Route choice on transit networks with on-line information at stops

Route choice on transit networks with on-line information at stops Guido Gentile Dipartimento di Idraulica Trasporti e Strade Università degli Studi di Roma La Sapienza, guido.gentile@uniroma1.it Sang Nguyen Département d Informatique et de Recherche Opérationnelle Université de Montréal, sang.nguyen@umontreal.ca Stefano Pallottino Dipartimento di Informatica Università di Pisa September 28, 24 Abstract Passengers on a transit network with common lines are often faced with the problem of choosing between either to board the arriving bus or to wait for a faster one. Many assignment models are based on the classical assumption that at a given stop passengers board the first arriving carrier of a certain subset of the available lines, often referred to as the attractive set. In this case, it has been shown that if the headway distributions are exponential then an optimal subset of lines minimizing the passenger travel time can be determined. However, when on-line information on future arrivals of buses are posted at the stop it is unlikely that the above classical assumption holds. Passengers may choose in this case to board a line that offers the best combination of displayed waiting time and expected travel time to destination once boarded. In this paper, we propose a general framework for determining the probability of boarding each line available at a stop when on-line information on bus waiting times is provided to passengers. We will also show that the classical model without on-line information may be interpreted as a particular instance of the proposed framework. The impact of the availability of information regarding bus arrivals and that of the regularity of transit lines on the network loads as well as on the passenger travel times will be illustrated with small numerical examples. Introduction Consider a transit network where every line has a fixed itinerary and is described by a sequence of stops. We are interested in urban networks with common lines, in the sense that multiple line routes share some stops and segments of itineraries. We will focus on the case where the transit lines service irregularity is high enough so that passengers may not be able to time their arrivals at stops in order to minimize their waiting times. It is reasonable in this case to assume that line headways are statistically independent and that passengers arrive randomly at stops (see, for example, Seddon & Day 1974, Jolliffe & Hutchinson 1975, Bowman & Turnquist 1981). Research grants by MIUR and INDAM-GNAMPA of Italy, and NSERC of Canada. Our colleague Stefano Pallottino sadly passed away on April the 11th, 24. 1

Earlier works on transit systems are devoted to the analysis of service regularity and to the waiting and boarding process in situations of congestion. They are mainly aimed at developing realistic bus headway distributions and passenger waiting time distributions, such as the Power and the Erlang distributions (e.g., Chriqui & Robillard 1975, Bowman & Turnquist 1981, Hasseltröm 1981, Larson & Odoni 1981, Gendreau 1984, Marguier & Ceder 1984). Recent researches focus more on the passenger s route choices and on the resulting assignment models. In terms of route choice, it is often assumed that each passenger has a traveling strategy that allows him to reach his destination at a minimum perceived cost. On the modeling network a strategy may be formally described by a hyperpath connecting the origin to the destination (see for instance Nguyen & Pallottino 1989) with the property that at each stop-node only transit lines that have a positive probability to be boarded by the passenger are explicitly considered. Existing assignment models using hyperpaths (e.g., Spiess 1984, Nguyen & Pallottino 1988, Spiess & Florian 1989, Wu, Florian & Marcotte 1994, Bouzaïene-Ayari, Gendreau & Nguyen 1995, Nguyen, Pallottino & Gendreau 1998) are all based on the following hypotheses: transit line headways are statistically independent, and have exponential distributions; passengers arrive randomly at stops, and board the first arriving carrier of their attractive set which is choosen among the available lines. These assumptions lead to a greedy algorithm for determining an optimal subset of the available lines that minimizes the passenger travel time (see Section 2). Simple methods for computing the line boarding probabilities for passengers sharing a given destination have been proposed as well as complete assignment procedures. The new ingredient here is the introduction of on-line information regarding the arrival times of the line carriers at transit stops. Specifically, at some stops it will be assumed that the waiting time for the next carrier of each line is directly displayed on a variable message sign, providing a reliable information that passengers may use locally to decide which line to board. More sophisticated technologies internet, mobile phones, wireless networks, etc. may allow passengers to retrieve the same information remotely from the stop. In general, the availability of prior information allows passengers to consider a variety of options and alternatives not addressed in this paper. For instance, with prior information a passenger at home may decide to change his departure time, the initial stop of his journey and even the transport mode. While a passenger riding on a carrier can choose the next alighting stop to make the most convenient transfer, based on information either displayed in the vehicle or retrieved through a mobile phone. The study of a more general behavioral model is a challenging topic of research beyond the scope of the present work. 1 Common lines with on-line information at stops 1.1 The problem At the core of any transit assignment model, there is either implicitly or explicitly a stop model. This latter model emulates the passenger s decision making at a transit stop. It is generally assumed that the passenger will choose to board a carrier that will minimize his expected total travel time to reach his destination. Given the headway distribution of each line and the expected travel time to destination, a stop model usually yields the probability of boarding each line, the associated expected waiting and travel times to destination. The line probabilities are in turn used to allocate passengers onto different lines (network loading phase) while the waiting times and the expected travel times enter into the computation of the average travel cost for each origin-destination pair. When congestion is a relevant issue, 2

