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:
In order to reasonably satisfy our targeted performance of Fast-Trips across these categories, we will utilize calibration techniques such as:
Run on: June 30, 2016
Run by: Lisa at MTC
Network:
Demand: OBS v0.1
Configs: Pathweights
Outputs
Vehicle Loads
Observed vs. Modeled Paths
Validation Summary Dashboard
No drive access or egress for now.
Memory issues
Addressed memory issues.
Run on: August 16, 2016
Run by: Lisa at MTC
Network: v1.4
Demand: v1.0
Configs: Pathweights
Walk Access/Egress to/from Rail Xfer links uni-directional rather than bi-directional Demand input error: person IDs not unique
3072 - No walk egress from teh N line to TAZ475 Person IDs 35, 377, 517
Run on: October 7, 2016
Run by: Lisa at MTC
Network:
Demand: v1.1
Configs: ..
Outputs
Observed vs. Modeled Paths
Validation Summary Dashboard
Get memory overflow with too many trips.
Run on: July 19, 2016
Run by: Lisa at MTC
Network:
Demand: CHTS v0.1
Configs: ..
Outputs
Observed vs. Modeled Paths
Validation Summary Dashboard
No BART Boards.
Run on: October 7, 2016
Run by: Lisa at MTC
Network: v1.9
Demand: CHTS v0.2
Configs: ..
Outputs
Observed vs. Modeled Paths
Validation Summary Dashboard
Run on: December 16th, 2016
Run by: Lisa at MTC
Network:
Demand: CHTS v0.3
Configs: ..
Outputs
Observed vs. Modeled Paths
Validation Summary Dashboard
Run on: Feb 6, 2017
Run by: Bhargava at SFCTA
Network: SFCTA v1.10
Demand: v0.4
Configs: ..
Outputs
Observed vs. Modeled Paths
Validation Summary Dashboard
Run on: March 21, 2017
Run by: Bhargava at SFCTA
Network: v1.10
Demand: CHTS v0.4
Configs: ..
Outputs
Observed vs. Modeled Paths
Validation Summary Dashboard
Routes where they transfer between different trips on same route
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 (???))Run on 2017-03-22
Run by Bhargava Sana at SFCTA
Network: v1.11
Access Egress Link Viewer**
Network Explorer**
Demand: CHTS v0.4
Parameters: ..
Configs: ..
Network:
Parameters:
Other:
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
Run on 2017-03-22
Run by Bhargava Sana at SFCTA
Network: v1.11
Access Egress Link Viewer**
Network Explorer**
Demand: CHTS v0.4
Parameters: ..
Configs: ..
Network:
Parameters:
Other:
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
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.
The final tier of calibration will assess the performance of the Fast-Trips parameters within a full SF-CHAMP model run.