Modelling for Estimation of Demand to Generate Feeder Bus Routes
of Mass Rapid Transit System
Prof B R Marwah, Non-member
Dr R Parti, Non-member
With the introduction of a high capacity rapid transit system in a metropolitan city, the commuters from the existing
public transport modes are generally attracted to the mass rapid transit system (MRTS) and the major portion of
demand is from the existing bus transport system. In addition, the implementation of such a fast efficient transport
system requires the feeder bus services for effective integration of the two modes. Planning of feeder bus transit system
to integrate with mass rapid transit system (MRTS) necessitates the estimation of demand which will shift to the MRTS
from the existing bus system. Delineation of influence area for MRTS stations and station loads. The present paper deals
with the mode choice analysis to estimate the proportion of demand shifted from the existing bus system to the MRTS
based on the logit model and is effectively implemented to New Delhi, the capital city of India.
Keywords: Mass rapid transit system; Feeder routes; Demand; Integration; Logit model
INTRODUCTION the mode choice analysis. To achieve this objective the study
Feeder bus routes to high capacity Rail Transit Network play methodology has two important stages, namely,
an important role in ensuring an integrated multi-model l Study of the existing road network, existing bus
public transport operation. Implementation of such a system transit system and proposed MRTS network.
will force the restructuring of existing bus network, and l Mode-choice analysis to estimate the passenger demand
provision of feeder bus services for effective integration of the matrix for combined MRTS and feeder network.
two modes. Planning of feeder bus transit system= a 0 + a1T _ time Bus + a 2T _
with mass rapid transit system (MRTS) requires estimation of Mode Choice Analysis
demand, which will shift to the MRTS from the existing bus In a metropolitan city, the existing modes of public
system, delineation of influence area for MRTS stations and transportation, before the introduction of MRTS, are
station loads. Integration of high capacity transit system with generally in different forms of buses and intermediate public
feeder public transport system therefore, needs a series a transport (IPT). With the introduction of a new public
heuristic optimization models1. transport mode, some share of trip makers will shift from the
The public transport system of Delhi2 is primarily road based existing bus system to the new public transport mode.
with a mix of public and private buses. At present the regular Similarly, some portion of demand from private modes may
bus system (stage carriage) comprises 6150 buses that carry also shift to the new system. However, the major component
about eight million commuters. of the demand coming to MRTS will be one transferred from
the existing bus transit system.
Besides about 5000 chartered buses provide point-to-point
service during peak hours supplementing the regular bus Logit model from the family of mode choice models has been
services. Public transport network is being improved and considered to split the travel demand between the existing bus
expanded by including a MRTS for a route length of 33 km in transit system and the newly introduced MRTS. The utility
the first phase. The commissioning of MRTS corridors started (or disutility) function is typically expressed as the linear
from 2002 and all the sections are likely to be operational by weighted sum of the independent variables or their
the year 2005. transformation. For sharing the inter stop travel demand
between MRTS and bus transit network, utility measure of the
STUDY METHODOLOGY two public transport modes is to be calculated. Measure of
The basic objective is to estimate the travel demand matrix for utility, a function of travel time, travel cost, comfort, transfer
the combined MRTS network and feeder bus services based on penalty etc, may be expressed as:
UCBus = a 0 + a1T _ time Bus + a 2T _ costBus
Prof B R Marwah is with the Department of Civil Engineering, IIT,
Kanpur 208 016; and Dr R Parti is with the Department of Civil + a 3C _ fort Bus + a 4 trans _ plt Bus (1)
Engineering, National Institute of Technology, Hamirpur, HP 177 005.
UCMRTS = b 0 + b1T _ time MRTS + b 2T _ cost MRTS
This paper is received on June 21, 2004. Written discussion will be received till
January 31, 2005. + b 3C _ fort MRTS + b 4 trans _ plt MRTS (2)
Vol 85, November 2004 169
where UCBus is utility measure of bus transit network; transfers are considered while planning for bus routes. For
UCMRTS is utility measure of MRTS; a0, a1, a2, a3, a4 are utility determining the travel time between any node pair (i, j ) on an
coefficients of bus transit network; b0, b1, b2, b3, b4 are utility existing bus route network, the inter-stop travel time is
coefficients of MRTS; T_timeBus is travel time of bus transit estimated through (i) no transfer cases, (ii) one-transfer routes,
network; T_timeMRTS is travel time of MRTS; T_costBus is and (iii) two transfer routes.
