DOWNLOADS


Mobile Century Data

UC Berkeley is pleased to release data from the Mobile Century field test. Learn more about the experiment at the project page.

In addition to the cell phone data collected during the experiment, the download includes data extracted from video recordings of the experiment, and data from in-road sensors that the project used to validate the cell-phone-generated traffic information.

As a public university, one of UC Berkeley's missions is to support research. Academic researchers may now download the field test data for non-commercial purposes. To download the data, use the form below to register your contact information and agree to the terms of use.

Mobile Century Manual Download: PDF (62 KB)

Mobile Century data documentation:  

1 USER AGREEMENT, TERMS OF USE

The Mobile Century data was collected on February 8, 2008, as part of a joint UC Berkeley - Nokia project, funded by the California Department of Transportation, to support the exploration of uses of GPS enabled phones to monitor traffic. In addition to the cell phone GPS data, two additional data sources are available for the experiment site. Inductive loop detector data obtained through the Freeway Performance Measurement System (PeMS), and travel time data obtained through vehicle re-identification using high resolution video data are included with this release. All identifiers assigned to the cell phones used during the Mobile Centuryexperiment have been randomized to protect the participants in the experiment. The video data is also processed and random number has been assigned to represent each vehicle. An extensive description of the experiment and data is available in the following article: Herrera, J.C., et al. "Evaluation of traffic data obtained via GPS-enabled mobile phones: The Mobile Century field experiment". Transport. Res. Part C (2009), doi:10.1016/j.trc.2009.10.006. Additional information about the successor project, known as Mobile Millennium, is available at http://traffic.berkeley.eduBy downloading the data, the user acknowledges that:

The data is available for use for research and analysis purposes only.
He/she will not redistribute the data.
Any publication using the data should refer the following article: Herrera, J.C., et al. "Evaluation of traffic data obtained via GPS-enabled mobile phones: The Mobile Century field experiment". Transport. Res. Part C (2009), doi:10.1016/j.trc.2009.10.006.

DOWNLOAD THE DATA (.zip 20 MB)

2  DESCRIPTION OF THE DATA

The data was collected during the Mobile Century experiment on Feb 8th 2008 between 10:00am and 18:00pm (PST) on Interstate 880, CA.

2.1  Inductive loop detector data

Inductive loop detector data, obtained from the Freeway Performance Measurement System (PeMS) database, consist of two sets: Northbound (NB) and Southbound (SB) 30 second raw loop detector data. Data are stored in two files:

  1. pems_rawdata_SB.csv and
  2. pems_rawdata_NB.csv.

These files contain the following columns: "pems_id", "unixtime", "flows" and "occs", see Table for the detailed explanation.

Table 1: PeMS raw data columns
"pems_id"
 

refers to the Vehicle Detector Station (VDS) ID of the loop detector station in the PeMS database

"unixtime"
 

refers to Unix time (seconds, UTC)

"flows"
 

refers to vehicle counts per lane over 30 second period

"occs"
 

refers to occupancy of the sensor per lane over a 30 second period

2.1.1  Inductive loop detector properties

Geographical locations of the Northbound and Southbound inductive loop detectors extracted from the PeMS database are given in two files:

  1. pems_prop_NB.csv and
  2. pems_prop_SB.csv.

These files have the following columns: "pems_id", "abs_pm", "lat" and "lon", see Table for detailed explanation.

Table 2: PeMS properties file columns
"pems_id"
 

refers to VDS ID of the loop detector station given by PeMS

"abs_pm"
 

refers to an absolute postmile (on Interstate 880) of the loop detector station given by PeMS

"lat", "lon"
 

refer to the latitude and longitude (degrees) of the loop detector station, respectively. Note, that lat/lon coordinates are approximate locations of the loop detector stations

2.2  Virtual Trip Line data

VTL (Virtual Loop Detector) data consist of two sets: Northbound (NB) and Southbound (SB) VTL speed data. Data are stored in two files:

  1. vtl_data_SB.csv and
  2. vtl_data_NB.csv.

These files contain the following columns: "vtl_id", "unixtime", "coordinate" and "vel_mph", see Table for detailed explanation.

Table 3: VTL data columns
"vtl_id"
 

refers to the integer ID of the VTL that the mobile device crossed when the data was sent

"unixtime"
 

refers to Unix time (seconds, UTC)

"coordinate"
 

refers to a latitude-longitude (degrees) coordinate pair that defines the geographical location of the measurement

"vel_mph"
 

refers to the reported speed (mph) of the mobile device when crossing the VTL

2.3  Ground truth travel time data

Ground truth travel time data is provided between Stevenson Blvd and Decoto Road and between Decoto Road and Winton Avenue for the Northbound direction of Interstate 880. Data are stored in files

  1. ground_truth_traveltimes_1_NB.csv and
  2. ground_truth_traveltimes_2_NB.csv,

respectively. The files contain two columns: "departure_time" and "travel_time", see Table for detailed explanation.

