huan li and robert l. bertini transportation research board 88th annual meeting washington, dc
DESCRIPTION
Assessment Of An Optimal Bus Stop Spacing Model Using High Resolution Archived Stop-level Data. Huan Li and Robert L. Bertini Transportation Research Board 88th Annual Meeting Washington, DC January 11-15, 2009. About TriMet. Serves 1.2 M population 63.9 M annual bus trips 95 bus routes - PowerPoint PPT PresentationTRANSCRIPT
TRB 88th Annual Meeting, Washington DC January, 2009
Huan Li and Robert L. Bertini
Transportation Research Board 88th Annual Meeting Washington, DCJanuary 11-15, 2009
Assessment Of An Optimal Bus Stop Spacing Model Using High Resolution Archived Stop-level Data
TRB 88th Annual Meeting, Washington DC January, 2009
About TriMet Serves 1.2 M
population 63.9 M annual bus
trips 95 bus routes 655 buses 8100 bus stops Also LRT, Commuter
Rail, Streetcar & Paratransit
TRB 88th Annual Meeting, Washington DC January, 2009
TriMet’s Bus Dispatch System
On-BoardComputer
Radio
DoorsLift
APC (Automatic Passenger Counter)Overhead Signs
OdometerSignal Priority Emitters
Stop Annunciation
Memory Card
RadioSystem
Garage PC’s
Radio AntennaGPS Antenna
Navstar GPS Satellites
Control Head
TRB 88th Annual Meeting, Washington DC January, 2009
TriMet’s Bus Dispatch System
Schedule deviation
Control Head
PCMIA Card
Infrared APC
Operator Input
Dispatching Arrival Prediction
TRB 88th Annual Meeting, Washington DC January, 2009
TriMet’s Bus Dispatch System
TRB 88th Annual Meeting, Washington DC January, 2009
One Year Stop-Level Data (2007)
Rou
te N
o.Se
rvic
e D
ate
Leav
e Ti
me
Stop
Tim
e
Arr
ive
Tim
e
Bad
geD
irect
ion
Trip
No.
Loca
tion
IDD
wel
lD
oor
Lift
Ons
Offs
Est.
Load
Max
Spe
edPa
ttern
D
ista
nce
X C
oor.
Y C
oor.
14 01NOV2001 8:53:32 8:49:15 8:53:28 285 0 1120 4964 0 0 0 0 0 21 41 10558.58 7644468 67600514 01NOV2001 8:55:00 8:51:41 8:54:46 285 0 1120 4701 4 0 0 0 1 20 50 15215.05 7649112 67632814 01NOV2001 8:56:22 8:52:00 8:55:08 285 0 1120 4537 36 3 0 6 0 26 34 15792.35 7649674 676220
Route Number Vehicle Number Service Date Actual Leave Time Scheduled Stop Time Actual Arrive Time Operator ID Direction Trip Number Bus Stop Location
Dwell Time Door Opened Lift Usage Ons & Offs (APCs) Passenger Load Maximum Speed on
Previous Link Distance Longitude Latitude
TRB 88th Annual Meeting, Washington DC January, 2009
2007 AM Boardings
TRB 88th Annual Meeting, Washington DC January, 2009
Background on Stop Location• Challenges in delivering reliable and timely bus
service • Financial constraints• Public transit operational issues• Transit service generally favors bus stop
accessibility• Sometimes based on past history and tradition
rather than rigorous ongoing analysis at the stop level
TRB 88th Annual Meeting, Washington DC January, 2009
Stop Spacing Service Standards
TriMet • Portland >80 units/acre: 400-600 ft 22-80 units/acre: 500-750 ft 4-22 units/acre: 600-1000 ft <4 units/acre: as needed Inner Portland has 200 ft blocks (264 ft street spacing) Route 19 mean stop spacing is 942 ft (3 blocks)
Objective: Develop and test a simple stop spacing model using this rich data
TRB 88th Annual Meeting, Washington DC January, 2009
Concept DerivationTrade off: person’s time in parallel access vs. another person’s time in riding.