travel cost of bus transit network; T_costMRTS is travel cost of Estimation of Inter-stop Travel Time through MRTS
MRTS; C_fortBus is comfort level of bus transit network;
C_fortMRTS is comfort level of MRTS; Trans_pltBus is transfer When the MRTS is newly introduced, the trip makers may
penalty of bus transit network; and Trans_pltMRTS is transfer not be able to clearly judge the nearest possible MRTS stations
penalty of MRTS. in the vicinity of their origin or destination. The approach to
determine the best path through MRTS, as shown in Figure 1, is:
The parameters of travel time and travel cost for MRTS
systems would also include travel time and travel cost of (i) For each stop, determine five closely located MRTS
feeder system. The data can be used to calibrate the stations, those with shortest distance from the stop.
coefficients of the parameters. Once these utility measures are (ii) For a bus stop pair (i, j ) let there be MRTS stations
calculated then proportion of demand on each mode is (mt1, mt2, mt3, mt4, mt5 ) in vicinity of stop i and let
calculated using logit model. there be MRTS stations (nt1, nt2, nt3, nt4, nt5 ) in
vicinity of stop j.
e −UC Bus
PBus = (3) (iii) To travel through the combination of MRTS and road
e −UC Bus + e −UC MRTS
between stops i and j, a total of 25 alternative paths are
now available. Total travel times are estimated for
e −UC MRTS
PMRTS = (4) each of the 25 alternatives and one with least time is
e −UC Bus + e −UC MRTS selected.
where PBus, PMRTS are the proportion of demand of the bus Total travel time through mt1, nt1 = TT_road (i, mt1 ) +
transit system and MRTS, respectively. TT_MRTS (mt1, nt1 ) + TT_road (nt1, j ) + transfer time at
The utility measures of bus transit system and MRTS include station mt1 + transfer time at station nt1 (5).
the parameters of travel time, travel cost and transfer penalty,
where TT_road(i, mt1 ) is travel time along the road from stop i
which need to be determined. Firstly, the various parameters
to MRTS station mt1 ; TT_road( j, nt1 ), travel time along the
with respect to MRTS are to be calibrated. The mode choice
road from stop j to MRTS station nt1 ; and TT_MRTS (mt1,
analysis determines the proportion of demand between the
nt1 ), travel time along MRTS between stations mt1 and nt1.
two modes for each origin and destination node pair. It is
therefore, desirable to calculate the travel time and travel cost Minimum of the Total Travel Time for 25 alternative paths
for each node pair for both the public transport modes. establishes the best path to travel through MRTS.
Estimation of Inter-stop Travel Time by Existing Bus Routes
Accessibility of the various inter-node transfer with respect to
The existing bus transit system in the city may have a large the MRTS are determined through two incidence matrices
number of bus routes serving the stops. A trip maker may able [con_mrts(i, j )] and [sec_mrts (i, j )]. If inter-nodal transfer
to perform a trip between an origin and destination either between i and j use MRTS system then [con_mrts (i, j )] depicts
through a directly connected route or by a series of bus routes the connectivity between the node i and MRTS station and
with some transfers4. As it may not be feasible to provide [sec_mrts (i, j )] gives the connectivity between the second
direct bus routes between each O-D pair, normally up to two MRTS station and node j.
mt1 mt2 mt3 mt4 mt 5
MRTS corridor nt 1 nt2 nt3 nt 4 nt 5
Figure 1 Estimation of inter-stop travel time
170 IE (I) JournalCV
Road network MRTS network Demand of matrix for
characteristics characteristics public transport
Existing bus routes
Pessimistic approach Optimistic approach characteristics
Inter-stop shortest travel Inter-stop travel time through
time through directly existing bus routes by
connected bus routes Directly connected bus routes
Identification of shortest time for inter-stop transfers through road
network or combined MRTS and road network
Availability of feeder bus Availability of feeder bus
routes at one end of MRTS routes at both ends of MRTS
Fare for: Fare for: Fare for: Fare for:
MRTS and feeder MRTS and feeder MRTS and bus Unified MRTS
services services system and feeder
—unit rate —slab rate —unit rate services
Bus system Bus system Feeder rate
—unit rate —slab rate services Bus system
—unit rate —unit rate
Demand matrix for
For an inter-stop transfer between MRTS
node pair (i, j), obtain
Demand matrix for
Figure 2 Model for mode choice analysis
Application of Mode Choice Analysis Model (iv) Availability of feeder routes on one end or both ends.
