Tier I: Quality of Service and Routes

The first tier of validation will use demand and chosen route information from the on-board surveys (OBS) and California Household Travel Survey (CHTS) and will evaluate the performance categories that require knowledge of routes including:

• Path-based quality of service,
• Path coverage, and
• Path probabilities.

In order to reasonably satisfy our targeted performance of Fast-Trips across these categories, we will utilize calibration techniques such as:

• Correction to base year networks or demand;
• Tweaking of run parameters such as the delta for the hyperpath;
• Addition or modification of the specification of various factors or their parameters.

OBS Run 1

Run on: June 30, 2016
Run by: Lisa at MTC

Summary of Changes

Inputs

Network:
Demand: OBS v0.1
Configs: Pathweights

Results

Outputs
Vehicle Loads
Observed vs. Modeled Paths
Validation Summary Dashboard

Known Issues

No drive access or egress for now.
Memory issues

Supporting Analysis

Bellwether Paths

Actions Taken

Addressed memory issues.

OBS Run 2

Run on: August 16, 2016
Run by: Lisa at MTC

Summary of Changes

Inputs

Network: v1.4
Demand: v1.0
Configs: Pathweights

Results

Outputs

Known Issues

Walk Access/Egress to/from Rail Xfer links uni-directional rather than bi-directional Demand input error: person IDs not unique

Supporting Analysis

Bellwether Paths

3072 - No walk egress from teh N line to TAZ475 Person IDs 35, 377, 517

Actions Taken

• Made xfers bi-directional

OBS Run 3

Run on: October 7, 2016
Run by: Lisa at MTC

Summary of Changes

Inputs

Network:
Demand: v1.1
Configs: ..

Results

Outputs
Observed vs. Modeled Paths
Validation Summary Dashboard

Known Issues

Get memory overflow with too many trips.

Supporting Analysis

Bellwether Paths

Next Steps

CHTS Run 0

Run on: July 19, 2016
Run by: Lisa at MTC

Summary of Changes

Inputs

Network:
Demand: CHTS v0.1
Configs: ..

Results

Outputs
Observed vs. Modeled Paths
Validation Summary Dashboard

Known Issues

No BART Boards.

Supporting Analysis

Bellwether Paths

CHTS Run 0.1

Run on: October 7, 2016
Run by: Lisa at MTC

Summary of Changes

Inputs

Network: v1.9
Demand: CHTS v0.2
Configs: ..

Results

Outputs
Observed vs. Modeled Paths
Validation Summary Dashboard

Known Issues

Supporting Analysis

Bellwether Paths

CHTS Run 1

Run on: December 16th, 2016
Run by: Lisa at MTC

Summary of Changes

Inputs

Network:
Demand: CHTS v0.3
Configs: ..

Results

Outputs
Observed vs. Modeled Paths
Validation Summary Dashboard

Known Issues

Supporting Analysis

Bellwether Paths

CHTS Run 2

Run on: Feb 6, 2017
Run by: Bhargava at SFCTA

Summary of Changes

Inputs

Network: SFCTA v1.10
Demand: v0.4
Configs: ..

Results

Outputs
Observed vs. Modeled Paths
Validation Summary Dashboard

Known Issues

Supporting Analysis

Bellwether Paths

CHTS Run 3

Run on: March 21, 2017
Run by: Bhargava at SFCTA

Inputs

Network: v1.10
Demand: CHTS v0.4
Configs: ..

Results

Outputs
Observed vs. Modeled Paths
Validation Summary Dashboard

Known Issues

• xfer not onerous enough
• egress penalty too high
• no dwell times
• no KNR
• no PNR lot capacities

Supporting Analysis

Routes where they transfer between different trips on same route

Bellwether Paths

Issue

BART paths for the following should be found; access/egress links appear to exist.

Peninsula -> SF
person_id: 1379781_2 person_trip_id: 6

East Bay <-> SF
1498957_1 person_trip_id: 14
3006352_2 person_trip_id: 7
3007326_2 person_trip_id: 5

7162320_2
Issue: Path that uses bart - but not sure how. CHTS must have coded a phantom BART trip or didn’t identify an intermediate stop.