•Minimize access cost: favors small s•Minimize riding cost: favors large s
TRB 88th Annual Meeting, Washington DC January, 2009
Assumptions• Origins & destinations distributed along route
in one dimension (ignore perpendicular access)…
• Average access distance (parallel only) =s/4
• Assume number of passengers boarding or alighting at a stop to be ~Poisson distributed
TRB 88th Annual Meeting, Washington DC January, 2009
Access Cost
Riding and Stopping Cost
TRB 88th Annual Meeting, Washington DC January, 2009
Access Cost Value of Passenger Travel Distance
p = density of trip origins plus density of trip destinations for passengers who board or alight the same vehicle (units: number/distance)
s/4= average access distance (unit: distance) ν = passenger access speed (unit: distance/time)• a = average cost per unit time per person for access
(unit: cost/time)
ps2a
4vin interval of length s
Ca= [avg. no. of pax] [avg. dist traveled] [cost/unit dist]
4s ps
va
TRB 88th Annual Meeting, Washington DC January, 2009
Access Cost
Riding and Stopping Cost
TRB 88th Annual Meeting, Washington DC January, 2009
Riding and Stopping CostValue of in-vehicle passenger lost time due to boardings and alightings
• N = expected number of passengers on vehicle• V = vehicle cruise speed• = time lost in stopping to serve passengers• Pr =1-e-ps = probability that vehicle actually stops (from
Poisson for number of ons and offs)• γr= average cost per unit time per person for riding
N= in interval length s
Cr= [avg. no. of pax] [riding time + lost time][cost/unit time]
sV+Pr
r
= NsrV
+Nr(1 e ps )
TRB 88th Annual Meeting, Washington DC January, 2009
[ ] / s
Average Cost Per Unit Length
Given that
[access]
= 4 ravpN
Average cost per unit length
+ [riding] + [stopping]
= ps2a
4v
NsrV
+Nr(1 e ps)
Average cost per unit length =
VNpN r
r
Independent of s !
Choice of s is independent of V, depends solely on
pseps ps )1(
TRB 88th Annual Meeting, Washington DC January, 2009
Objective Function
psepsC
ps )1(0
Coverage for >2
If β > 2:pse ps )1(
ps1
TRB 88th Annual Meeting, Washington DC January, 2009
pspsC 1
0
pN
p
Npv
ps a
r
44
ps = expected number of passengers to board or alight per stop
nsDestinatio and Origins Trip ofDensity Buson Pax No.ofTimeLost 4Spacing Optimal
pspsC 1
0
TRB 88th Annual Meeting, Washington DC January, 2009
Case Study: Inbound Route 19 All Day
• 370 days (2/20/07 - 1/5/08) • 19,344 trips• 33.2 ons and offs/trip:• Average passenger load/stop:
7.9
• Route 19 Glisan to Portland• Route Length: 9.27 mi• Number of stops: 52• Mean delay due to stopping:
33.6 s• Use 4ft/s walking speed
TRB 88th Annual Meeting, Washington DC January, 2009
343
780
1,26
42,
045
807
1,49
955
179
359
996
6 1,01
783
760
581
971
0 758 80
5 859
762
1,02
464
879
61,
096
623
613
638
660
890
861
505
991
420
704
630 76
1 793
1,18
778
256
597
472
6 814
2,76
81,
950
728
1,57
952
82,
108
890
2,47
995
246
0
0
500
1,000
1,500
2,000
2,500
3,000
0.0
0.1
0.2
0.5
0.8
1.0
1.3
1.4
1.5
1.6
1.8
2.0
2.2
2.3
2.4
2.6
2.7
2.9
3.0
3.2
3.4
3.5
3.7
3.9
4.0
4.1
4.2
4.3
4.5
4.7
4.8
5.0
5.0
5.2
5.3
5.4
5.6
5.8
6.0
6.1
6.2
6.4
6.5
7.1
7.4
7.6
7.9
8.0
8.4
8.5
9.0
9.2
9.3
9.5
Distance Along Route Toward Downtown (miles)
Bus
Sto
p Sp
acin
g (ft
)
0
2
4
6
8
10
12
14
Num
ber o
f Pas
seng
ers
Route 19 Inbound Spacing Status
TRB 88th Annual Meeting, Washington DC January, 2009
Optimized Spacing Calculation• No. of passengers on vehicle • Passenger ons and offs• Lost time
Rou
te N
o.
Serv
ice
Dat
e
Leav
e Ti
me
Stop
Tim
e
Arr
ive
Tim
e
Bad
geD
irec
tion
Trip
No.
Loca
tion
IDD
wel
lD
oor
Lift
Ons
Offs
Est.
Load
Max
Spe
edPa
ttern
Dis
tanc
e
X C
oor.
