Supplementary Technical Details and Results

Size: px
Start display at page:

Download "Supplementary Technical Details and Results"

Transcription

1 Supplementary Technical Details and Results April 6, Introduction This document provides additional details to augment the paper Efficient Calibration Techniques for Large-scale Traffic Simulators. In what follows, we provide sections that include additional technical details on the implementation of experiments, specifications on the simulation model, and additional analysis on the case studies. Section 2 provides additional technical details on implementation of the analytical traffic model, the simulation model, and the Simulation-based Optimization (SO) algorithm. Section 3 provides additional analysis for the toy and Berlin network case studies. Section 4 provides a detailed description of the route choice model used in the simulator, MATSim. 2 Implementation Details 2.1 Estimation of the exogenous parameters of the analytical traffic model We solve the analytical traffic model to get the analytical approximation of link flows λ i (θ). In order to solve the analytical traffic model, we first need to estimate the exogenous parameters. The exogenous parameters of the system of nonlinear differentiable equations are as follows. d s expected travel demand for OD pair s; µ i service rate of queue i; l i space capacity of queue i; l i length of link i (road length); v i maximum speed of link i; S set of OD pairs; Q set of queues; R set of routes; R s set of routes of OD pair s; G ij set of routes that consecutively go through queues i and j; H i set of routes that go through queue i; T i set of routes that start with queue i; Ψ r set of links of route r. In general, the exogenous parameters of the analytical traffic model are inferred mainly based on the input to the simulation counterpart, MATSim [2]. The input to the simulation model consists of: (1) configuration. (2) network which describes the simulation network infrastructure 1

2 in terms of links and nodes. In the network, the attributes associated with links and nodes are specified such as link length, maximum speed, and flow capacity, etc. (3) plans/population which describes the demand to the simulation model in terms of a list of day plans. A day plan is a chained activities and trips which a traveler intend to go through for a whole day. Each trip is a sequence of links which has its first link as origin and the last link as destination. A time stamp is attached to the trip as well. The multiple realizations of trips between a given OD pair make the route choice set. For link i, exogenous parameters such as link length l i, maximum speed v i, service rate µ i as the flow capacity can be obtained directly from the network input to the simulation model. The space capacity l i of queue i is approximated by the following equation: l i = l i /d, (1) where d is the average distance between the front bumpers of two successive vehicles which is set to 7.5 meters. This value is adopted from the effective cell size defined in the simulator, MATSim. The fraction is rounded up to the nearest integer. The expected travel demand d s for OD pair s S is estimated based on the day plans of all travelers of the simulation model. We first extract all trips being selected (there is only one trip in use out of the multiple alternative routes in every assignment iteration) which depart within a given period (i.e., the time step we are interested in, for example, 8-9 am), then obtain the OD pairs for all these trips, and at last summarize the frequencies of all unique OD pairs. The frequency of a specific OD pair s represents the number of travelers traveled during the given period which is the expected travel demand d s in the unit of vehicles per time period (e.g., vehicles/hour). The set S consists of all unique OD pairs. For the simulation model, the set of route alternatives for a given OD pair is endogenous and the number of alternative routes for a given OD pair is fixed (i.e., 10). For the purpose of tractability, a route choice set of fixed size is synthesized for the analytical traffic model. Specifically, we synthesize a route choice set R s of fixed size (i.e., 10) for any given OD pair s S. This resulting route choice set is exogenous to the analytical traffic model. For details regarding the generation of route choice set, please refer to Section 2.3. The pool of all route choice sets for all OD pair s S constitutes the set of routes R. For each route r R, the links on this route constitutes the set of links Ψ r of route r. All links involved in these trips consists the set of queues Q. The set of routes T i that start with queue i is obtained by find the subset of routes from the set of routes R which have link i as their first link. For all routes in R which depart during the given period, we filter out a subset of routes which contain link i. This subset of routes constitutes set H i which is the set of routes go through queue i. Similarly, G ij which is the set of routes that consecutively go through links i and j can be formed by finding out the subset of routes which have link i and link j consecutively. 2.2 Feasible region of behavioral parameter θ The behavioral parameter θ represents the travelers sensitivity to travel time. The behavioral parameter is negative and upper bounded by zero since additional travel time decreases the 2

3 utility of traveling. The lower bound of the behavioral parameter depends on the network and is calculated using Equation (2) which was proposed in [1]. θ L = ln2 (1 ξ)ν sp, (2) where ξ is a constant which is greater than 1. ν sp is the shortest route cost among all OD pairs. Equation (2) defines a scale parameter θ L such that the probability of choosing a route of cost ξν sp is half the probability that the shortest path being chosen. By applying this method with ξ = 1.042, the lower bound for the behavioral parameter θ for the toy network and the Berlin network is calculated to be 60 for travel time in the unit of hours. Therefore, we obtain the feasible region Θ for the behavioral parameter θ as [ 60, 0]. The resulting feasible region for the behavioral parameter not only contains the pre-calibrated behavioral parameter value (i.e., -6) defined in the simulator, MATSim, but also retains a relatively large region to take possible local variance of the behavioral parameter into consideration. 2.3 Route choice set of the analytical traffic model The simulator, MATSim, makes route choice from a choice set of fixed size (i.e., 10). Each route in a choice set is assigned the score (i.e., utility) which it was realized last time. In each assignment iteration of the simulation, most of the travelers (i.e., 83.3%) make route choice according to the Multinomial Logit (MNL) model based on the scores of routes, and a small portion of travelers (i.e., 16.7%) generates new routes and eliminates one of the current routes with worst score. The travelers who generate new route are randomly sampled in each assignment iteration. Therefore, the route choice set is endogenous which varies with behavioral parameter θ and across assignment iterations. As a matter of convenience, we consider an exogenous route choice set for the analytical traffic model. To keep consistent with the simulator, we derive a route choice set of fixed size (i.e., 10) for each OD pair. The route choice set for the analytical traffic model does not vary with behavioral parameter θ and across iterations. However, a desired route choice set should take into account the variability of route choice sets in the simulator. We propose an algorithm for generating the route choice set for the analytical traffic model. By running simulation until convergence with multiple different behavioral parameters, we obtain a route choice set from the last assignment iteration of each of these experiments. These route choice sets are different from each other due to different behavioral parameters. We put the route choice sets associated with different behavioral parameters for the same OD pair s together. The route choice set R s for the analytical traffic model for a given OD pair s is derived by selecting the most representative routes of the fixed size (i.e., 10) from the pool of simulated route choice sets associated with different behavioral parameters. The most representative route choice set for a given OD pair s is the one which has the maximum overlap with any of the simulated route choice sets for the OD pair s. The overlap of two route choice sets are measured by the proportion of total distance of common unique links of these two route choice sets. The resulting route choice set R for all OD pairs is provided as exogenous input for the analytical 3