1407624_2
Issue: High probability paths have a transfer to the 19 Polk rather than just getting on the 47 or 49. The 47 and 49 are found, but given <1% probabilities. This is b/c the 19 has an egress distance of 0.06 whereas 47/49 have egress distance of 0.21. This is a cost difference of ~13 whereas the xfer only costs 5

1474991_2
Issue: Tradeoff between egress distance and transfers. Transfers are too cheap.

3006352_2 person_trip_id: 7
Issue: Tradeoff between egress distance and transfers. Transfers are too cheap.

Issue: Suspicious Caltrain-to-Caltrain transfers

• 2625922_1, 3 (Bb2N to Bb1N)
• 2630962_2, 11 (AB2N to AB1N)
• 2630962_2, 19 (AB2N to AB1N)
• 3016440_1. 7 (AB2N to AB1S (???))

CHTS Run 4

Run on 2017-03-22
Run by Bhargava Sana at SFCTA

Summary of Changes

• Same demand/networks as CHTS Run 3;
• 50% more penalty on xfers

Inputs

Network: v1.11
Access Egress Link Viewer**
Network Explorer**

Demand: CHTS v0.4
Parameters: ..
Configs: ..

Results

Observed vs. Modeled Paths

Validation Summary Dashboard

Box Location

Known Issues

Network:

• No KNR access links
• No PNR station capacities
• Missing access and xfer links

Parameters:

• Transfers still to easy relative to access/egress distance
• Access and Egress parameters are not even; egress far too onerous.

Other:

• Strange switching between Caltrain routes

Supporting Analysis

Path Traces

Bellwether Paths

1710214_1 : FT doesn’t find anything N of Geary, including the observed route downtown via the 1AX

1407624_2

Previous Issue High probability paths have a transfer to the 19 Polk rather than just getting on the 47 or 49. The 47 and 49 are found, but given <1% probabilities. This is b/c the 19 has an egress distance of 0.06 whereas 47/49 have egress distance of 0.21. This is a cost difference of ~13 whereas the xfer only costs 5

Run 4: FT still strongly preferring the strange path with the transfer to the 9 and 19.

1474991_2

Previous Issue: Tradeoff between egress distance and transfers. Transfers are too cheap.

Run 4: Fixed!

3006352_2 person_trip_id: 7

Previous Issue: Tradeoff between egress distance and transfers. Transfers are too cheap.

Run 4: Still there

1710214_1 Run 3: FT doesn’t find anything N of Geary, including the observed route downtown via the 1AX Run 4: same issue

CHTS Run 4

Run on 2017-03-22
Run by Bhargava Sana at SFCTA

Summary of Changes

• Same demand/networks as CHTS Run 3;
• 50% more penalty on xfers

Inputs

Network: v1.11
Access Egress Link Viewer**
Network Explorer**

Demand: CHTS v0.4
Parameters: ..
Configs: ..

Results

Observed vs. Modeled Paths

Validation Summary Dashboard

Box Location

Known Issues

Network:

Parameters:

• No overlap calculation
• Dispersion 1.0

Other:

• Strange switching between Caltrain routes still there.

Supporting Analysis

Path Traces - none yet

Bellwether Paths

1710214_1 : FT doesn’t find anything N of Geary, including the observed route downtown via the 1AX

1407624_2

Run 4: High probability paths have a transfer to the 19 Polk rather than just getting on the 47 or 49. The 47 and 49 are found, but given <1% probabilities. This is b/c the 19 has an egress distance of 0.06 whereas 47/49 have egress distance of 0.21. This is a cost difference of ~13 whereas the xfer only costs 5. FT still strongly preferring the strange path with the transfer to the 9 and 19.

3006352_2 person_trip_id: 7

Run 4: Tradeoff between egress distance and transfers. Transfers are too cheap.

1710214_1 Run 4: FT doesn’t find anything N of Geary, including the observed route downtown via the 1AX

Tier II: Regional Assignment and Computation Feasibility

The second tier of calibration will utilize the full regional demand in order to assess the two remaining performance targets: the assignment as a whole as well as the computational efficiency.

Tier III: Regional Assignment with Feedback

The final tier of calibration will assess the performance of the Fast-Trips parameters within a full SF-CHAMP model run.