Optimal sensor placement for structural health monitoring ...
Optimal Sensor Placement for Traffic Data Collection: Case Studies and...
Transcript of Optimal Sensor Placement for Traffic Data Collection: Case Studies and...
Optimal Sensor Placement for Traffic Data Collection: Case Studies and Challenges
Jeff Ban & JD Margulici March 4, 2008 California 511 Workshop
PDF created with pdfFactory trial version www.pdffactory.com
Traffic Data Collection Techniques
PDF created with pdfFactory trial version www.pdffactory.com
Collected thoughts about traffic
Most sensors and communication infrastructure are installed on a case-by-case basis without knowing whether the associated benefits are fully realized
Caltrans does not have a decision support tool to help evaluate and justify sensor deployment from a system-wide perspective
Optimal sensor deployment strategies should be developed in the context of specific applications for different types of corridors such as rural/mid-size/urban
sensors
PDF created with pdfFactory trial version www.pdffactory.com
Deploying traffic sensors
¤ Questions:
⁄ How many?
⁄ Which kinds?
⁄ Where?
⁄ To what benefits?
⁄ At what cost?
¤ Why they matter:
⁄ Attaining functional objectives
⁄ Making the right choices
⁄ Providing standard guidelines
⁄ Justifying budget changes
PDF created with pdfFactory trial version www.pdffactory.com
Case Study: Optimal Sensor Placement for Freeway Travel Time Estimation
PDF created with pdfFactory trial version www.pdffactory.com
Application: Displaying Travel Times on CMS
SJ DWNTWN VIA 101 60 MIN VIA 880 40 MIN
Loop detectors
ETC readers 60 min
40 min
PDF created with pdfFactory trial version www.pdffactory.com
A Dynamic Programming Model
¤ Empirical Studies ⁄ Thomas (1999), Eisenman et al. (2006), Liu et al. (2006), Fujito et al.
(2006), Kown et al. (2006), Ban et al. (2007) ⁄ Based on existing sensor deployment, investigate how changes of sensor
locations impact the performance of travel time estimation. ¤ Optimal Sensor Placement Study Sponsored by Caltrans
⁄ Investigate the requirements for numbers and locations of sensors to collect traffic data for 1) travel time estimation, 2) ramp metering control, and 3) freeway performance monitoring.
¤ Current Findings ⁄ We formulate the problem using Dynamic Programming, which can be
solved optimally in polynomial time ⁄ Test the model and solution algorithm using both simulation and real world
data from GPS-Enabled Cell Phones.
PDF created with pdfFactory trial version www.pdffactory.com
Numerical Results Using Micro-Simulation Data
PDF created with pdfFactory trial version www.pdffactory.com
Simulation Network (I-405 in LA)
PDF created with pdfFactory trial version www.pdffactory.com
Optimal Locations for 6 Sensors
PDF created with pdfFactory trial version www.pdffactory.com
Evolution of Optimal Sensor Locations
Major BN
Free Flow
Minor BN
Free Flow
PDF created with pdfFactory trial version www.pdffactory.com
Numerical Results Using GPS-Equipped Cellular Phone Data
PDF created with pdfFactory trial version www.pdffactory.com
Cell Phone Data
PDF created with pdfFactory trial version www.pdffactory.com
Data Collection
¤ 20 Cars Equipped with Nokia GPS N95 Cell Phone, looping between Alvarado Niles Rd and CA-92 (about 3 miles) from 1:00 pm to 5:00 pm on Nov. 02, 2007.
¤ Average loop travel time is about 20 minutes, equivalently 60 veh/hour, which is about 1% of the total freeway volume on that day (6000 veh/hour).
¤ Collect trajectories of looping vehicles
¤ Estimated speed fields of the study route for 2:00 pm – 4:00 pm.
PDF created with pdfFactory trial version www.pdffactory.com
Optimal Locations for 6 Sensors
15:00 15:30 16:00 14:30 14:00
20-Car Experiment
PeMS Data
PDF created with pdfFactory trial version www.pdffactory.com
Evolution of Optimal Locations
PDF created with pdfFactory trial version www.pdffactory.com
Performance Comparison of Optimally Deployed Sensors with Evenly Spaced Sensors
PDF created with pdfFactory trial version www.pdffactory.com
Challenges: How to Value Information Quality (Accuracy, Reliability, etc.)?
PDF created with pdfFactory trial version www.pdffactory.com
Travel Time Estimation Error vs. # of Sensors (via Simulation Data)
MSE
0 2 4 6 8 10 12 14 16 18 20
# of Sensors
0.11
0.1
0.09
0.08
0.07
0.06
PDF created with pdfFactory trial version www.pdffactory.com
Potential Directions
¤ System Perspective: information coverage (geographical and demographic), system performance improvement (delay reduction, accident/incident reduction)
¤ User Perspective: travel time reduction, willingness to pay
PDF created with pdfFactory trial version www.pdffactory.com
Optimal Sensor Placement for Traffic Data Collection: Case Studies and Challenges
Jeff Ban & JD Margulici March 4, 2008 California 511 Workshop
PDF created with pdfFactory trial version www.pdffactory.com