4 traffic model. 2.4 Initialization of the endogenous variables of the analytical traffic model The system of nonlinear equations of the analytical traffic model is solved iteratively. For each iteration of the solution process, we first solve the assignment model given fixed network conditions, and then update the network conditions given a fixed assignment. Therefore, we need to estimate the expected link travel time t i of link i at the very beginning of the iterative solution process. The expected link travel time t i is initialized as the free flow travel time on link i which is defined as: t i = l i v i. (3) With the initial expected link travel time estimated, we can proceed to solve for the whole system of equations. For details regarding the iterative process of solving the analytical traffic model, please refer to Section Numerical evaluation of the analytical traffic model With the route choice set generated, exogenous parameters estimated, and endogenous variables initialized, we proceed to solve the analytical traffic model to get the analytical approximation of link flows. In this section, we state technical details of how the system of equations of the analytical traffic model is numerically evaluated. At first, we recap the formulation of the analytical traffic model used in our SO calibration algorithm. f r = s S d s p sr r R (4) p sr = e θtr j R s e θt j s S, r R s (5) t r = i Ψ r t i r R (6) t i = l i v i + ñi λ i i Q (7) 4

5 ñ i = ρ i (l i + 1) ρ li+1 i 1 ρ i 1 ρ l i+1 i i Q (8) ρ i = λ i µ i i Q (9) λ i = γ i + j Q p ji λ j i Q (10) p ij = r G ij f r k H i f k i Q, j Q (11) γ i = r T i f r i Q (12) To apply the iterative heuristic, we first discompose the above system of equations into two separate systems of equations. The first system of equations consists of Equations 4, 5, 6, 11, and 12. The second system of equations consists of Equations 7 through 10. A typical iteration of the heuristic consists of two main steps. At step one, the first system of equations is solved for turning probabilities p ij given the expected link travel time t i (free flow travel time for the very first iteration). Then, the second system of equations is solved for the updated network conditions (i.e.,expected link travel time t i ) with the turning probabilities p ij fixed. We use stopping (or convergence) criteria that stop the iterative process when a pre-defined condition is satisfied. Commonly used stopping criteria are based on the change of solution achieved during one iteration. For instance, given a small ϵ > 0, the iterative process is stopped at iteration a when the following condition is met: x a x a 1 < ϵ, (13) where we denote the solution of iteration a of the iterative process by x a. maximum acceptable number of iterations is usually specified. Additionally, a Specifically, we keep track of the change of variables t i and p ij of the system of equations of two consecutive iterations a 1 and a of the iterative heuristic until it is less than a threshold ϵ as shown in (14). i Q ( t a i t a 1 ) 2 i + ( ) p a ij p a 1 2 ij < ϵ, (14) t p i,j Q 5

6 where t and p are the maxima of t i and p ij respectively and defined as: which are used to normalize these two variables t = max t i, i Q (15) p = max i,j Q p ij. (16) The threshold ϵ is set to 10 3 and the maximum number of iteration is set to 100. The iterative solution process terminates either when the threshold ϵ is satisfied or the number of iteration reaches the maximum iteration. 2.6 Estimation of f(θ) Due to stochasticity, the stochastic user equilibrium flows are calculated as the expectation over multiple simulation replications. Each simulation replication is independent and depended on a different random seed. And all simulation replications are simulated afresh in parallel. Since each simulation is computational expensive, we choose a total number of simulation replications U for evaluating the expectation as a tradeoff between the expensive simulation cost and the stochasticity of the simulator. Before the simulation starts, we use Matlab function randi() to generate U pseudo-random integers uniformly between 1 and the maximum integer as random seeds with which to start each simulation replication. The expected stochastic user equilibrium flow for link i which is denoted by E[F i (θ; x)] for the behavioral parameter θ is calculated as the average of the simulated flows over all U independent simulation replications for link i as: E[F i (θ; x)] = U 1 U F j i (θ; x), (17) j=1 where F j i (θ; x) is the stochastic user equilibrium link flows for a given behavioral parameter θ of simulation replication j where j [1, U]. Assuming the total number of assignment iterations for a given simulation replication is M, the replication-specific stochastic user equilibrium link flow F j i (θ; x) for link i for a given behavioral parameter θ of simulation replication j where j [1, U] are calculated as the average of flows from the last P assignment iterations for link i: M F j i (θ; x) = k=m P +1 1 P F jk i (θ; x), (18) where F jk i (θ; x) is the simulated flow on link i of the kth assignment iteration of the jth replication for a given behavioral parameter θ. The motivation of using Equation (18) is to smoothen out the simulation observations within a given simulation replication. 6

7 By applying Equations 17 and 18 together, we are able to achieve an evaluation of the stochastic user equilibrium flows with relatively low variance. 2.7 SO algorithm The proposed calibration algorithm uses the Simulation-based Optimization (SO) algorithm in [3]. In this section, we state the changes made to the traditional SO algorithm. In general, the traffic measurement we obtain by solving the analytical traffic model can be a good approximation to the simulation-based traffic measurement, and is usually better than the initial guess θ 0. Different from the traditional SO algorithm, we firstly solve the calibration problem with only the analytical traffic model. To be specific, we solve a similar optimization problem to the metamodel optimization problem using a metamodel with only the analytical problem-specific approximation λ i (θ). The optimization problem solved in the first iteration is formulated as follows: min θ (y i m i (θ)) 2 i I (19) θ L θ θ U h 2 (θ, ṽ; q) = 0. (20) (21) And the metamodel for link i used here takes the following mathematical form: m i (θ) = λ i (θ). (22) In addition, the optimization problem is solved despite the trust region constraint so as to be able to search over the full feasible region Θ. For solving the system of nonlinear equations of the analytical traffic model, we use the iterative method as described in Section 2.5 instead of the Matlab routine fsolve. Similar to the traditional SO algorithm, we use the Matlab routine fmincon for solving the metamodel optimization problem, however, the implementation is different. We implement the metamodel optimization problem with only the behavioral parameter θ as the decision variable. We don t explicitly formulate the analytical traffic model as constraints and only relate it to the objective function. Thus, all endogenous variables of the analytical traffic model do not appear to be decision variables in this problem. The number of decision variables will always be the same regardless of the scale of the network. This makes our calibration algorithm applicable regardless of the scalability issues. The optimization objective function is formulated as a function of the analytical model so that for any behavioral parameter θ proposed by the solver the analytical traffic model yields the corresponding link flows λ i (θ), and hence the corresponding value of the objective function is calculated. 7

8 Toy network Berlin network max min γ inc γ dec θ L θ U 0 0 η τ µ n max Table 1: Algorithmic parameters initialization A list of the algorithmic parameters used for the SO calibration algorithm is displayed below in Table 1. In this table, max, min, and 0 are respectively the maximum, minimum and initial value of trust region radius. γ inc and γ dec are the factors for increasing and decreasing the trust region radius. η is the threshold for accepting new trial points. τ is the threshold for sampling new points. µ is the threshold on the number of consecutive rejected trial points required to reduce the trust region radius. And n max is the maximum number of simulation runs (i.e., computational budget). 2.8 Simulator settings and APIs The Application Program Interface (API) of the traffic simulator, MATSim, is used for interacting with the simulator such as starting simulation externally from Matlab, configure the simulation settings, and extracting simulation observations. In this section, we briefly describe the simulator settings and how we control the simulation model in terms of APIs. In order to obtain the user equilibrium flows for a given behavioral parameter θ, we need to modify the corresponding simulation model parameter by configuring the settings and specify the random seed for different simulation replication. This is achieved by modifying the parameters traveling car and random seed in the configuration file. Then, we start the simulation by supplying the network infrastructure (network file) and demand (plans file). During the simulation, we obtain the stochastic link traffic measurement by counting the number of vehicles on a link during a time period. This is achieved by using the MATSim API LinkLeaveEventHandler to keep track of all vehicles leaving this specific link during this time period. To get the simulation observations, we listen to the completion of an assignment iteration by using the MATSim API EventListener and then calculate the simulated link flows for each link in I for this assignment iteration. We also listen to the end of simulation (i.e., shutdown ) and write the collected simulated link flows for all assignment iterations into the output file. This output file is used for analyzing the objective function and fitting metamodel parameters. The strategy modules used for routing in the simulation are SelectExpBeta and ReRoute. SelectExpBeta doesn t create new plan and selects one of the existing plans of the traveler based on the Multinomial Logit (MNL) model. For details regarding the route choice model in 8