that is when the travel time (or cost) is a function of the flow pattern resulting from the network loading, these computations must be repeated to reach a stable network load. For equilibrium assignment models, the iterative adjustments lead to an equilibrium passenger flow pattern. In this work we generalize the behavioral model at the stop to the case where, in addition to the headway distributions and the line travel times, we also have on-line information displayed at the stop about the waiting times of the available lines. We also investigate the impact of service regularity on the line loads by considering the family of Erlang headway distributions, which encompasses both the exponential and the deterministic distributions. This new stop model provides straightforward mathematical formulas for the computation of line probabilities and expected travel time, and thus may be embedded into many existing assignment models to handle transit networks with on-line information at stops and general headway distributions. The basic assumptions of the model are the following: transit line headways are statistically independent with given continuous distributions; passengers arrive randomly at the stop; passengers have a good estimate of each line travel time, i.e., the expected travel time from the stop to destination once boarded a carrier of the line including subsequent transfers; passengers can retrieve the line waiting times from the information displayed at the stop; passengers choose to board the line that shows the minimum total time to destination, i.e., the sum of the line waiting time and of the line travel time. The aim is to develop a tool that can measure the impact of providing on-line information about carrier arrivals upon the average passenger flows on the transit network. To this end the line waiting times displayed at the stop will be considered independent random variables. In this perspective, the line shares and (minus) the expected travel time are respectively the choice probabilities and the satisfaction that can be obtained from a discrete choice model where the randomness of the alternatives utility is intrinsic to the supply instead of being derived as usual from the user s and/or the modeler s errors. Note that the attractive set is reduced here to a singleton identified by the line with the best total time. As a consequence, the line share is now a measure of the line attractiveness in contrast to being the probability that a carrier of the line is the first one to arrive at the stop among the lines of the attractive set as in the classical models. For the sake of simplicity, we have assumed that time is the only measure of disutility. However, extending the framework to handle generalized costs is a straightforward task. 1.2 The basic framework For each line i L n, where L n = {1, 2,..., n} is the set of lines serving a given stop, we assume that the following attributes are exogenously given: s i the line travel time; f i (w) the probability density function of the line waiting time, with f i (w) =, w <. Note that for passengers who arrive randomly at a stop, the line waiting time probability density function is related to the headway distribution through the formula (Larson & Odoni 1981): f i (w) = w g i (h) dh = λ i (1 G i (w)), (1) E[h i ] 3