Table 4: Ground truth travel time data columns
"departure_time"
 

refers to Unix time (seconds, UTC) when vehicle started the trip (from Stevenson Blvd or Decoto Rd)

"travel_time"
 

refers to the travel time (seconds) experienced by the vehicle recorded (at Decoto Rd or Winton Ave)

2.4  Vehicle trajectory data

Two different types of vehicle trajectory data are provided. The first type described in Section consists of individual trips on one direction (NB and SB) of the highway. The second type described in Section consists of 77 individual phone logs. These logs contain the unprocessed latitude and longitude coordinates of the vehicles throughout the experiment.

2.4.1  Individual trip data

These data consist of individual "trips" on one direction of the highway. Northbound trips are in the "NB_veh_files" folder and the southbound trips are in the "SB_veh_files". Each file contains the following five columns: "unixtime", "latitude", "longitude", "postmile"1 and "speed"2, see Table for detailed explanation.

Table 5: Vehicle trajectory data columns
"unixtime"
 

refers to Unix time (seconds, UTC)

"latitude"
"longitude"
 

refer to latitude and longitude coordinates (degrees), respectively

"postmile"
 

refers to the number of miles from the start of Interstate 880 in California; a positive value represents the vehicle is on the Northbound side of the highway, a negative value represents the vehicle is on the Southbound side of the highway

"speed"
   refers to the speed (mph) derived from the lat-lon coordinates
2.4.2  Individual phone log data

Individual phone logs consist of 77 GPS log files extracted from the Nokia N95 mobile devices. Data are stored under the GPS_logsfolder in 77 files named vehXXX.csv where XXX is ranging from 101 to 177. Each file contains the following four columns: "unixtime", "latitude", "longitude" and "speed" , see Table for detailed explanation.

Table 6: Individual phone log data columns
"unixtime"
 

refers to Unix time (seconds, UTC)

"latitude"
"longitude"
 

refers to an absolute postmile (on Interstate 880) of the loop detector station given by PeMS

"speed"
 

refers to the speed (mph) derived from the lat-lon coordinates

 

 

 


 

LWR solver

The Lighthill-Whitham-Richards Partial Differential Equation (LWR PDE) is a seminal equation in traffic flow theory. It leads to common yet widely used traffic flow models for highways. This package proposes a sample implementation for a LWR solver using a new Lax-Hopf method. In the literature, the LWR PDE is typically solved using the Cell Transmission Model (CTM), a Godunov scheme, which requires a grid to compute the solution numerically, and induces specific approximations of the solution (in addition to the errors of the numerical computation).

The proposed approach has the following features:

  • It is greed free: in order to compute the value at a given point x and a given time t, it is not necessary to grid space and time (only to prescribe initial and boundary data, and to enter the point where the solution should be computed).
  • It is exact: the program presented here computes the solution by evaluating a function numerically (rather than approximating derivatives of the PDE by finite differences).
  • It is has small computational cost: since it is not needed to grid the space, the computation is much faster than with finite difference methods (Godunov, CTM, etc.)

To get started:

30-second how-to

  • script.m contains a sample script which creates a Lax-Hopf solver for a given Greenshields fundamental diagram and solves the Cauchy problem associated with an arbitrary set of initial densities, upstream and downstream flows;
  • two functions allow you to plot your results: LH_plot3D and LH_plot2D;
  • for advanced users, it is possible to solve the Cauchy problem associated with any concave fundamental diagram. Refer to the manual (PDF) for more details.

Important notes

  • The Matlab functions require Matlab version R2008A or newer. They may work under earlier Matlab versions, but have not been tested.
  • No additional toolboxes are required.
  • By downloading this software, you agree with the license terms.

References

More details about the foundations of this method can be found in the following publications:

 

 


 

Java Toolbox for Data Assimilation with Scalar Conservation Laws

The Riemann problem is a building block for constructing solutions to the Cauchy problem, and the initial-boundary value problem associated with a hyperbolic conservation law. The solution to the Riemann problem characterizes the propagation of discontinuities in the solution, corresponding to the propagation of queues of vehicles in partial differential equation models of traffic flow.

This package proposes a Java implementation of several advanced filtering methods for data assimilation with a macroscopic traffic flow model discretized using the Godunov finite volume numerical scheme and initial conditions corresponding to a Riemann problem.

A graphical interface is designed to allow efficient comparative analysis of the estimates produced by the EKF, the EnKF, and a Monte-Carlo forward simulation, in the case a scalar conservation law with flux function corresponding to a Greenshields, Newell-Daganzo, or quadratic-linear flux function. The values of the initial and boundary conditions noise, state noise, observation noise, and sampling rate can be set through the graphical interface.

A specific module is designed to illustrate the emergence of multi-modal distributions of the uncertainty on the true state, in the case of entropic shock waves.

Reproducible research

The joint use of the two testers allows reproduction of the main results presented in the manuscript “On sequential data assimilation for scalar macroscopic traffic flow models”, by Sebastien Blandin, Adrien Couque, Alexandre Bayen, and Daniel Work, accepted for publication to Physica D: Nonlinear Phenomena on May 2012.

To get started

Alexandre M. Bayen

Department of Civil & Environmental Engineering
Department of Electrical Engineering and Computer Sciences
University of California, Berkeley