9 MATSim, please refer to Section 4. And ReRoute uses the routing algorithm (i.e., Dijkstra) to calculate new least cost routes using travel times from the previous iteration and create new plans. In other words, ReRoute module is used to take care of the cases when the existing choice set is not suitable for a given behavioral parameter θ. The probability of using SelectExpBeta module is set to 1.0 and using ReRoute module is set to 0.2. In the simulation, each traveler chooses exactly one module per iteration, and this setting defines the probability that a module will be chosen by a traveler. Hence, a traveler has a probability of 5/6 to choose SelectExpBeta module and 1/6 to choose ReRoute module. The selection of these numerical values are to achieve the goal of route choice which updates the choice set slowly and following the multinomial logit model. These settings are also helpful for stabilizing the simulation process. These two modules are in effect until the last assignment iteration for each simulation. 3 Additional results and analysis In this section, we augment the analyses in the paper by providing additional tables and discussions for both the toy network and Berlin network case studies. In order to further analyze and compare the algorithm performance in terms of convergence of model Am and model Aϕ, we consider the convergence statistics listed as follows. Stat 1: Number of convergences. This is the total number of cases which converge for a method and a true value. Stat 2: Number of enter/exit/re-enter-converge. The number of cases which enter, leave, and re-enter the green region (once or for multiple times) but finally stay within the green region. Stat 3: Number of enter/exit/do-not-converge. The number of cases which enter and leave the green region (once or for multiple times) and finally stay off the green region. Stat 4: Number of never-enter/do-not-converge. The number of cases which never enter the green region and stay off the green region all the time. For the toy network, Table 2 displays the results for each method and a pair of true value and initial value (i.e., each experiment). Each row of this table corresponds to one of the figures of algorithmic solutions versus iterations in the paper. Table 3 summarizes the results for each method and a true value. And the values in the table are written as percentage of all cases in a certain category. It is worth noting that the total number of cases for a method and a true value is 9 for the toy network in Table 3. Based on Table 3, we can see that all cases for method Am converge, while there are 3 cases for method Aϕ do not converge. For the general-purpose method Aϕ, there are cases which they first enter the green region, then left the green region, and end up with a un-converged solution. However, this never happens for method Am. 9

10 stat 1 stat 2 stat 3 stat 4 Method Am Aϕ Am Aϕ Am Aϕ Am Aϕ θ 0 = θ = 5 θ 0 = θ 0 = θ 0 = θ = 20 θ 0 = θ 0 = θ 0 = θ = 55 θ 0 = θ 0 = Table 2: Case specific convergence statistics for the toy network stat 1 stat 2 stat 3 stat 4 Method Am Aϕ Am Aϕ Am Aϕ Am Aϕ θ = 5 100% 100% 11.1% 22.2% 0% 0% 0% 0% θ = % 77.8% 33.3% 0% 0% 22.2% 0% 0% θ = % 88.9% 11.1% 0% 0% 11.1% 0% 0% Table 3: Convergence statistics for the toy network On the whole, both method Am and method Aϕ are satisfying with similar performance for the toy network. However, method Am outperforms method Aϕ since it identifies good solutions within less number of SO iterations and always converges. We further analyze and compare the algorithm performance in terms of the convergence statistics for the Berlin network. We first present the convergence statistics for each case with a different initial value of θ in Table 4. Rows in this table correspond to Figures respectively. We then present the convergence statistics for a given initial value in Table 5 as percentages. The total number of cases for each category is 3 in Table 5. Based on Table 5, we can see that all cases for method Am converges, while there are only 2 cases for method Aϕ converge. Different from the toy network, there are 3 cases of method Aϕ which never enter the green region and eventually do not converge. On the whole, method Am outperforms method Aϕ in terms of convergence. In the paper, we carried out a more detailed analysis within the initial green region [ 6, 1]. The stat 1 stat 2 stat 3 stat 4 Method Am Aϕ Am Aϕ Am Aϕ Am Aϕ θ 0 = θ 0 = θ 0 = Table 4: Case specific convergence statistics for the Berlin network 10

11 stat 1 stat 2 stat 3 stat 4 Method Am Aϕ Am Aϕ Am Aϕ Am Aϕ θ 0 = 0 100% 33.3% 33.3% 0% 0% 0% 0% 66.7% θ 0 = % 33.3% 33.3% 0% 0% 66.7% 0% 0% θ 0 = % 0% 0% 0% 0% 0% 0% 100% Table 5: Convergence statistics for the Berlin network green region based on the more detailed estimation of the simulation-based objective function is re-defined as: [ 5, 4.5], [ 3.75, 2.75], and [ 2.25, 2]. The convergence in terms of the newly estimated green region for each case of the Berlin network is summarized in Table 6. The percentage in the table is calculated over all experiments (i.e., 3 experiments) for a specific method and initial value θ 0. In the last row of the table, we also display the overall convergence performance for a method regarding all experiments. Method Am Aϕ θ 0 = % 0% θ 0 = % 33.3% θ 0 = % 0% Overall 55.6% 11.1% Table 6: The convergence percentages for each method and each initial value for the Berlin network In terms of the more detailed analysis, the convergence percentages decrease for both methods. However, method Am still outperforms method Aϕ with a much higher chance of convergence. 4 Route choice model in the simulation model We specify the route choice model used in MATSim in this section as a reference for readers based on the MATSim book [2]. As we mentioned previously, the SelectExpBeta module in MATSim makes route choice between plans according the Multinomial Logit (MNL) model. The probability for a traveler to choose plan i among all alternative plans is formulated mathematically as: p i = ebeta brain score i j ebeta brain score j (23) where score i is the score of plan i from the previous assignment iteration. The scores are taken as utilities. The parameter beta brain is treated as a scale parameter in this model. In our case, we fix beta brain at 1. The score/utility of plan is computed as the sum of all activity utilities V act,w plus the sum of all travel utilities V trav,mode(w). Mathematically, it is formulated as: 11

