1. Arterial Loop+Racetrack No-Incident Model
July 23, 2014

2. Outline Arterial Loop+Racetrack No-Incident Model
Model components (with calibration set)
Model run
Model validation (with validation set)

3. Model Components Generate FD Fixed capacity Fixed jam density
Data Processing
Model Components: FD / BF / SR / Signals
CTM Forward Simulation
Metrics: TT, Delay, LOS, VHT, VMT
Comparison: Travel Time, Flow
Model Components
Generate FD
Fixed capacity
Fixed jam density
Speed limit from
Nokia map
cc-network FD
SR/BF Calibration
Racetrack flow 1
Racetrack SR 2
SR, BF
cc-scenario building
Loop flow 3
Loop SR 4
Signal Timing 6
cc-scenario

4. Family of Fundamental Diagrams
[ Input for CTM Forward Simulation of Arterial Loop+Racetrack No-Incident Model ]
Family of Fundamental Diagrams
Fixed jam density = 150 veh/km/lane
Fixed capacity = 1800 veh/hr/lane
Free flow speed = speed limit from Nokia maps
Flow
Density

5. [ Input for CTM Forward Simulation of Arterial Loop+Racetrack No-Incident Model ]
SR/BF Calibration
Check whether side street BF reasonable
Apply engineering judgment SR
Set SR
BF Optimization
Generate BF profile
Racetrack SR 2 SR BF
loop SR 4
Default SR
Flow
Time
Target flows
loop flow 3
Racetrack flow
where loop is not available at the same location 1b
Scale racetrack flow
Generate
Generic flow profile
Racetrack flow
where loop is available
at the same location 1a
loop flow 3
Default BF

6. Process to Generate Split Ratios
[ Input for CTM Forward Simulation of Arterial Loop+Racetrack No-Incident Model ]
Process to Generate Split Ratios
Use racetrack data
Use default values when racetrack data unavailable
Apply engineering judgment to obtain reasonable boundary flows in subsequent steps

7. [ Input for CTM Simulation of Arterial Loop+Racetrack No-Incident Model ]
90%
10%
I210W
Split Ratios
69%
31%
Color coding:
Racetrack data
49%
24%
58%
18%
Default values
44%
90%
Engineering judgment 7%
100%
10%
Split ratio from loop
11%
90%
90%
Colorado Blvd
55%
45%
93%
77%
10%
10% 7%
Colorado Blvd
12%
50%
50%
50%
50% 7%
71%
22%
Colorado Pl
94% 6%
Michillinda
Baldwin / Oxford
Huntington
72%
17% 8%
73%
19%
50%
50%
18%
82%
11%
10%
90%
50%
50%
100%
74%
26%
Colorado Pl
100% 8%
92%
10%
90%
10% 5% 6%
23%
46% 2%
73%
90%
75%
10%
10%
10%
35%
Colorado
77%
54%
98%
17% 5%
19%
90%
90%
90%
65% 5% 5% 5% 3%
90%
100%
100%
10%
95%
68%
90%
60%
91%
10%
90%
27% 5%
37% 6%
Huntington
90%
10%
50%
50% 7%
37%
56%
17%
64%
19% 8%
10%
51%
41%
90%
50%
33%
50%
50%
Santa Clara
Santa Anita
1st
2nd
Gateway
5th
I210E
I210W

8. Reasons for finetuning split ratios with Engineering judgment
[ Input for CTM Forward Simulation of Arterial Loop+Racetrack No-Incident Model ]
Reasons for finetuning split ratios with Engineering judgment
@ 1st both Northern and Southern approaches: change split ratio from [ left=10%, thru=90%, right=0% ] to [50%, 50%, 0%]
Side street is small; it is therefore unlikely that many drivers go thru; instead, they most likely turn towards the main street
@ 5th N/S: ditto

9. Reasons for finetuning split ratios with Engineering judgment
[ Input for CTM Forward Simulation of Arterial Loop+Racetrack No-Incident Model ]
Reasons for finetuning split ratios with Engineering judgment
@ Santa Anita South: change from [ left=17%, thru=64%, right=19% ] to [17%, 50%, 33%]
Too little traffic predicted on Huntington at 2nd W
Further hypothesis: After racetrack event, people leave via Santa Anita (among others) and want to go to the freeway. The split ratio estimated by the racetrack study are therefore overestimated in the thru direction.