where h i is the headway of line i, g i (h) its probability density function, G i (h) its cumulative distribution function, and 1/E[h i ] = λ i is the inverse of the mean headway, generally referred to as the frequency. Let F i (w) denote the cumulative distribution function of the line waiting time and F i (w) its complement, then we have: F i (w) = w f i (v) dv, F i (w) = 1 F i (w) = w f i (v) dv. The probability of boarding line i is equal to the probability that i is the line with the best total time. Since the line headways and consequently the line waiting times are assumed to be independent from each other, the latter may be expressed as: π i = f i (w) j L n\{i} Prob(w j w + s i s j ) dw, where j L n\{i} Prob(w j w + s i s j ) is the joint probability that the total time of line i is better than or equal to that of any other line, when the waiting time displayed for the next carrier of line i is equal to w. Using: Prob(w j w + s i s j ) = F j (w + s i s j ), we can rewrite the probability of boarding line i as: where γ i (w) denotes: π i = γ i (w) = f i (w) γ i (w) dw, (2) F j (w + s i s j ), (3) j L n\{i} and may be interpreted as the probability density function of the waiting time at the stop conditional to boarding a carrier of line i. The expected waiting time conditional to boarding a carrier of line i is therefore: EW i = wγ i (w) dw. Summing EW i over all lines i L n gives the expected waiting time at the stop: EW = wγ i (w) dw = w γ i (w) dw. (4) Thus, the sum γ i (w) is the probability density function of the waiting time at the stop. By definition the expected travel time once boarded is: ES = π i s i, and then the expected travel time is: ET = EW + ES = 4 (w + s i )γ i (w) dw.

It is worth noting that a different expression of the expected travel time can be obtained. Indeed, the travel time at the stop y can be seen as the minimum of the total times of the lines serving the stop: y = min{w i + s i : i L n }. Let φ(y) denote the probability density function of y and Φ(y) the complement of its cumulative distribution function. Since the line headways are statistically independent, the probability that the travel time at the stop y is greater than or equal to a given value τ: w i + s i τ is given by the product of the probabilities that for each line i: Φ(τ) = Prob(y τ) = Prob(w i τ s i ) = F i (τ s i ). (5) Furthermore, since φ(y) = for y < E[y] = Thus, combining (5) and (6), we have: 1.3 Bounded headway ET = E[y] = τ φ(τ) dτ = Φ(τ) dτ. (6) F i (τ s i ) dτ. (7) If there exists a finite upper bound u j of the probability density function f j (w), for at least one line j L n, that is f j (w) =, w / [, u j ], then it may be useful to specify the upper and lower limits of the integrals of equations (2) and (4) for computational purposes. First, observe that the following holds for any line i L n : 1, if w s j s i ; uj F j (w + s i s j ) = w+s i s j f j (v) dv, if s j s i < w < u j + s j s i ; (8), if w u j + s j s i. In the first case, line i dominates line j regardless of the value of the line waiting time w j (the total time w + s i proves to be not greater than the line travel time s j ), while in the last case the reverse is true. For added insights, assume that the lines in L n are indexed in non-decreasing order of the line travel time, that is: s 1 s 2... s n, and denote L i = {j L n : j i} and us = min{u j + s j : j L n }. This leads to the following expression for the line probability: { us si π i = γ i (w) dw, if s i < us;, otherwise. Then, the subset of the lines with positive probability is L t, where t = max{i : i L n, s i < us}. 2 Relationships with the classical stop model without on-line information 2.1 The classical model viewed as a particular instance of the proposed framework The classical stop model is based on the assumption that the passenger will board the first arriving carrier which belongs to the attractive set L L n. Therefore, the probability of 5