Mode choice analysis model, as shown in Figure 2, is evolved Considering the inter-nodal transfers between existing bus
to assess the number of commuters who may shift from the transit system and MRTS/feeder bus service, the utility
existing bus system to MRTS-feeder bus combination based on measures for each of the two modes is estimated. The total
the parameters of travel time, travel cost and modal travel time by the existing bus system and by the best possible
characteristics of the two modes. A commuter may opt to path from the 25 different combinations of MRTS stations and
travel through MRTS subject to the conditions: O-D nodes is estimated. Accessibility of the inter-nodal
transfers with respect to the MRTS is established and
(i) Maximum distance from the origin or destination to accessibility materials are calculated.
the MRTS station from where demand can be
Optimistic approach deals with limited existing bus transit
attracted and feeder routes are to be generated should
network that is already available whereas pessimistic approach
be within a certain limit.
deals with unlimited bus transit network. The model
(ii) The distance travelled on MRTS corridor should facilitates eight options for both the approaches based on the
be at-least some proportion of the total distance combinations of bus fare system, MRTS and feeder bus
travelled between the origin and destination nodes. system, availability of feeder bus service either on one end of
(iii) The distance travelled on the MRTS corridor is MRTS or on both ends. Table 1 presents the options available
greater than a certain minimum distance. in the model for the analysis of mode choice.
Vol 85, November 2004 171
Table 1 Options available for mode choice analysis The mode choice analysis between any O-D pair involves the
Option Existing/ MRTS and feeder Availability
number unlimited bus system of feeder (i) In vehicle travel time from origin to destination.
bus system MRTS fare Feeder bus service
(ii) Transfer time at MRTS stations.
fare bus fare
(iii) Waiting time on the bus and MRTS system.
1 Unit Rate Unit Rate Unit Rate One end
(iv) Travel cost through bus and MRTS/feeder system.
2 Unit Rate Unit Rate Unit Rate Both ends
As the mode choice model assigns the demand between
3 Slab system Slab system Slab system One end MRTS/feeder system and bus system, the value of different
parameters assume considerable importance. To test the
4 Slab system Slab system Slab system Both ends
sensitivity of the model with respect to different parameters
5 Slab system Inter MRTS- Slab system One end and decision thresholds, the parameters, which are of
station fare considerable importance in mode choice, that need to be
matrix studied under different scenarios are:
6 Slab system Inter MRTS- Slab system Both ends l Policy time headway for feeder bus routes.
station fare l Maximum distance from MRTS station from where
matrix demand is attracted to MRTS.
7 Slab system Unified MRTS One end l Minimum ratio of MRTS travel distance to total
and feeder bus travel distance between origin and destination.
l Availability of feeder service at one end or both ends.
matrix) l Type of fare structure slab system or unit rate.
8 Slab system Unified MRTS Both ends Table 2 gives the values of different parameters adopted for
and feeder bus sensitivity analysis. In this table, each of the parameters at
fare (Inter-stop serial 3, 5 and 7 have three values; those at serial 8 and 9 have
unified fare two values, while others have only one value. A full factorial
matrix) design for these values will result in 108 combinations. Mode
Considering these operational constraints, the mode choice Table 2 Value of parameters for mode choice analysis
analysis estimates the share of demand between the bus transit
network and MRTS/feeder bus service for each O-D pair. The Serial Parameters Values
proportionate share of inter-nodal demand through MRTS is Number
determined by the formulated Logit model and matrix 1 Transfer time from feeder
[share_mrts (i, j )] is generated. If there is a feasible path to MRTS, s 300
through MRTS between i and j, then share_mrts (i, j ) lies
2 Transfer time from MRTS
between 0 to 1, otherwise share_mrts (i, j ) = 0.