Y C
oor.
14 01NOV2001 8:53:32 8:49:15 8:53:28 285 0 1120 4964 0 0 0 0 0 21 41 10558.58 7644468 67600514 01NOV2001 8:55:00 8:51:41 8:54:46 285 0 1120 4701 4 0 0 0 1 20 50 15215.05 7649112 67632814 01NOV2001 8:56:22 8:52:00 8:55:08 285 0 1120 4537 36 3 0 6 0 26 34 15792.35 7649674 676220
N = 7.9 pax/stopps = 33.2 pax/trip
=33.6sec
TRB 88th Annual Meeting, Washington DC January, 2009
2007 Time Space Passenger Load Plot Route 19
Time (hour)
Dis
tanc
e (m
i)9
8
7
6
5
4
3
2
1
05 10 15 20
30
25
20
15
10
5
0
TRB 88th Annual Meeting, Washington DC January, 2009
Route 19 Inbound Optimized Spacing
0 10 20 30 400
500
1000
1500
2000
2500
3000
3500
4000
N (number of passengers on bus)
Stop
Spa
cing
(ft)
20 ons/offs
60 ons/offs
40 ons/offs
Feb.20th,2007-Jan.5th,2008consider stop circle delay: Stop circle time - free flow timeFor n=790,392, mean = 33.6 sec, stdev = 22.8 sec (range 1-100)Mean ons/offs=33.2Also show spacing plot for 20 ,40 and 60 ons/offsStep function based on 20 blocks/mile
33.2 ons/offs
Optimal
TRB 88th Annual Meeting, Washington DC January, 2009
AM Peak Analysis• Direction: all inbound trips• Analyzed time period: AM
peak hours (6:00-9:00 defined by TriMet) in weekdays
• Number of trips: 3,658 • Mean headway in peak hour:
12 minutes• Mean trip time: 32.6 min
2007 AM Passenger Load Plot Route 19
2007 AM Passenger Ons&Offs Plot Route 19
TRB 88th Annual Meeting, Washington DC January, 2009
AM Peak AnalysisFree Trip Time Acc&Dec Time Passenger On Passenger Off
All Day 22.2 min 17.0 sec 3.0 sec 4.1 sec
AM Peak 23.9 min 18.4 sec 0.8 sec 3.0 sec
Passenger Load Ons and Offs Lost Time Opt Spacing
7.9pax
33.2pax 33.6sec
1222ft
12.8pax
48.7pax
37sec
1258ft
All Day AM Peak
TRB 88th Annual Meeting, Washington DC January, 2009
AM Peak Analysis
0 10 20 30 400
500
1000
1500
2000
2500
3000
3500
4000
N (number of passengers on bus)
Stop
Spa
cing
(ft)
20 ons/offs
60 ons/offs
40 ons/offs
48.7 ons/offs
Feb.20th,2007-Jan.5th,2008 AM Peak (6:00-9:00)consider stop circle delay: Stop circle time - free flow timeFor n=169,702, mean = 37.03 sec, stdev = 22.8 sec (range 1-100)Mean ons/offs=37.6Also show spacing plot for 20 ,40 and 60 ons/offsStep function based on 20 blocks/mile
Optimal
TRB 88th Annual Meeting, Washington DC January, 2009
Conclusions• 12 (14) stops are recommended for consolidation• The trip time would be reduced by 3.4 (4.0) min/trip• The total savings due to consolidation could be up
to 3.7 (4.4) hours of service time per day • Allow the addition of approximately 7.6 (9)
additional trips per weekday • Mean weekday headway would drop from 18.0 min
to 16.1 (15.8) min • Total of 17,076 inbound trips, the time saved would
be 980 (1140) hours during the year • Assuming $60/hr operating cost, about $60,000
($68,000) could be saved by TriMet
TRB 88th Annual Meeting, Washington DC January, 2009
Next Steps• Automate process for all routes• Produce quarterly reports for TriMet• Verify “real” cost savings• Check model assumptions (e.g. Poisson) • Consider “real” relationship to demand and equity• Connect to scheduling
TRB 88th Annual Meeting, Washington DC January, 2009
Acknowledgements• David Crout of TriMet
for providing the rich data set that facilitated this analysis
• Prof. Gordon Newell• Prof. Michael Cassidy,
University of California at Berkeley, for his assistance in developing the analytical framework for this paper [email protected]
www.its.pdx.edu