10. Reasons for finetuning split ratios with Engineering judgment
[ Input for CTM Forward Simulation of Arterial Loop+Racetrack No-Incident Model ]
Reasons for finetuning split ratios with Engineering judgment
@ Huntington & Colorado East: change from [ left=77%, right=23% ] to [ 54%, 46% ]
Too little traffic predicted on Colorado & Colorado SE
Analysis of loop data at advance and left-turn locations showed that splitratios differ a lot between loops and racetrack study

11. Reasons for finetuning split ratios with Engineering judgment
[ Input for CTM Forward Simulation of Arterial Loop+Racetrack No-Incident Model ]
Reasons for finetuning split ratios with Engineering judgment
@ 2nd West: change from [ left=3%, thru=91%, right=6% ] to [3%, 60%, 37%] :
Too much traffic predicted on Huntington at I210E and I210W
Further hypothesis: After racetrack event, people leave via Huntington (among others) and want to go to the freeway. The split ratio estimated by the racetrack study are therefore overestimated in the thru direction.

12. [ Input for CTM Forward Simulation of Arterial Loop+Racetrack No-Incident Model ]
SR/BF Calibration
Check whether side street BF reasonable
Apply engineering judgment SR
Set SR
BF Optimization
Generate BF profile
Racetrack SR 2 SR BF
loop SR 4
Default SR
Flow
Time
Target flows
loop flow 3
Racetrack flow
where loop is not available at the same location 1b
Scale racetrack flow
Generate
Generic flow profile
Racetrack flow
where loop is available
at the same location 1a
loop flow 3
Default BF

13. Process to Generate Boundary Flows Overview
[ Input for CTM Forward Simulation of Arterial Loop+Racetrack No-Incident Model ]
Process to Generate Boundary Flows Overview
Generate static (constant in time) boundary flows using an optimization procedure
Generate generic flow profile based on loop data
Scale generic flow profile of step 2 by optimal values obtained in step 1

14. Process to Generate Boundary Flows Step 1
[ Input for CTM Forward Simulation of Arterial Loop+Racetrack No-Incident Model ]
Process to Generate Boundary Flows Step 1
Generate static boundary flows using an optimization procedure
Decision variables: boundary flows
Data:
Split ratios as specified previously
Target flows to match
Measured flows from loop (calibration dataset), where available
Flows from racetrack study, then scaled down by 6% or down by 35%
Default values, when loop and racetrack data unavailable
Objective: minimize RMSE between simulated and target flows
weight of measurement
Set of locations where target flows are available

15. Process to Generate Boundary Flows Step 1
[ Input for CTM Forward Simulation of Arterial Loop+Racetrack No-Incident Model ]
Process to Generate Boundary Flows Step 1
Choice of weight values wij in objective function
Based on data credibility
Loops 2014: are considered very credible  w = 1
Racetrack 2006: used if loop not available  w = 1
Default values: not very credible, used to nudge system towards realistic values  w = 0.2
weight of measurement
Set of locations where target flows are available

16. [ Input for CTM Forward Simulation of Arterial Loop+Racetrack No-Incident Model ]
Color coding:
I210W
Target Flows used in Boundary Flow Optimization Program to Minimize Error to Simulated Flow
Hourly flow from loop
(calibration dataset)
Hourly flow from racetrack,
scaled down by 35%
812
Hourly flow from racetrack,
scaled down by 6%
517
Default values
1297
Colorado Blvd
1087
210
553
Colorado Blvd
200
200
Colorado Pl
Michillinda
Baldwin / Oxford
477
Huntington
Colorado Pl
670
200
223
200
200
230
519
Colorado
322
803
1193
792
708
766
940
1065
1030
736
1035
1270
814
1016
415
100
370
200
Huntington
Santa Clara
Santa Anita
1st
2nd
Gateway
5th
I210E
I210W

17. [ Input for CTM Forward Simulation of Arterial Loop+Racetrack No-Incident Model ]
823
Optimal Static Boundary Flows that minimize error between simulated and target flows
I210W
Baldwin / Oxford
1347
Colorado Blvd
Colorado Blvd
127
567 78
129
Colorado Pl
Michillinda
Huntington
709
220 89
150
518
Colorado Pl
332
173
228
Colorado
1028
200
Huntington
1059
487
364
166
Santa Clara
Santa Anita
1st
2nd
Gateway
5th
I210E
I210W

18. Process to Generate Boundary Flows Steps 2 and 3
[ Input for CTM Forward Simulation of Arterial Loop+Racetrack No-Incident Model ]
Process to Generate Boundary Flows Steps 2 and 3
Boundary Flow Profile used in Forward Simulation
Step 2: Create generic flow profile
Step 3: Scale generic flow profile of step 2 by optimal values obtained in step 1
Flow
220
16:00
20:00
Time

19. Signal Timings 11 intersections: use given plans
[ Input for CTM Forward Simulation of Arterial Loop+Racetrack No-Incident Model ]
Signal Timings
11 intersections: use given plans
2 intersection: invented reasonable plans
Colorado Pl
Michillinda
Baldwin / Oxford
Santa Clara
2nd
5th
1st
Gateway
I210E
I210W
Santa Anita

20. Outline Arterial Loop+Racetrack No-Incident Model
Model components (with calibration set)
Model run
Model validation (with validation set)