12 W 1 score i = w=0 W 1 Vact,w i + w=0 V i trav,mode(w) (24) where W is the number of activities associated with plan i. Trip w is the trip that follows activity w. For scoring, the last activity is merged with the first activity to produce an equal number of trips and activities. Specifically, the score/utility of an activity w is calculated as: V act,w = V dur,w + V wait,w + V late.ar,w + V early.dp,w + V short.dur,w (25) where V dur,w is the utility of performing activity w, V wait,w is the utility of waiting for performing activity w, V late.ar,w and V early.dp,w specifies the late arrival penalty and early departure respectively, and V short.dur,w is the penalty for a too short activity. The score/utility for traveling on leg w is calculated as follows: V trav,w = C mode(w) + α trav,mode(w) t trav,w + α m m w + (α d,mode(w) + α m γ d,mode(w) )d trav,w + α transfer x transfer,w (26) where C mode(w) is a mode-specific constant, α trav,mode(w) is the direct marginal utility of time spent traveling by mode, t trav,w is the travel time between activities w and w + 1, α m is the marginal utility of money, m w is the change in monetary budget caused by fares/tolls, α d,mode(w) is the marginal utility of distance, γ d,mode(w) is the mode-specific monetary distance rate, d trav,w is the distance traveled between activities w and w + 1, α transfer is the public transit transfer penalty, and x transfer,w is the binary variable indicating whether a transfer is involved in the current leg. In our model, only the car mode is considered. References [1] M. Bierlaire and G. Flötteröd. Metropolis-Hastings sampling of alternatives for route choice models. In Proceedings of the International Choice Modelling Conference, Leeds, England, July [2] A. Horni, K. Nagel, and K. W. Axhausen. The multi-agent transport simulation MATSim. Zurich and Berlin, [3] C. Osorio and M. Bierlaire. A simulation-based optimization framework for urban transportation problems. Operations Research, 61(6): ,

We can now formulate our model as showed in equation 2:

We can now formulate our model as showed in equation 2: Simulation of traffic conditions requires accurate knowledge of travel demand. In a dynamic context, this entails estimating time-dependent demand matrices, which are a discretised representation of the

More information

Logistical and Transportation Planning. QUIZ 1 Solutions

Logistical and Transportation Planning. QUIZ 1 Solutions QUIZ 1 Solutions Problem 1. Patrolling Police Car. A patrolling police car is assigned to the rectangular sector shown in the figure. The sector is bounded on all four sides by a roadway that requires

More information

MEZZO: OPEN SOURCE MESOSCOPIC. Centre for Traffic Research Royal Institute of Technology, Stockholm, Sweden

MEZZO: OPEN SOURCE MESOSCOPIC. Centre for Traffic Research Royal Institute of Technology, Stockholm, Sweden MEZZO: OPEN SOURCE MESOSCOPIC SIMULATION Centre for Traffic Research Royal Institute of Technology, Stockholm, Sweden http://www.ctr.kth.se/mezzo 1 Introduction Mesoscopic models fill the gap between static

More information

Traffic Flow Simulation using Cellular automata under Non-equilibrium Environment

Traffic Flow Simulation using Cellular automata under Non-equilibrium Environment Traffic Flow Simulation using Cellular automata under Non-equilibrium Environment Hideki Kozuka, Yohsuke Matsui, Hitoshi Kanoh Institute of Information Sciences and Electronics, University of Tsukuba,

More information

Simulation-based travel time reliable signal control

Simulation-based travel time reliable signal control Simulation-based travel time reliable signal control Paper accepted for publication in Transportation Science Xiao Chen School of Highway, Chang an University, Xi an, China Carolina Osorio Department of

More information

Session-Based Queueing Systems

Session-Based Queueing Systems Session-Based Queueing Systems Modelling, Simulation, and Approximation Jeroen Horters Supervisor VU: Sandjai Bhulai Executive Summary Companies often offer services that require multiple steps on the

More information

CIV3703 Transport Engineering. Module 2 Transport Modelling

CIV3703 Transport Engineering. Module 2 Transport Modelling CIV3703 Transport Engineering Module Transport Modelling Objectives Upon successful completion of this module you should be able to: carry out trip generation calculations using linear regression and category

More information

Simulation-based travel time reliable signal control

Simulation-based travel time reliable signal control Simulation-based travel time reliable signal control Carolina Osorio Department of Civil and Environmental Engineering, Massachusetts Institute of Technology (MIT), Cambridge, MA239, USA, osorioc@mit.edu,

More information

Application of Monte Carlo Simulation to Multi-Area Reliability Calculations. The NARP Model

Application of Monte Carlo Simulation to Multi-Area Reliability Calculations. The NARP Model Application of Monte Carlo Simulation to Multi-Area Reliability Calculations The NARP Model Any power system reliability model using Monte Carlo simulation consists of at least the following steps: 1.

More information

More on Input Distributions

More on Input Distributions More on Input Distributions Importance of Using the Correct Distribution Replacing a distribution with its mean Arrivals Waiting line Processing order System Service mean interarrival time = 1 minute mean

More information

Intuitionistic Fuzzy Estimation of the Ant Methodology

Intuitionistic Fuzzy Estimation of the Ant Methodology BULGARIAN ACADEMY OF SCIENCES CYBERNETICS AND INFORMATION TECHNOLOGIES Volume 9, No 2 Sofia 2009 Intuitionistic Fuzzy Estimation of the Ant Methodology S Fidanova, P Marinov Institute of Parallel Processing,

More information

IV. Violations of Linear Programming Assumptions

IV. Violations of Linear Programming Assumptions IV. Violations of Linear Programming Assumptions Some types of Mathematical Programming problems violate at least one condition of strict Linearity - Deterministic Nature - Additivity - Direct Proportionality

More information

A Framework for Dynamic O-D Matrices for Multimodal transportation: an Agent-Based Model approach

A Framework for Dynamic O-D Matrices for Multimodal transportation: an Agent-Based Model approach A Framework for Dynamic O-D Matrices for Multimodal transportation: an Agent-Based Model approach Nuno Monteiro - FEP, Portugal - 120414020@fep.up.pt Rosaldo Rossetti - FEUP, Portugal - rossetti@fe.up.pt

More information

A simulation-based optimization algorithm for dynamic large-scale urban transportation problems

A simulation-based optimization algorithm for dynamic large-scale urban transportation problems A simulation-based optimization algorithm for dynamic large-scale urban transportation problems Linsen Chong, Carolina Osorio Civil and Environmental Engineering Department, Massachusetts Institute of

More information

14 Random Variables and Simulation

14 Random Variables and Simulation 14 Random Variables and Simulation In this lecture note we consider the relationship between random variables and simulation models. Random variables play two important roles in simulation models. We assume

More information

Assortment Optimization under the Multinomial Logit Model with Nested Consideration Sets

Assortment Optimization under the Multinomial Logit Model with Nested Consideration Sets Assortment Optimization under the Multinomial Logit Model with Nested Consideration Sets Jacob Feldman School of Operations Research and Information Engineering, Cornell University, Ithaca, New York 14853,

More information

UNIVERSITY OF YORK. MSc Examinations 2004 MATHEMATICS Networks. Time Allowed: 3 hours.

UNIVERSITY OF YORK. MSc Examinations 2004 MATHEMATICS Networks. Time Allowed: 3 hours. UNIVERSITY OF YORK MSc Examinations 2004 MATHEMATICS Networks Time Allowed: 3 hours. Answer 4 questions. Standard calculators will be provided but should be unnecessary. 1 Turn over 2 continued on next

More information

Traffic Modelling for Moving-Block Train Control System

Traffic Modelling for Moving-Block Train Control System Commun. Theor. Phys. (Beijing, China) 47 (2007) pp. 601 606 c International Academic Publishers Vol. 47, No. 4, April 15, 2007 Traffic Modelling for Moving-Block Train Control System TANG Tao and LI Ke-Ping

More information

Statistical Inference

Statistical Inference Chapter 14 Confidence Intervals: The Basic Statistical Inference Situation: We are interested in estimating some parameter (population mean, μ) that is unknown. We take a random sample from this population.

