeston/los Statewde, used to develop speed table by type of road both average and runnng speeds (latter to start up some juncton-based model networks across Oho)

14 hgh sample sze floatng car Can use to measure varablty n travel tme as well as havng more confdence n the average (along wth how varablty changes movng from lnk-level to path-level)

15 Some state DOT PATR statons (mostly rural Interstate) and portable counts ITS montorng network (same as used for VMS) Weather montorng statons (RWIS) Relatonshp between spot speeds and the space mean speeds s usually good on freeways, but often a problem on nterrupted flow facltes.

16 Average & 85 th Percentle Speeds (3/2008)

17 Example from local agency (can jon to sgnal/stop sgn nventory to determne for whch locatons/drectons the spot speeds are not lkely to be useful).

18 Doppler Statons so densely spaced that ITS secton uses to estmate segment as well as pont speeds (combned wth GPS probe data), RWIS statons collect volume and speed as well

19

20 Archve data from vehcle fleets & cell probes Extensve road network coverage, could replace or reduce/redeploy need for floatng car surveys

21 Prvate vendor (TMC) network extends to urban arterals n larger urban areas (over ½ mllon pop) NPMRDS (FHWA program) on the NHS only, but provdes separate car and truck speeds

22

23 Focus s on mnmzng the amount of tme a state hghway has an average travel speed below the posted speed lmt. Both ODOT s sensors and GPS data used.

24 Of course not... - due to: Samplng errors Measurng (slghtly?) dfferent thngs MAYBE some sources have more QC ssues than others The best source could well depend on what exactly s your applcaton

25 Dfferences exsts n how these are measured (spot vs space mean speeds) Statewde, average speeds hgher on the ATR s (about 7%) Items to watch for: caps on reported speeds, data flterng, any bases n the wred drvers and vehcles compared to general traffc

26 Dfferences exst n route segmentaton Very small sample szes n the floatng car surveys Statewde, n close agreement on overall average speeds

27 Dfferences n TMC segmentaton frustrate lnk-level comparsons. Statewde, average speeds outsde the Interstate system are slower usng NPMRDS data (how much so depends on level of flterng )

28 Far hgher (f weghted?) sample szes, more versatlty on hour of day / day of week / season of year Good for overall speed valdaton of model on average values by segment however: Dependng on level of access, mght not have ablty to see full journey tmes or dstrbutons of segment tmes Buffer ndex values reflect system-level, not user-level varablty

29

30 From Phoenx (by hour & day of week): Natonwde: And from Florda research:

31 Volume-based mpacts offset by dfferences n drver and vehcle characterstcs, Sgnal tmng plans, parkng management Is V/C that mportant?

32

33 Local control over route coverage and segmentaton Example corrdor below (US40 n Columbus nsde I- 270): there s 33 TMC segments each way (good major street coverage), but also 78 sgnals Smaller areas (hstorcally) have even less TMC coverage

34 From Merdan Rd to I-680, there s 6 TMC segments each way (lttle local road coverage), and 17 sgnals

35 MPO models already exstng & under development Travel paths account for (and weght) varablty as well as average travel tmes therefore, can forecast future levels of varablty as well as average congeston. Data needed for lnk & corrdor level varatons n tme, and correlatons between adjacent road segments (RR can be ether asserted, or revealed n the valdaton)

36 Total Path Impedance (RR = relablty rato) Varance of Path Travel Tme (Thru movements on successve lnks are correlated.) Margnal change n travel tme standard devaton from selectng the next lnk: So that for any path between an orgn and destnaton: And for each lnk: (t=tme, L=length, coeffcents vary by road type) path path path path y t RR I + + σ = = + + = σ σ + σ = σ , M M path r , , σ σ + σ σ σ + σ = = + + = = + + = k k k k k r r R = = σ M path R 1 φ δ γ = L t t CV 0

37

38 Started wth fndngs from Brtsh GPS study, but w/o access to ther data can use local GPS and (hgh-sample) floatng car to develop local values. CV equaton coeffcents qute varable by local source, tests va model performance vared wthn statstcal nose. Adjacent-lnk travel tme correlatons we removed f connected by left or rght turn. Local data so far fnds lower values for arterals than Brtsh study but hgher for freeways, both dependent on tme of day and type of data used. (No correlaton means path tme varance s the sum of the lnk tme varances re earler chart.)

39 The need for speed n travel modelng Sources for average road segment speeds (floatng car, spot speeds, GPS (va data vendors or local nfrastructure), other) The need for (and modelng) varablty/relablty Data sources on varablty n speeds/travel tmes (n addton to average values) Fnally, don t elmnate but reposton any efforts you re makng on local floatng car surveys

40

41 TMIP Updates For future webnar announcement, please sgn up for GovDelvery at f you have not done so.

42 TMIP Contacts If you have any questons or comments about today s presentaton or TMIP, or f you are nterested n sharng your experence, please contact me at: sarah.sun@dot.gov or feedback@tmp.org.