1 / 30

Freeway Segment Traffic State Estimation

Freeway Segment Traffic State Estimation. Heterogeneous Data Sources and Uncertainty Quantification: A Stochastic Three-Detector Approach. Wen Deng Xuesong Zhou University of Utah Prepared for INFORMS 2011. Needs for Traffic State Estimation. Sensor Data. Traffic State Estimation.

tahlia
Download Presentation

Freeway Segment Traffic State Estimation

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. Freeway Segment Traffic State Estimation Heterogeneous Data Sources and Uncertainty Quantification: A Stochastic Three-Detector Approach Wen Deng Xuesong Zhou University of Utah Prepared for INFORMS 2011

  2. Needs for Traffic State Estimation Sensor Data Traffic State Estimation Traffic Flow/Control Optimization

  3. Motivating Questions • How to estimate freeway segment traffic states from heterogeneous measurements? • Point mean speed • Bluetooth travel time records • Semi-continuous GPS data Semi-continuous path trajectory Continuous path trajectory Point Point-to-point Loop Detector Automatic Vehicle Identification Automatic Vehicle Location Video Image Processing

  4. Motivating Questions • How much information is sufficient? • How to locate point sensors on a traffic segment? • How to locate Bluetooth reader locations? • How much AVI/GPS market penetration rate is sufficient?

  5. Existing Method 1: Kalman Filtering • Eulerian sensing framework • Muñoz et al., 2003; Sun et al., 2003; Sumalee et al., 2011 • Linear measurement equations to incorporate flow and speed data from point detectors • Extended Kalman filter framework • second-order traffic flow model • Wang and Papageorgiou (2005)

  6. Existing Method 2: Cell Transmission Model • Cell inflow inequality qi,j(t) = Min { vfreeki,j(t) , qmaxi,j(t) , w  (kjam - ki,j(t)) Δ x } • Switching-mode model (SMM) • set of piecewise linear equations • qi,j(t) =  [vfreeki,j(t) ] +  [vfreeki,j(t) ]

  7. Existing Method 3: Lagrangian sensing • Nanthawichit et al., 2003; Work et al., 2010; Herrera and Bayen, 2010 • Establish linear measurement equations • Utilize semi-continuous samples from moving observers or probes

  8. Existing Method 4: Interpolation method • Treiber and Helbing, 2002 • “kernel function” that builds the state equation for forward and backward waves • Linear state equation through a speed measurement-based weighting scheme Figure Source: Treiber and Helbing, 2002

  9. Challenge No.1 • 1. Unified measurement equations to incorporate • Point, point-to-point and semi-continuous data

  10. Our New Perspective • Dr. Newell’s three-detector model provides a unified framework • N(t,x)=Min {Nupstream(t-BWTT)+Kjam*distance, Ndownstream(t-FFTT)}

  11. 1: From Point Sensor Data to Boundary N-curves • Cell density and flow are all functions of cumulative flow counts

  12. 2: From Bluetooth Travel Time to Boundary N-curves • Downstream and upstream N-Curves between two time stamps are connected

  13. 3: From to GPS Trajectory Data to Boundary N-curves • Under FIFO conditions, GPS probe vehicle keeps the same N-Curve number (say m) m m m m m

  14. 4: From Boundary N-curves to Everything inside

  15. Challenge No. 2 • All sensors have errors error propagation

  16. The Question We have to Answer • Under error-free conditions, Newell’s model provides a good traffic state description tool N(t,x) =Min {Nupstream(*), Ndownstream(*)} • With measurement errors • What are the mean and variance of • Min {Nupstream(*)+eu, Ndownstream(*) +ed}

  17. Quick Review: Probit Model and Clark’s Approximation • Probit model (discrete choice model for min of two alternatives’ random utilities ) • U = min (U1+e1, U2+e2) • Route choice application • Clark’s approximation minimization of two random variables can be approximated by a third random variables

  18. Proposed Stochastic 3-Detector Model

  19. Discussion 1: Consistency Checking When Uncertainties of boundary values are 0, the stochastic 3-detector model reduces to deterministic 3-detertor model

  20. Discussion 2: Weights under Different Traffic Conditions

  21. Discussion 3: Quantify Uncertainty of Inside-Traffic-State Estimates • Variance or trace of estimates determine the value of information

  22. Numerical Example

  23. Input: Queue Spillback

  24. Ground truth Arrive-departure Curves

  25. Estimated Density Profile

  26. Estimated Uncertainty profile Before After

  27. Impact of Additional Sensors

  28. Possible (Un-captured) Modeling Errors Stochastic free-flow speed, Stochastic backward wave speed; Heterogeneous driving behavior • Upper plot: original NGSIM vehicle trajectory data • Lower plot: reconstructed vehicle trajectory based on flow count measurements

  29. Conclusions • Proposed stochastic 3-detertor Model • Estimate freeway segment traffic states from heterogeneous measurements • Quantify the degree of estimation uncertainty and value of information, under different sensor deployment plans

More Related