to Feeder, s 300
Application of Mode Choice Analysis for New Delhi 3 Time headway for feeder
The demand is estimated for generation of feeder route plan in bus routes, s 300, 600, 900
accordance with the data available from Rail India Technical 4 Time headway for MRTS
and Economical Services. The daily demand matrix for inter- routes, s 300
stop demand of (1542 × 1542) size represents total demand of
7.67 million passengers for the year 2000. For the 5 Maximum distance (in km) from
metropolitan city of Delhi, the share of public and private MRTS station from where demand
transport mode data is available and 62% of the travellers is attracted on to MRTS 8, 10, 12
move through public transport. In this paper, the study is 6 Minimum distance travelled on
restricted to only two parameters of travel time and travel MRTS, m 2000
cost, because the realistic data is available only for these two
7 Ratio of MRTS travel distance to
parameters. Analysis was done with the estimated coefficients
total travel distance between origin
for different trip lengths and trip cost between public and
and destination 0.2, 0.3, 0.4
private mode. Knowing the number of travellers for different
trip lengths, the total share of public transport system is 8 Availability of feeder bus service One end, both ends
estimated. This analysis helped to calibrate the coefficients of
9 Fare structure Slab rate, unit rate
travel time and travel cost.
172 IE (I) JournalCV
choice analysis is carried out for these 108 cases to study the from 12 km to 8 km and increasing ratio of MRTS travel
behaviour of the model under different scenarios. distance to total distance between O-D pairs from 0.2 to 0.4,
the daily MRTS rider-ship will decrease from 1.82 million to
The fare structure considered for sensitivity analysis is slab
1.24 million, a drop of 31.87%. For the cases, when the feeder
system and unit rate/km.
service is available only at one end, the rider-ship is estimated
Application of the mode choice model for a scenario estimates to range between 1.65 million and 1.16 million.
the expected MRTS rider-ship for each O-D pair and the total
Estimated MRTS demands for different scenarios indicate that
MRTS demand matrix (1542 × 1542) is generated. These
the formulated mode choice appears to give realistic results.
matrices are generated for all the 108 cases of experimental
The estimated MRTS demand matrix obtained from the mode
design. Study of these results indicate that
choice model is to be used for planning of the feeder bus
l MRTS rider-ship decreases as the time headway for network.
the feeder system increases from 300 s to 900 s. This is
because higher time headway increases the travel time SUMMARY
making MRTS less attractive. This paper deals with mode choice analysis to estimate the
l MRTS rider-ship increases as the influence area for proportion of demand shifted from the existing bus system to
MRTS increases from 8 km to 12 km. If feeder service the MRTS based on the logit model and is effectively
is provided for longer distance, it will have more implemented to New Delhi, the capital city to India. The
attraction to MRTS. coefficients of parameters for travel time and travel cost is
l As the minimum travel distance over MRTS, estimated based on the modal split for public transport in
expressed as ratio of the total trip length, increases Delhi. To test the sensitivity of the mode choice model with
from 0.2 to 0.4, the MRTS rider-ship decreases. respect to different parameters and decision thresholds, the
various parameters, which are of considerable importance, are
l When the feeder service is available only at one end of studied under different scenarios.
the trip, MRTS rider-ship will be less as compared to
when service is available at both ends. REFERENCES
These trends are similar for both types of fare structure. In case
1. P Raman. Bus Transit Planning for a Large City and Decision Support
of slab fare structure, the highest daily MRTS rider-ship of
System of Feeder Bus Routes for Rail Transit Network. Ph D Thesis
1.82 million is estimated when
(unpublished work), IIT, Kanpur, September 2002.
l Feeder routes have time headway of 300 s.
2. RITES. Planning of Feeder Public Transport System for Phase-I of Delhi
l Maximum distance from MRTS station to stop within
MRTS. New Delhi, India, 2001.
which demand can be attracted is 12 km.
l MRTS travel distance is at least 0.2 times the total trip 3. C S Papacostas. Fundamentals of Transportation Engineering. Prentice-Hall
length between O-D pairs. of India Private Limited, New Delhi, 1990.
l Feeder routes are available at both ends. 4. M H Bajj and H S Mahamassani. TRUST: a LISP Program for Analysis of
With increase of time headway for feeder routes from 300 s to Transit Route Configurations. Transportation Research Record 1283,
900 s, decreasing maximum distance of MRTS station to steps Transportation Research Board, Washington, DC 1990, pp 125-135.
Vol 85, November 2004 173