More information

On the Approximate Linear Programming Approach for Network Revenue Management Problems

On the Approximate Linear Programming Approach for Network Revenue Management Problems On the Approximate Linear Programming Approach for Network Revenue Management Problems Chaoxu Tong School of Operations Research and Information Engineering, Cornell University, Ithaca, New York 14853,

More information

1.225 Transportation Flow Systems Quiz (December 17, 2001; Duration: 3 hours)

1.225 Transportation Flow Systems Quiz (December 17, 2001; Duration: 3 hours) 1.225 Transportation Flow Systems Quiz (December 17, 2001; Duration: 3 hours) Student Name: Alias: Instructions: 1. This exam is open-book 2. No cooperation is permitted 3. Please write down your name

More information

Lecture 8 Network Optimization Algorithms

Lecture 8 Network Optimization Algorithms Advanced Algorithms Floriano Zini Free University of Bozen-Bolzano Faculty of Computer Science Academic Year 2013-2014 Lecture 8 Network Optimization Algorithms 1 21/01/14 Introduction Network models have

More information

Figure 8.2a Variation of suburban character, transit access and pedestrian accessibility by TAZ label in the study area

Figure 8.2a Variation of suburban character, transit access and pedestrian accessibility by TAZ label in the study area Figure 8.2a Variation of suburban character, transit access and pedestrian accessibility by TAZ label in the study area Figure 8.2b Variation of suburban character, commercial residential balance and mix

More information

Line search methods with variable sample size for unconstrained optimization

Line search methods with variable sample size for unconstrained optimization Line search methods with variable sample size for unconstrained optimization Nataša Krejić Nataša Krklec June 27, 2011 Abstract Minimization of unconstrained objective function in the form of mathematical

More information

Integrated Network Design and Scheduling Problems with Parallel Identical Machines: Complexity Results and Dispatching Rules

Integrated Network Design and Scheduling Problems with Parallel Identical Machines: Complexity Results and Dispatching Rules Integrated Network Design and Scheduling Problems with Parallel Identical Machines: Complexity Results and Dispatching Rules Sarah G. Nurre 1 and Thomas C. Sharkey 1 1 Department of Industrial and Systems

More information

Chapter 1. Trip Distribution. 1.1 Overview. 1.2 Definitions and notations Trip matrix

Chapter 1. Trip Distribution. 1.1 Overview. 1.2 Definitions and notations Trip matrix Chapter 1 Trip Distribution 1.1 Overview The decision to travel for a given purpose is called trip generation. These generated trips from each zone is then distributed to all other zones based on the choice

More information

Route choice models: Introduction and recent developments

Route choice models: Introduction and recent developments Route choice models: Introduction and recent developments Michel Bierlaire transp-or.epfl.ch Transport and Mobility Laboratory, EPFL Route choice models: Introduction and recent developments p.1/40 Route

More information

Optimal Adaptive Departure Time Choices with Real-Time Traveler Information Considering Arrival Reliability

Optimal Adaptive Departure Time Choices with Real-Time Traveler Information Considering Arrival Reliability University of Massachusetts Amherst ScholarWorks@UMass Amherst Masters Theses 1911 - February 2014 2009 Optimal Adaptive Departure Time Choices with Real-Time Traveler Information Considering Arrival Reliability

More information

Inventory optimization of distribution networks with discrete-event processes by vendor-managed policies

Inventory optimization of distribution networks with discrete-event processes by vendor-managed policies Inventory optimization of distribution networks with discrete-event processes by vendor-managed policies Simona Sacone and Silvia Siri Department of Communications, Computer and Systems Science University

More information

Preferred citation style. Schüssler, N. (2009) Challenges of route choice models derived from GPS, Workshop on Discrete Choice Models, EPF

Preferred citation style. Schüssler, N. (2009) Challenges of route choice models derived from GPS, Workshop on Discrete Choice Models, EPF Preferred citation style Schüssler, N. (2009) Challenges of route choice models derived from GPS, Workshop on Discrete Choice Models, EPF Lausanne, Lausanne, August 2009. Challenges of route choice models

More information

Departure time choice equilibrium problem with partial implementation of congestion pricing

Departure time choice equilibrium problem with partial implementation of congestion pricing Departure time choice equilibrium problem with partial implementation of congestion pricing Tokyo Institute of Technology Postdoctoral researcher Katsuya Sakai 1 Contents 1. Introduction 2. Method/Tool

More information

Lecture 1. Behavioral Models Multinomial Logit: Power and limitations. Cinzia Cirillo

Lecture 1. Behavioral Models Multinomial Logit: Power and limitations. Cinzia Cirillo Lecture 1 Behavioral Models Multinomial Logit: Power and limitations Cinzia Cirillo 1 Overview 1. Choice Probabilities 2. Power and Limitations of Logit 1. Taste variation 2. Substitution patterns 3. Repeated

More information

Extracting mobility behavior from cell phone data DATA SIM Summer School 2013

Extracting mobility behavior from cell phone data DATA SIM Summer School 2013 Extracting mobility behavior from cell phone data DATA SIM Summer School 2013 PETER WIDHALM Mobility Department Dynamic Transportation Systems T +43(0) 50550-6655 F +43(0) 50550-6439 peter.widhalm@ait.ac.at

More information

Appendix A.0: Approximating other performance measures

Appendix A.0: Approximating other performance measures Appendix A.0: Approximating other performance measures Alternative definition of service level and approximation. The waiting time is defined as the minimum of virtual waiting time and patience. Define

More information

A Polynomial-Time Algorithm to Find Shortest Paths with Recourse

A Polynomial-Time Algorithm to Find Shortest Paths with Recourse A Polynomial-Time Algorithm to Find Shortest Paths with Recourse J. Scott Provan Department of Operations Research University of North Carolina Chapel Hill, NC 7599-380 December, 00 Abstract The Shortest

More information

The design of Demand-Adaptive public transportation Systems: Meta-Schedules

The design of Demand-Adaptive public transportation Systems: Meta-Schedules The design of Demand-Adaptive public transportation Systems: Meta-Schedules Gabriel Teodor Crainic Fausto Errico ESG, UQAM and CIRRELT, Montreal Federico Malucelli DEI, Politecnico di Milano Maddalena

More information

Trip Distribution Modeling Milos N. Mladenovic Assistant Professor Department of Built Environment

Trip Distribution Modeling Milos N. Mladenovic Assistant Professor Department of Built Environment Trip Distribution Modeling Milos N. Mladenovic Assistant Professor Department of Built Environment 25.04.2017 Course Outline Forecasting overview and data management Trip generation modeling Trip distribution

More information

WAITING-TIME DISTRIBUTION FOR THE r th OCCURRENCE OF A COMPOUND PATTERN IN HIGHER-ORDER MARKOVIAN SEQUENCES

WAITING-TIME DISTRIBUTION FOR THE r th OCCURRENCE OF A COMPOUND PATTERN IN HIGHER-ORDER MARKOVIAN SEQUENCES WAITING-TIME DISTRIBUTION FOR THE r th OCCURRENCE OF A COMPOUND PATTERN IN HIGHER-ORDER MARKOVIAN SEQUENCES Donald E. K. Martin 1 and John A. D. Aston 2 1 Mathematics Department, Howard University, Washington,

More information

RANDOM SIMULATIONS OF BRAESS S PARADOX

RANDOM SIMULATIONS OF BRAESS S PARADOX RANDOM SIMULATIONS OF BRAESS S PARADOX PETER CHOTRAS APPROVED: Dr. Dieter Armbruster, Director........................................................ Dr. Nicolas Lanchier, Second Committee Member......................................