21. Westbound traffic on Huntington and Colorado
[ Output from CTM Forward Simulation of Arterial Loop+Racetrack No-Incident Model ]
Westbound traffic on Huntington and Colorado

22. Eastbound traffic on Huntington and Colorado
[ Output from CTM Forward Simulation of Arterial Loop+Racetrack No-Incident Model ]
Eastbound traffic on Huntington and Colorado
TODO: Scale the verticle axis properly

23. Outline Arterial Loop+Racetrack No-Incident Model
Model components (with calibration set)
Model run
Model validation (with validation set)

24. Data Processing
Model Components: FD / BF / SR / Signals
CTM Forward Simulation
Metrics: TT, Delay, LOS, VHT, VMT
Comparison: Travel Time, Flow
Comparison
Data Processing
Calibration Data Set
Validation Data Set
The Model
Bluetooth Travel time 5
Loop flow 3
Comparison: Travel Time, Flow
March: 3, 11, 13, 17, 18, 19, 25, 26, 31
April: 3, 14, 16, 23, 28, 29
May: 5, 6, 8, 12, 13, 15
Calibration Days
Validation Days

25. Simulated vs. Measured Flows
[ Validation of CTM Forward Simulation of Arterial Loop+Racetrack No-Incident Model ]
Simulated vs. Measured Flows
Based on calibration dataset
Based on validation dataset
Huntington

26. Simulated vs. Measured Flows, 16:00-17:00
[ Validation of CTM Forward Simulation of Arterial Loop+Racetrack No-Incident Model ]
Simulated vs. Measured Flows, 16:00-17:00
Individual link flows
Passed cases
Targets
Flow within 100 vph for link flows < 700 vph
4/5 = 80%
> 85%
Flow within 15% for 700 vph < link flows < 2700 vph
6/9 = 67%
Flow within 400 vph for link flows > 2700 vph
0/0
GEH statistics < 5
10/14 = 71%
Sum of all link flows
Results
Targets
Relative Error in Total Flow
3.7%
< 5%
GEH
3.1
< 4

27. Simulated vs. Measured Flows, 17:00-18:00
[ Validation of CTM Forward Simulation of Arterial Loop+Racetrack No-Incident Model ]
Simulated vs. Measured Flows, 17:00-18:00
Individual link flows
Passed cases
Targets
Flow within 100 vph for link flows < 700 vph
4/5 = 80%
> 85%
Flow within 15% for 700 vph < link flows < 2700 vph
6/9 = 67%
Flow within 400 vph for link flows > 2700 vph
0/0
GEH statistics < 5
10/14 = 71%
Sum of all link flows
Results
Targets
Relative Error in Total Flow
2.9%
< 5%
GEH
3.1
< 4

28. Simulated vs. Measured Flows, 18:00-19:00
[ Validation of CTM Forward Simulation of Arterial Loop+Racetrack No-Incident Model ]
Simulated vs. Measured Flows, 18:00-19:00
Individual link flows
Passed cases
Targets
Flow within 100 vph for link flows < 700 vph
10/10 = 100%
> 85%
Flow within 15% for 700 vph < link flows < 2700 vph
3/4 = 75%
Flow within 400 vph for link flows > 2700 vph
0/0
GEH statistics < 5
13/14 = 93%
Sum of all link flows
Results
Targets
Relative Error in Total Flow
5.0%
< 5%
GEH
4.7
< 4

29. Simulated vs. Measured Flows, 19:00-20:00
[ Validation of CTM Forward Simulation of Arterial Loop+Racetrack No-Incident Model ]
Simulated vs. Measured Flows, 19:00-20:00
Individual link flows
Passed cases
Targets
Flow within 100 vph for link flows < 700 vph
8/12 = 67%
> 85%
Flow within 15% for 700 vph < link flows < 2700 vph
2/2 = 100%
Flow within 400 vph for link flows > 2700 vph
0/0
GEH statistics < 5
10/14 = 71%
Sum of all link flows
Results
Targets
Relative Error in Total Flow
22.4%
< 5%
GEH
17.0
< 4

30. Simulated vs measured travel times
[ Validation of CTM Forward Simulation of Arterial Loop+Racetrack No-Incident Model ]
Simulated vs measured travel times
Simulated WB travel time from Gateway to Santa Clara averages 147s between 4pm – 6pm
Measured WB Bluetooth travel time (validation dataset) averages 153s between 4pm – 6pm

31. Simulated vs measured travel times
[ Validation of CTM Forward Simulation of Arterial Loop+Racetrack No-Incident Model ]
Simulated vs measured travel times
16:00-17:00
Journey time within network
Passed cases
Targets
Within 15% or 1 minute, whichever criteria is higher
1/1 = 100%
> 85%
17:00-18:00
Journey time within network
Passed cases
Targets
Within 15% or 1 minute, whichever criteria is higher
1/1 = 100%
> 85%