Using Inductive Loop Signature Re-Identification for Travel Time Measurement Use Case Bremerhaven

Transcription

1 Using Inductive Loop Signature Re-Identification for Travel Time Measurement Use Case Bremerhaven Jonas Lüßmann TU München, Chair of Traffic Engineering and Control (TUM-vt) Florian Schimandl TUM-vt Friedrich Maier Fritz Busch TUM-vt 6 th International Symposium Networks for Mobility Stuttgart, September 27,

2 Verstimmung in %o Detuning %o Inductive loops, vehicle signatures Nummerierung der Stützstellen (unnormiert) Time [s] ISAR: Inductive loops Signature Analysis for vehicle Re-identifcation and travel time measurement project at TUM-vt

3 ISAR-Method basics Standardised signatures!!! Relevant signature features: Signature derivative maximum detuning Filter criteria: Vehicle class Temporal distance Small similarity 2

4 ISAR-Method basics y A y B y A y B U uniformness, similarity cal calibration factor > x U Matching equation: A,B x End min(max y cal ;max y cal ) 1 min(y, n; y ) A A B B A,n B,n max(max y cal ;max y cal ) (x x 1) n x max( y ; y ) End Start A A B B Start A,n B,n Matching of the max. detuning (Signatures not standardised) Matching of the signature derivatives shapes 3

5 Field test Munich 2006 >2.000 vehicles at each cross section ~250 vehicle crossing both cross sections 38 correct reidentifications, 7 errors Error Examples 4

6 Project AMONES Aim: Evaluation of different methods of network control by analysing simulation results and collected data a Test site Bremerhaven (Northern Germany): 9 intersections controlled by traffic lights Intersections equipped with inductive loops Some ANPR-systems temporarilly installed Side effects of collecting inductive signatures: Evaluation of the ISAR-method using ANPRdata as reference If the ISAR-method works fine: additional data to evaluate the traffic light control algorithms Adapted from Openstreetmap 5

7 Test site installations (1) (2) (3) (4) (1) Control box with detectors and electric power supply (2) Computer to collect the loop detuning (10 PCs in Bremerhaven placed in the control boxes and connected to 20 detectors, sampling rate 125 Hz) (3) ANPR-Camera with Picture: Example from Munich (4) Camera computer (in Bremerhaven integrated in the camera body, supervised by the Institute of Road and Transportation Science of the University of Stuttgart) 6

8 AMONES: data collection reality Signature analysis for driving direction south: 2 PCs at (1) 2 approaches and (5) 1 PC at (3), (8) and (9) 3 PCs at (4), 2 approaches ANPR-detection at (1), (5) and (9) wrong volumes and useless signatures at (8) High in- and outflow rate between (5) and (9) due to large parking decks Bad prospects for vehicle re-identification ~ 25 re-identifications at (5)-(9) per day Between 80 and 90 re-identifications on the other sections No improvements with the data from (4) Let s see the results from (1)-(3) and (3)-(5) X X X X Adapted from Openstreetmap 7

9 Travel time [s] AMONES: Route (1,3) Travel times at route R(1,3) on February 17th 2009 Time of day No ANPR-Installation at (3) Qualitatively similar results on route (3,5) 8

10 Travel time [s] AMONES: Route (1,5) Travel times at route R(1,5) on February 17th 2009 Time of day ISAR travel times: addition of routes (1,3) and (3,5) 9

11 AMONES: Re-ident. frequency 10

12 Potential of travel times v/v 0 [-] ANPR-travel time [min] (route with many links) Function to estimate link-related v/v 0 : Using a segemented regression approach Using ANPR travel times as input data 11

13 Potential of travel times Temporal offset Temporal offset Functions to estimate link-related v/v 0 [-]: Using ANPR-travel times [min] as input data With different temporal offsets Functions to estimate travel times on a route [min]: Using local occupancy [%] as input data With different temporal offsets 12

14 Potential of travel times Estimation of link-related v/v 0 with ANPR-travel times as input data Sun,24:00 v/v 0 Mon, 0:00 A9 B13 A99 13

15 Last slide There is more information in vehicle signatures than only classification: they also allow the collection of travel time data to a certain extend These travel times can enrich the data basis for the evaluation of traffic management measures But: the data collection with our algorithm and equipment is still difficult Travel times include more than only travel times between two cross sections: in combination with historic fleet data they also offer estimations of the current spatial distribution of travel time loss on the detected route There is still a lot of information covered in the data we already collect we just have to elaborate it! 14

16 The end. Contact information: 15