More information

On the control of highly congested urban networks with intricate traffic patterns: a New York City case study

On the control of highly congested urban networks with intricate traffic patterns: a New York City case study On the control of highly congested urban networks with intricate traffic patterns: a New York City case study Carolina Osorio Xiao Chen Jingqin Gao Mohamad Talas Michael Marsico 1 Introduction This paper

More information

ANALYSING ROUTE CHOICE DECISIONS ON METRO NETWORKS

ANALYSING ROUTE CHOICE DECISIONS ON METRO NETWORKS ANALYSING ROUTE CHOICE DECISIONS ON METRO NETWORKS ABSTRACT Sebastián Raveau Department of Transport Engineering and Logistics Pontificia Universidad Católica de Chile Avenida Vicuña Mackenna 4860, Santiago,

More information

Dynamic Call Center Routing Policies Using Call Waiting and Agent Idle Times Online Supplement

Dynamic Call Center Routing Policies Using Call Waiting and Agent Idle Times Online Supplement Submitted to imanufacturing & Service Operations Management manuscript MSOM-11-370.R3 Dynamic Call Center Routing Policies Using Call Waiting and Agent Idle Times Online Supplement (Authors names blinded

More information

MASSACHUSETTS INSTITUTE OF TECHNOLOGY Department of Electrical Engineering and Computer Science

MASSACHUSETTS INSTITUTE OF TECHNOLOGY Department of Electrical Engineering and Computer Science MASSACHUSETTS INSTITUTE OF TECHNOLOGY Department of Electrical Engineering and Computer Science 6.262 Discrete Stochastic Processes Midterm Quiz April 6, 2010 There are 5 questions, each with several parts.

More information

TRANSMISSION STRATEGIES FOR SINGLE-DESTINATION WIRELESS NETWORKS

TRANSMISSION STRATEGIES FOR SINGLE-DESTINATION WIRELESS NETWORKS The 20 Military Communications Conference - Track - Waveforms and Signal Processing TRANSMISSION STRATEGIES FOR SINGLE-DESTINATION WIRELESS NETWORKS Gam D. Nguyen, Jeffrey E. Wieselthier 2, Sastry Kompella,

More information

Note special lecture series by Emmanuel Candes on compressed sensing Monday and Tuesday 4-5 PM (room information on rpinfo)

Note special lecture series by Emmanuel Candes on compressed sensing Monday and Tuesday 4-5 PM (room information on rpinfo) Formulation of Finite State Markov Chains Friday, September 23, 2011 2:04 PM Note special lecture series by Emmanuel Candes on compressed sensing Monday and Tuesday 4-5 PM (room information on rpinfo)

More information

How to Estimate, Take Into Account, and Improve Travel Time Reliability in Transportation Networks

How to Estimate, Take Into Account, and Improve Travel Time Reliability in Transportation Networks University of Texas at El Paso DigitalCommons@UTEP Departmental Technical Reports (CS) Department of Computer Science 11-1-2007 How to Estimate, Take Into Account, and Improve Travel Time Reliability in

More information

Statistics Boot Camp. Dr. Stephanie Lane Institute for Defense Analyses DATAWorks 2018

Statistics Boot Camp. Dr. Stephanie Lane Institute for Defense Analyses DATAWorks 2018 Statistics Boot Camp Dr. Stephanie Lane Institute for Defense Analyses DATAWorks 2018 March 21, 2018 Outline of boot camp Summarizing and simplifying data Point and interval estimation Foundations of statistical

More information

Minimizing Total Delay in Fixed-Time Controlled Traffic Networks

Minimizing Total Delay in Fixed-Time Controlled Traffic Networks Minimizing Total Delay in Fixed-Time Controlled Traffic Networks Ekkehard Köhler, Rolf H. Möhring, and Gregor Wünsch Technische Universität Berlin, Institut für Mathematik, MA 6-1, Straße des 17. Juni

More information

Pareto-Improving Congestion Pricing on General Transportation Networks

Pareto-Improving Congestion Pricing on General Transportation Networks Transportation Seminar at University of South Florida, 02/06/2009 Pareto-Improving Congestion Pricing on General Transportation Netorks Yafeng Yin Transportation Research Center Department of Civil and

More information

Chapter 1 Statistical Inference

Chapter 1 Statistical Inference Chapter 1 Statistical Inference causal inference To infer causality, you need a randomized experiment (or a huge observational study and lots of outside information). inference to populations Generalizations

More information

Capacitor Placement for Economical Electrical Systems using Ant Colony Search Algorithm

Capacitor Placement for Economical Electrical Systems using Ant Colony Search Algorithm Capacitor Placement for Economical Electrical Systems using Ant Colony Search Algorithm Bharat Solanki Abstract The optimal capacitor placement problem involves determination of the location, number, type

More information

Homework 5 ADMM, Primal-dual interior point Dual Theory, Dual ascent

Homework 5 ADMM, Primal-dual interior point Dual Theory, Dual ascent Homework 5 ADMM, Primal-dual interior point Dual Theory, Dual ascent CMU 10-725/36-725: Convex Optimization (Fall 2017) OUT: Nov 4 DUE: Nov 18, 11:59 PM START HERE: Instructions Collaboration policy: Collaboration

More information

A STAFFING ALGORITHM FOR CALL CENTERS WITH SKILL-BASED ROUTING: SUPPLEMENTARY MATERIAL

A STAFFING ALGORITHM FOR CALL CENTERS WITH SKILL-BASED ROUTING: SUPPLEMENTARY MATERIAL A STAFFING ALGORITHM FOR CALL CENTERS WITH SKILL-BASED ROUTING: SUPPLEMENTARY MATERIAL by Rodney B. Wallace IBM and The George Washington University rodney.wallace@us.ibm.com Ward Whitt Columbia University

More information

Optima and Equilibria for Traffic Flow on Networks with Backward Propagating Queues

Optima and Equilibria for Traffic Flow on Networks with Backward Propagating Queues Optima and Equilibria for Traffic Flow on Networks with Backward Propagating Queues Alberto Bressan and Khai T Nguyen Department of Mathematics, Penn State University University Park, PA 16802, USA e-mails:

More information

1 [15 points] Search Strategies

1 [15 points] Search Strategies Probabilistic Foundations of Artificial Intelligence Final Exam Date: 29 January 2013 Time limit: 120 minutes Number of pages: 12 You can use the back of the pages if you run out of space. strictly forbidden.