boarding line i L is equal to the probability that a carrier of i is the first one to reach the stop. Since the line headways are assumed to be independent from each other, this probability is given by: π i = f i (w) j L \{i} F j (w) dw, (9) where f i (w) is the density of probability that a carrier of line i arrives after w units of time, while j L \{i} F j(w) is the probability that none of the other carriers arrives before that of line i. Using the same technique that has been adopted to derive equation (7), the expected waiting time can be expressed as (Gendreau 1984): EW = i L F i (w) dw. (1) Similarly to the case with on-line information, the expected travel time is in turn obtained by summing to the expected waiting time the product of the line travel time and the line probability for each attractive line. When there is some bounded headway, the upper limit of integration + in (9) and (1) can be replaced by u = min{u j : j L n }, because at least one carrier must arrive within time u. In contrast to the case with on-line information, the line travel times here are excluded from the expression of line boarding probabilities which depends solely on the headway distributions but are implicitly considered in the definition of the attractive set. It is worth noting then if all lines had the same line travel time, s i = s j, i, j L n, equations (2)-(3) would reduce to equation (9) and (4) to (1). The passenger choosing to board the carrier of the line which provided the shortest total travel time would indeed board the first arriving carrier. Therefore, the classical common lines model where the attractive set includes all available lines can be seen as a particular instance of the general model with on-line information. 2.2 The line travel times and the attractive set Note that, in general, the above expressions of the line probabilities and of the expected waiting time can be applied to any given subset L L n of all the lines serving a stop. Let ET L denote the expected travel time from the considered stop to the destination corresponding to subset L. The attractive set L is a subset of L n that yields the least expected travel time, thus we have: ET L ET L, L L n. The determination of the attractive set L requires in general the calculation of ET L for all possible subsets of L n. However, from a behavioral standpoint it is somewhat counter-intuitive to consider an attractive set where any member of this set may have a larger line travel time than the one of an excluded line and thus it is generally assumed that passengers only consider the feasible sets L j, with j = 1,..., n. Under this assumption, a simple algorithm to determine the attractive set is to calculate in turn ET L i, for i = 1,..., n, and then choose: L = argmin{et L i : i = 1,..., n}. It is well known (Spiess 1984, Nguyen & Pallottino 1988) that in the particular case of exponential headways a greedy approach can be adopted, where the progressive calculation of the values ET L i can be stopped as soon as the addition of the next line produces an increase of the expected travel time. Hence, in contrast to the case with on-line information, the line travel times do not enter directly into the computation of the line probabilities and expected waiting times but rather indirectly through the definition of the attractive set. 6

3 The Erlang headway distributions Depending on the probability density functions of the line waiting times, expressions (2)-(3) and (4) may or may not have a closed form, and hence may or may not be easy to compute. For practical purposes, we will focus on a particular family of headway distributions that can be applied to a large number of cases, namely the Gamma functions with integer parameter, which are also known as Erlang distributions. Let the headway h i of each line i be a random variable with an Erlang distribution having integer shape parameter m i and expected value 1/λ i ; that is: { e m i λ i h (m i λ i ) m i h m i 1 g i (h) = (m i 1)!, if h ;, otherwise. The above distribution describes the sum of m i independent Poisson processes and thus the memory of the overall process lasts for m i elementary events. The shorter is the memory of the process the higher is its irregularity, for m i = 1, the Erlang distribution reduces to the exponential distribution, while for m i +, it yields the deterministic headway. Both these extremal distributions are widely used in transit models, while the Erlang distribution with general shape parameter has been extensively investigated in Gendreau (1984) and in Marguier & Ceder (1984). In Bouzaïene-Ayari, Gendreau & Nguyen (21) Erlang distributions are used to describe the variability of the line headway in a network loading model for transit systems with explicit vehicle capacity constraints. Introducing the Erlang headway into formula (1) yields the following probability density function of the line waiting time: λ i e m iλ i w m i 1 (m i λ i w) k k!, if w ; f i (w) = k=, otherwise, and the following complement of the cumulative distribution function: e m iλ i w m i 1 (1 k m F i (w) = i ) (m iλ i w) k k!, if w > ; k= 1, otherwise. It is not difficult to show that here π i, EW and ET may be written as a sum of integrals where each term has the following form: β e αx x r dx = β e αx r k= r! k! x k + c, αr k+1 where α and β are real parameters, while r is an integer parameter and c is an additive constant. Therefore, for the Erlang distributions the computations required by the proposed model can be performed in closed form. In Gendreau (1984) an efficient algorithm is proposed to compute the expected waiting time in the standard case without on-line information. This scheme can be extended to the computation of the line probabilities and of the expected travel time in the present case with on-line information. In the next two sections, we will take a closer look at the two extremal cases that pervade existing transit models. 7

