Travel Time And Alternate Route On CMS

1
Travel Time And Alternate Route On CMS
Chao Chen, Karl Petty PeMS Development
Group University of California,
Berkeley 10/24/2002
2
Provide real-time information where its most
helpful

• Travel times
• Alternate routes
• Incidents
• Updated automatically

3
Benefits

• Minimize travel time by picking the best route
• Reduce uncertainty in travel time
• Distribute load to less congested corridors
• Quantify delays

4
Feasibility

• There are many CMS on CA freeways
• PeMS already measures and predicts travel times
• The information is easily made available
• A similar CMS travel time service exists in Paris
  and works well

District 12 CMS locations
5
CMS inventory
6
Objective of this study

• Quantify benefit of travel time on CMS
• Choose locations where potential benefits are
  greatest
• Alternate routes are available
• Lots of congestion

7
Set-up of the problem
2
1
8
More set-up
9
Travel time variability

• Travel times for 20 days in October, 2000, on
  I-10E, between pm 1.3 and 48.5, starting every
  5-min, between 5 am and 8 pm
• Shows large variation between 42 and 130 minutes
• Permits prediction of travel time

10
Travel time prediction30 min forecast
11
Travel times on two routes in San Diego

• Congestion occurs on both routes
• Travel time varies between 15 and 45 minutes

• Average travel times are similar these are
  equally good alternatives

12
Travel Time Prediction Algorithm

• Given V(d,x,t), velocity at location x on day d
  at time t
• Estimate Te (t s), travel time starting at t s
• M(t s) historical average of T(t s)
• T(t) travel time with frozen velocity field
  V(l, t )
• T(t s ) a(t, s) T(t) b (t, s) M(t s)
  error
• Obtain a(t, s), b (t, s) by least-squares
  regression

13
Travel Time Estimation

• Speed contours from loops dark regions show
  times and places of low speed
• White lines show computed vehicle trajectories
• Estimated travel times match those of TACH
  measurements

14
TACH And PeMS Travel Time Comparison

• 25 TACH runs, mean travel time 12 min
• RMS error 1.7 min
• R2 0.75

15
Prediction minimizes travel time

• Each dot represents one trip
• Prediction almost always picks the best route
• Even when prediction doesnt pick the best route,
  the travel difference is small

16
Average travel time reduction

• 2-minute reduction for 23 minute trip, about 9,
  during peak hour
• Predictoin achieves close to minimum travel time

17
Prediction reduces uncertainty

• Prediction is accurate to 2 minutes on 23 minute
  trip
• More than 50 reduction in uncertainty in peak
  hours

18
List of study areas

• District 7 Los Angeles and Ventura Counties,
  from I-10 and I-210 to downtown
• District 11 San Diego, from I-5 and I-805 to
  downtown (I-5 and I-163)
• District 11 from SR-125 and I-8 to downtown
• District 12 Orange County, from I-5 and I-405
  to I-5 and SR-91
• District 12 from I-5 and I-405 to I-405 and
  I-605

19
Summary of several study areas
20
Still to do

• Do same study for more O-D locations
• Decide how to interface PeMS and Caltrans CMS
  system
• Make sure PeMS travel times are accurate

21
Conclusion

• Travel time prediction on CMS can reduce travel
  time when alternate routes exist
• It greatly improves travel time predictability