More information

Decision Theory: Markov Decision Processes

Decision Theory: Markov Decision Processes Decision Theory: Markov Decision Processes CPSC 322 Lecture 33 March 31, 2006 Textbook 12.5 Decision Theory: Markov Decision Processes CPSC 322 Lecture 33, Slide 1 Lecture Overview Recap Rewards and Policies

More information

Computer Algorithms in Systems Engineering Spring 2010 Problem Set 8: Nonlinear optimization: Bus system design Due: 12 noon, Wednesday, May 12, 2010

Computer Algorithms in Systems Engineering Spring 2010 Problem Set 8: Nonlinear optimization: Bus system design Due: 12 noon, Wednesday, May 12, 2010 Computer Algorithms in Systems Engineering Spring 2010 Problem Set 8: Nonlinear optimization: Bus system design Due: 12 noon, Wednesday, May 12, 2010 Problem statement The analytical equations in lecture

More information

Urban Link Travel Time Estimation Using Large-scale Taxi Data with Partial Information

Urban Link Travel Time Estimation Using Large-scale Taxi Data with Partial Information Urban Link Travel Time Estimation Using Large-scale Taxi Data with Partial Information * Satish V. Ukkusuri * * Civil Engineering, Purdue University 24/04/2014 Outline Introduction Study Region Link Travel

More information

Technical Note: Capacitated Assortment Optimization under the Multinomial Logit Model with Nested Consideration Sets

Technical Note: Capacitated Assortment Optimization under the Multinomial Logit Model with Nested Consideration Sets Technical Note: Capacitated Assortment Optimization under the Multinomial Logit Model with Nested Consideration Sets Jacob Feldman Olin Business School, Washington University, St. Louis, MO 63130, USA

More information

CHAPTER 5 DELAY ESTIMATION FOR OVERSATURATED SIGNALIZED APPROACHES

CHAPTER 5 DELAY ESTIMATION FOR OVERSATURATED SIGNALIZED APPROACHES CHAPTER 5 DELAY ESTIMATION FOR OVERSATURATED SIGNALIZED APPROACHES Delay is an important measure of effectiveness in traffic studies, as it presents the direct cost of fuel consumption and indirect cost

More information

Delay management with capacity considerations.

Delay management with capacity considerations. Bachelor Thesis Econometrics Delay management with capacity considerations. Should a train wait for transferring passengers or not, and which train goes first? 348882 1 Content Chapter 1: Introduction...

More information

MASSACHUSETTS INSTITUTE OF TECHNOLOGY Department of Electrical Engineering and Computer Science

MASSACHUSETTS INSTITUTE OF TECHNOLOGY Department of Electrical Engineering and Computer Science MASSACHUSETTS INSTITUTE OF TECHNOLOGY Department of Electrical Engineering and Computer Science 6.262 Discrete Stochastic Processes Midterm Quiz April 6, 2010 There are 5 questions, each with several parts.

More information

Dynamic Call Center Routing Policies Using Call Waiting and Agent Idle Times Online Supplement

Dynamic Call Center Routing Policies Using Call Waiting and Agent Idle Times Online Supplement Dynamic Call Center Routing Policies Using Call Waiting and Agent Idle Times Online Supplement Wyean Chan DIRO, Université de Montréal, C.P. 6128, Succ. Centre-Ville, Montréal (Québec), H3C 3J7, CANADA,

More information

Recap. Probability, stochastic processes, Markov chains. ELEC-C7210 Modeling and analysis of communication networks

Recap. Probability, stochastic processes, Markov chains. ELEC-C7210 Modeling and analysis of communication networks Recap Probability, stochastic processes, Markov chains ELEC-C7210 Modeling and analysis of communication networks 1 Recap: Probability theory important distributions Discrete distributions Geometric distribution

More information

Simulation. Where real stuff starts

Simulation. Where real stuff starts 1 Simulation Where real stuff starts ToC 1. What is a simulation? 2. Accuracy of output 3. Random Number Generators 4. How to sample 5. Monte Carlo 6. Bootstrap 2 1. What is a simulation? 3 What is a simulation?

More information

Algorithms for a Special Class of State-Dependent Shortest Path Problems with an Application to the Train Routing Problem

Algorithms for a Special Class of State-Dependent Shortest Path Problems with an Application to the Train Routing Problem Algorithms fo Special Class of State-Dependent Shortest Path Problems with an Application to the Train Routing Problem Lunce Fu and Maged Dessouky Daniel J. Epstein Department of Industrial & Systems Engineering

More information

THE ROYAL STATISTICAL SOCIETY 2009 EXAMINATIONS SOLUTIONS GRADUATE DIPLOMA MODULAR FORMAT MODULE 3 STOCHASTIC PROCESSES AND TIME SERIES

THE ROYAL STATISTICAL SOCIETY 2009 EXAMINATIONS SOLUTIONS GRADUATE DIPLOMA MODULAR FORMAT MODULE 3 STOCHASTIC PROCESSES AND TIME SERIES THE ROYAL STATISTICAL SOCIETY 9 EXAMINATIONS SOLUTIONS GRADUATE DIPLOMA MODULAR FORMAT MODULE 3 STOCHASTIC PROCESSES AND TIME SERIES The Society provides these solutions to assist candidates preparing

More information

Notes on Dantzig-Wolfe decomposition and column generation

Notes on Dantzig-Wolfe decomposition and column generation Notes on Dantzig-Wolfe decomposition and column generation Mette Gamst November 11, 2010 1 Introduction This note introduces an exact solution method for mathematical programming problems. The method is

More information

EEC 686/785 Modeling & Performance Evaluation of Computer Systems. Lecture 19

EEC 686/785 Modeling & Performance Evaluation of Computer Systems. Lecture 19 EEC 686/785 Modeling & Performance Evaluation of Computer Systems Lecture 19 Department of Electrical and Computer Engineering Cleveland State University wenbing@ieee.org (based on Dr. Raj Jain s lecture

More information

Moment-based Availability Prediction for Bike-Sharing Systems

Moment-based Availability Prediction for Bike-Sharing Systems Moment-based Availability Prediction for Bike-Sharing Systems Jane Hillston Joint work with Cheng Feng and Daniël Reijsbergen LFCS, School of Informatics, University of Edinburgh http://www.quanticol.eu

More information

APPENDIX IV MODELLING

APPENDIX IV MODELLING APPENDIX IV MODELLING Kingston Transportation Master Plan Final Report, July 2004 Appendix IV: Modelling i TABLE OF CONTENTS Page 1.0 INTRODUCTION... 1 2.0 OBJECTIVE... 1 3.0 URBAN TRANSPORTATION MODELLING

More information

Formulas to Estimate the VRP Average Distance Traveled in Elliptic Zones

Formulas to Estimate the VRP Average Distance Traveled in Elliptic Zones F. Robuste, M. Estrada & A. Lopez-Pita 1 Formulas to Estimate the VRP Average Distance Traveled in Elliptic Zones F. Robuste, M. Estrada & A. Lopez Pita CENIT Center for Innovation in Transport Technical

More information

Economics 701 Advanced Macroeconomics I Project 1 Professor Sanjay Chugh Fall 2011

Economics 701 Advanced Macroeconomics I Project 1 Professor Sanjay Chugh Fall 2011 Department of Economics University of Maryland Economics 701 Advanced Macroeconomics I Project 1 Professor Sanjay Chugh Fall 2011 Objective As a stepping stone to learning how to work with and computationally

More information

An Optimization-Based Heuristic for the Split Delivery Vehicle Routing Problem

An Optimization-Based Heuristic for the Split Delivery Vehicle Routing Problem An Optimization-Based Heuristic for the Split Delivery Vehicle Routing Problem Claudia Archetti (1) Martin W.P. Savelsbergh (2) M. Grazia Speranza (1) (1) University of Brescia, Department of Quantitative