3.1 The exponential case Consider first the exponential headways. In this case, we have for each line i an exponential waiting time distribution with probability density function: { λ i e λiw, if w ; f i (w) =, otherwise. We then have the following complement of the cumulative distribution function: { e λ j(w+s i s j ), if w > s j s i ; F j (w + s i s j ) = 1, otherwise. To obtain the results for the case without on-line information it suffices to set the line travel times of all the different lines to a fixed value in the above expression, or equivalently to eliminate them all together. This leads respectively to the well known expressions for the line probabilities and the expected waiting time: π i = f i (w) EW = j L n\{i} F j (w) dw = F i (w) dw = λ i e λ iw j L n\{i} e λ jw dw = λ i e λiw dw = e ( e ( P j Ln where Λ n = n j=1 λ j denotes the combined frequency of all lines. 3.2 The deterministic case P λ j )w λ j Ln i dw =, Λ n λ j )w dw = 1 Λ n, With deterministic headways, we have a uniformly distributed waiting time for each line i with probability density function: { λ i, if w 1/λ i ; f i (w) =, otherwise, where 1/λ i = h i = u i. Consequently, equation (8) becomes: 1, if w s j s i ; F j (w + s i s j ) = 1 λ j (w + s i s j ), if s j s i < w < u j + s j s i ;, if w u j + s j s i. Again, we may easily derive the following results for the case without on-line information: π i = u f i (w) j L n\{i} F j (w) dw = λ i u j L n\{i} (1 λ j w) dw, EW = u F i (w) dw = u j L n (1 λ j w) dw. 8

4 Embedding the proposed stop model in transit assignment procedures So far, we have proposed a method for determining the probabilities of the available lines and the corresponding expected travel time for a single transit stop, assuming that the line travel times s j, j L n, are exogenously given. However, a transit trip may involve several transfers, so that the line travel times themselves are expected travel times resulting from downstream stop models. In order to obtain an optimal strategy in the framework of a transit assignment without recurring to an explicit enumeration of hyperpaths, the following optimality principle must hold: all sub-strategies of an optimal strategy are optimal themselves. To prove that the optimality principle holds, we will investigate the sensitivity of the proposed stop model with respect to changes of the line travel times. In the case with on-line information, by definition, we have: ET = E[min{w j + s j : j L n }] =... min{w j + s j : j L n } j L n f j (w j ) dw 1... dw n. The function ET is continuous in the space R n of the line travel time variables s j, j L n. Meanwhile, note that the function min{w j + s j : j L n } is not everywhere differentiable. Indeed, for any given w i, we have: min{w j + s j : j L n } 1, if w j (w i + s i s j, + ) j L n \ {i}; = s i, if j L n \ {i} : w j [, w i + s i s j ). Where i is not the only line yielding the minimum total travel time, so that there exists another line j L n \ {i} such that w j = w i + s i s j, the derivative is not well defined. However, at these points the left partial derivative is 1 and the right partial derivative is. By splitting each integral defined on [, + ) into the sum of two integrals defined on [, w i + s i s j ] and on [w i + s i s j, + ), all the points such that w j = w i + s i s j, with j L n \ {i}, are on the boundary of the intervals of integration. These points form a set of null measure. Thus, by eliminating the zero valued integrals we obtain: ET + = f i (w) f j (w j ) dw j dw. s i w+s i s j j L n\{i} The nested form reduces to a product of integrals since the probability density functions are independent from each other. Taking into account equation (2)-(3), we obtain: ET s i = π i. (11) Equation (11) also holds in the case without on-line information, as we have: ET = EW + π i s i, where, based on (9) and (1), EW and π i, i = 1,..., n, are independent of the line travel times. Based on (11), since line probabilities are non-negative, the optimality principle holds for the proposed stop models. Indeed, the expected travel time is minimized when all the line travel times are minimized. 9

5 Numerical examples 5.1 Common lines at a single stop To illustrate the impact of on-line information at a stop on the distribution of loads, we firstly apply the proposed framework to a simple case with three lines having the same headway E[h] = 2 min and the following line travel times: s 1 = 8 min, s 2 = 14 min, s 3 = 16 min. In the classical case without on-line information, regardless the level of irregularity, all the three lines are attractive, and since they have the same mean headway, they will have the same choice probability. On the contrary, in the case with on-line information it may happen that, based on the line waiting times displayed at the stop, it is preferable to wait for a faster line instead of boarding an incoming slower one. This produces higher probabilities for faster lines. Indeed, the probability of the fastest line has been found to vary from 52.9% for the exponential case to 66.5% for the deterministic case; conversely, the probability of the slowest line varies from 2.1% to 12.5%. Thus, in the deterministic case, the on-line information doubles the number of passengers choosing the fastest line and almost 2/3 of passengers leave the slowest line. This small example shows that the on-line information may have a significant impact on the use of the transit lines, which must be taken into consideration when simulating and dimensioning the transit network for planning purposes. We now analyze the impact of the on-line information on the expected travel and waiting times and the sensitivity of the two stop models with respect to the service irregularity. Consider the example given in Table 1, where for simplicity all lines are assumed to have the same Erlang shape parameter m. The results are reported in Tables 2-3. line i mean headway E[h i ] travel time s i 1 2 min 3 min 2 15 min 4 min 3 1 min 45 min Table 1: Line attributes m π 1 π 2 π 3 EW ER ET 1.587.257.156 6.81 34.92 41.73 2.653.233.114 6.52 34.22 4.74 3.687.218.95 6.55 33.61 4.16 4.77.29.84 6.6 33.35 39.95 1.755.185.6 6.85 32.75 39.57 +.85.16.35 7.27 32.12 39.39 Table 2: Line probabilities and travel times with on-line information For the same service regularity, the expected travel time when on-line information is available is systematically lower than that of the classical case. The relative difference of the expected travel time between the two cases may decrease or increase when the regularity increases, and in general is not monotonic, as shown by the above example where we obtain the improvements ET = (ET without info ET with info )/ET without info reported in Table 4. With on-line information, we note that the fastest line, which is also the one with lower frequency, has a high share (always greater than 5%) which increases monotonically with the regularity of the service. For the standard case in which on-line information is not available, the 1

m π 1 π 2 π 3 EW ER ET 1.429.571 8.42 35.45 43.87 2.419.581 6.98 35.73 42.71 3.413.587 6.48 35.79 42.28 4.49.591 6.25 35.84 42.9 1 1 11.2 3. 41.2 + 1 1. 3. 4. Table 3: Line probabilities and travel times without on-line information m ET 1 4.88% 2 4.61% 3 5.1% 4 5.8% 1 3.53% + 1.53% Table 4: Improvements due to the on-line information fastest line has a lower share (always smaller than 5%) when the service is irregular, while it is the only one to be used when the service is regular. As expected, in both cases the more regular the service is the lower the expected travel time is. In addition, this improvement is monotonic. It also appears that the impact of service irregularity on the expected travel time is considerably lower when on-line information is available at the stop. Indeed, the gap in terms of expected travel time between the deterministic and the exponential headways decreases from 9.67% to 5.94%. In the case with on-line information, the expected waiting time increases with the regularity of the service as passengers accept to wait longer in order to board a line that takes less time to reach the destination. Note that, in the case without on-line information, the number of attractive lines decreases when the regularity increases. 5.2 The attractive set In the case without on-line information, the determination of the attractive set can be solved heuristically by the sequential approach proposed in subsection 2.2. That is, all the lines are added in increasing order of their line travel times and the attractive set is the ordered set with the least expected travel time. It s worth noting that the greedy approach (Chriqui & Robillard 1975), based on stopping the processes as soon as the addition of the next line produces an increase of the expected travel time, can be applied only to the case with exponential headways (Marguier 1981). For instance, consider the example given in Table 5. For exponential headways, as shown in Table 6, we have: ET 1 = 57., ET 1,2 = 49.1 and ET 1,2,3 = 4.47. While for deterministic headways, as shown in Table 7, we have: ET 1 = 42., ET 1,2 = 42.3 and ET 1,2,3 = 41.14. In the latter case, the greedy approach produces the line set {1} while the attractive set is {1, 2, 3}. Note that by rearranging the order of the lines according to the expected travel time of the line considered separately ({3, 1, 2} in Table 6 and {1, 3, 2} in Table 7), the greedy approach is successful. Moreover, the ordered sets obtained in this way are optimal compared to any other set with the same number of lines. So far, we are neither able to prove that this result holds in 11

line i mean headway E[h i ] travel time s i 1 3 min 27 min 2 5 min 38 min 3 5 min 4 min Table 5: Line attributes line set info π 1 π 2 π 3 EW ER ET 1 N 1 3. 27. 57. 2 N 1 5. 38. 88. 3 N 1 5. 4. 45. 1 2 N.618.372 18.25 3.85 49.1 1 3 N.136.83 3.82 36.89 4.71 2 3 N.81.875 3.93 38.7 42.1 1 2 3 N.125.68.771 3.68 36.79 4.47 1 2 3 Y.43.71.48 5.6 33.49 39.9 Table 6: The exponential case line set info π 1 π 2 π 3 EW ER ET 1 N 1 15. 27. 42. 2 N 1 25. 38. 63. 3 N 1 2.5 4. 42.5 1 2 N.7.3 12. 3.3 42.3 1 3 N.83.917 2.36 38.92 41.28 2 3 N.5.95 2.42 39.9 42.32 1 2 3 N.81.47.872 2.29 38.86 41.14 1 2 3 Y.51.49.441 5.8 33.29 38.37 Table 7: The deterministic case general nor to find a counterexample. When the on-line information is provided, the set of lines with positive probability is {1, 2, 3} for both exponential and deterministic headways. Comparing the results obtained respectively for the two cases with and without on-line information, we observe a remarkable difference in the line loads. In particular, in the case with deterministic headways and without on-line information almost all users (87.2%) choose the slowest and more frequent line 3 and just a few (8.1%) chooses the fastest and less frequent line 1, while in the case with on-line information line 1 is choosen by 51.% of passengers and the share of users choosing line 3 decreases to 44.1%. 5.3 Assignment on a small network The transit network of Sioux Falls has been considered for a numerical application of the proposed framework. On a base network of 76 directed arcs and 24 nodes (all of which are centroids), the itineraries of 5 bus lines are defined, covering the entire network (each node is served by at least one line). In particular, the original network presented in LeBlanc (1988) has been slightly modified (see Figure 1) by moving the south terminal of line B from node 5 to node 21 so that line A and line B overlap along the North-South direction between nodes 5 and 21. In addition, 12

line B is designed as a low frequency express line, while all other lines have a slow speed and a frequency that is twice higher. This yields a supply configuration more suitable for the present investigation. 1 2 3 4 5 6 7 9 8 12 11 1 16 17 18 14 15 19 23 22 13 24 21 2 line i mean headway E[h i ] speed V i itinerary of line i A 1 min 15 km/h 1 3 4 5 9 1 15 22 21 B 2 min 4 km/h 2 6 5 9 1 15 22 21 C 1 min 15 km/h 12 11 1 16 17 19 2 D 1 min 15 km/h 13 24 23 14 15 E 1 min 15 km/h 7 8 16 18 Figure 1: Transit network nodes and arcs The analysis is performed considering three Erlang headway distributions with different shape parameter: m = 1 (exponential headway), m = 4, and m + (deterministic headway). The results reported in Table 8 refer to an all or nothing network loading. The total time decreases substantially with the service regularity, but the improvement due to the presence of on-line information is not significant (less than 1%). The share of the fast line B increases with the service regularity and with the presence of on-line information. This result is more pronounced at the southern stops where the common lines A and B are in direct competition for a greater number of O/D pairs. Other numerical results, not reported here, also show that the impact of service regularity and the presence of on-line information is less relevant for short distance O/D pairs where the relative weight of line waiting time with respect to the line travel time decreases. For this reason, on real networks the impact of providing on-line information on flow patterns may be light, yet definitely relevant. Conclusion In this paper we have developed a general framework for investigating passenger s route choice in transit networks when on-line information about carriers arrival times at stops are available. 13

info Y Y Y N N N m 1 4 + 1 4 + total time 4419 h 3946 h 367 h 4441 h 398 h 373 h stops passengers share B share B share B share B share B share B 5-9 838.345.36.387.321.331.365 9-1 111.375.388.46.325.326.351 1-15 1648.443.466.472.322.297.348 15-22 795.491.537.572.33.314.445 22-21 275.516.593.615.331.345.673 21-22 275.531.582.67.342.327.673 22-15 795.58.548.557.342.316.425 15-1 1648.475.496.492.35.318.355 1-9 111.496.52.55.424.438.43 9-5 838.55.515.525.452.492.48 Table 8: Total times and shares on line B Assuming that the ultimate passenger s objective is to minimize the total travel time to destination, a new stop model is proposed for determining the probability of boarding each line available at a given stop and the corresponding expected waiting time when headways have independent general distributions. In addition, by showing that the classical common lines model may be interpreted as a particular instance of the proposed framework we extend the existing results to cover general headway distributions. Small size numerical examples are presented to illustrate the proposed framework. Using the Erlang family of distributions with various shape parameters, and in particular with the extremal parameters corresponding to deterministic and exponential headways, we compare the proposed model with the classical one. Drastic differences are observed in terms of proportions of passengers boarding slow and fast lines, while the differences on travel times are less important yet definitely relevant. Moreover, the impact of service regularity is significant for both cases. These results imply that the existing models are inadequate to deal with the presence of online information at stops and service regularity, and thus may lead to distorted traffic patterns for planning purposes. Acknowledgment We are grateful to the referees for their constructive comments that have lead to a much improved paper. References Bouzaïene-Ayari, B., M. Gendreau & S. Nguyen 1995, An equilibrium-fixed point model for passenger assignment in congested transit networks, Publication 95-57, Centre de Recherhce sur les Transports, Université de Montréal, Montréal, Canada. Bouzaïene-Ayari, B., M. Gendreau & S. Nguyen 21, Modelling bus stops in transit networks: A survey and new formulations, Transportation Science 35, 34 321. 14

Bowman, L.A. & M.A. Turnquist 1981, Service frequency, schedule reliability and passenger wait times at transit stops, Transportation Research A 15(6), 465 471. Chriqui, C. & P. Robillard 1975, Common bus line, Transportation Science 9(2), 115 121. Gendreau, M. 1984, Une étude approfondie d un modèle d équilibre pour l affectation des passagers dans les réseaux de transport en commun, PhD thesis, Département d informatique et de recherche opérationnelle, Université de Montréal, Canada. Hasseltröm, D. 1981, Public Transportation Planning A Mathematical Programming Approach, PhD thesis, Department of Business Administration, University of Gothenburg, Sweden. Jolliffe, J.K. & T.P. Hutchinson 1975, A behavioral explanation of the association between bus and passenger arrivals at a bus stop, Transportation Science 9(3), 248 282. Larson, R.C. & A.R. Odoni 1981, Urban Operations Research, Prentice-Hall, Englewoods Cliffs. LeBlanc, L.J. 1988, Transit system network design, Transportation Research B 22, 383 39. Marguier, P.H.J. 1981, Optimal strategies in waiting for common bus lines, Master s thesis, Department of Civil Engineering, M.I.T., Cambridge, USA. Marguier, P.H.J. & A. Ceder 1984, Passenger waiting strategies for overlapping bus routes, Transportation Science 18(3), 27 23. Nguyen, S. & S. Pallottino 1988, Equilibrium traffic assignment for large scale transit networks, European Journal of Operational Research 37(2), 176 186. Nguyen, S. & S. Pallottino 1989, Hyperpaths and shortest hyperpaths, in B. Simeone, ed., Combinatorial Optimization, Vol. 143 of Lecture Notes in Mathematics, Springer-Verlag, Berlin, pp. 258 271. Nguyen, S., S. Pallottino & M. Gendreau 1998, Implicit enumeration of hyperpaths in logit models for transit networks, Transportation Science 32(1), 54 64. Seddon, P.A. & M.P. Day 1974, Bus passenger waiting times in Greater Manchester, Traffic Engineering and Control 15(9), 442 445. Spiess, H. 1984, Contributions à la théorie et aux outils de planification des réseaux de transport urbain, PhD thesis, Département d informatique et de recherche opérationnelle, Université de Montréal, Canada. Spiess, H. & M. Florian 1989, Optimal strategies: A new assignment model for transit networks, Transportation Research B 23(2), 83 12. Wu, J.H., M. Florian & P. Marcotte 1994, Transit equilibrium assignment: a model and solution algorithms, Transportation Science 28(3), 193 23. 15