More information

The common-line problem in congested transit networks

The common-line problem in congested transit networks The common-line problem in congested transit networks R. Cominetti, J. Correa Abstract We analyze a general (Wardrop) equilibrium model for the common-line problem in transit networks under congestion

More information

Traffic Demand Forecast

Traffic Demand Forecast Chapter 5 Traffic Demand Forecast One of the important objectives of traffic demand forecast in a transportation master plan study is to examine the concepts and policies in proposed plans by numerically

More information

Conservation laws and some applications to traffic flows

Conservation laws and some applications to traffic flows Conservation laws and some applications to traffic flows Khai T. Nguyen Department of Mathematics, Penn State University ktn2@psu.edu 46th Annual John H. Barrett Memorial Lectures May 16 18, 2016 Khai

More information

Managing Call Centers with Many Strategic Agents

Managing Call Centers with Many Strategic Agents Managing Call Centers with Many Strategic Agents Rouba Ibrahim, Kenan Arifoglu Management Science and Innovation, UCL YEQT Workshop - Eindhoven November 2014 Work Schedule Flexibility 86SwofwthewqBestwCompanieswtowWorkwForqwofferw

More information

Computational statistics

Computational statistics Computational statistics Combinatorial optimization Thierry Denœux February 2017 Thierry Denœux Computational statistics February 2017 1 / 37 Combinatorial optimization Assume we seek the maximum of f

More information

Routing. Topics: 6.976/ESD.937 1

Routing. Topics: 6.976/ESD.937 1 Routing Topics: Definition Architecture for routing data plane algorithm Current routing algorithm control plane algorithm Optimal routing algorithm known algorithms and implementation issues new solution

More information

Distributed Optimization. Song Chong EE, KAIST

Distributed Optimization. Song Chong EE, KAIST Distributed Optimization Song Chong EE, KAIST songchong@kaist.edu Dynamic Programming for Path Planning A path-planning problem consists of a weighted directed graph with a set of n nodes N, directed links

More information

Machine Learning, Fall 2012 Homework 2

Machine Learning, Fall 2012 Homework 2 0-60 Machine Learning, Fall 202 Homework 2 Instructors: Tom Mitchell, Ziv Bar-Joseph TA in charge: Selen Uguroglu email: sugurogl@cs.cmu.edu SOLUTIONS Naive Bayes, 20 points Problem. Basic concepts, 0

More information

Inferring Passenger Boarding and Alighting Preference for the Marguerite Shuttle Bus System

Inferring Passenger Boarding and Alighting Preference for the Marguerite Shuttle Bus System Inferring Passenger Boarding and Alighting Preference for the Marguerite Shuttle Bus System Adrian Albert Abstract We analyze passenger count data from the Marguerite Shuttle system operating on the Stanford

More information

Markov Chain Monte Carlo The Metropolis-Hastings Algorithm

Markov Chain Monte Carlo The Metropolis-Hastings Algorithm Markov Chain Monte Carlo The Metropolis-Hastings Algorithm Anthony Trubiano April 11th, 2018 1 Introduction Markov Chain Monte Carlo (MCMC) methods are a class of algorithms for sampling from a probability

More information

Encapsulating Urban Traffic Rhythms into Road Networks

Encapsulating Urban Traffic Rhythms into Road Networks Encapsulating Urban Traffic Rhythms into Road Networks Junjie Wang +, Dong Wei +, Kun He, Hang Gong, Pu Wang * School of Traffic and Transportation Engineering, Central South University, Changsha, Hunan,

More information

Optimal Adaptive Routing and Traffic Assignment in Stochastic Time-Dependent Networks. Song Gao

Optimal Adaptive Routing and Traffic Assignment in Stochastic Time-Dependent Networks. Song Gao Optimal Adaptive Routing and Traffic Assignment in Stochastic Time-Dependent Networks by Song Gao B.S., Civil Engineering, Tsinghua University, China (1999) M.S., Transportation, MIT (22) Submitted to

More information

CS261: A Second Course in Algorithms Lecture #12: Applications of Multiplicative Weights to Games and Linear Programs

CS261: A Second Course in Algorithms Lecture #12: Applications of Multiplicative Weights to Games and Linear Programs CS26: A Second Course in Algorithms Lecture #2: Applications of Multiplicative Weights to Games and Linear Programs Tim Roughgarden February, 206 Extensions of the Multiplicative Weights Guarantee Last

More information

The Multi-Agent Rendezvous Problem - The Asynchronous Case

The Multi-Agent Rendezvous Problem - The Asynchronous Case 43rd IEEE Conference on Decision and Control December 14-17, 2004 Atlantis, Paradise Island, Bahamas WeB03.3 The Multi-Agent Rendezvous Problem - The Asynchronous Case J. Lin and A.S. Morse Yale University

More information

Learning Bayesian Networks for Biomedical Data

Learning Bayesian Networks for Biomedical Data Learning Bayesian Networks for Biomedical Data Faming Liang (Texas A&M University ) Liang, F. and Zhang, J. (2009) Learning Bayesian Networks for Discrete Data. Computational Statistics and Data Analysis,

More information

Network Analysis with ArcGIS Online. Deelesh Mandloi Dmitry Kudinov

Network Analysis with ArcGIS Online. Deelesh Mandloi Dmitry Kudinov Deelesh Mandloi Dmitry Kudinov Introductions Who are we? - Network Analyst Product Engineers Who are you? - Network Analyst users? - ArcGIS Online users? - Trying to figure out what is ArcGIS Online? Slides

More information

MnDOT Method for Calculating Measures of Effectiveness (MOE) From CORSIM Model Output

MnDOT Method for Calculating Measures of Effectiveness (MOE) From CORSIM Model Output MnDOT Method for Calculating Measures of Effectiveness (MOE) From CORSIM Model Output Rev. April 29, 2005 MnDOT Method for Calculating Measures of Effectiveness (MOE) From CORSIM Model Output Table of

More information

On the Partitioning of Servers in Queueing Systems during Rush Hour

On the Partitioning of Servers in Queueing Systems during Rush Hour On the Partitioning of Servers in Queueing Systems during Rush Hour This paper is motivated by two phenomena observed in many queueing systems in practice. The first is the partitioning of server capacity

More information

Typical information required from the data collection can be grouped into four categories, enumerated as below.

Typical information required from the data collection can be grouped into four categories, enumerated as below. Chapter 6 Data Collection 6.1 Overview The four-stage modeling, an important tool for forecasting future demand and performance of a transportation system, was developed for evaluating large-scale infrastructure

More information

Efficient MCMC Samplers for Network Tomography

Efficient MCMC Samplers for Network Tomography Efficient MCMC Samplers for Network Tomography Martin Hazelton 1 Institute of Fundamental Sciences Massey University 7 December 2015 1 Email: m.hazelton@massey.ac.nz AUT Mathematical Sciences Symposium

More information

ArcGIS Online Routing and Network Analysis. Deelesh Mandloi Matt Crowder

ArcGIS Online Routing and Network Analysis. Deelesh Mandloi Matt Crowder ArcGIS Online Routing and Network Analysis Deelesh Mandloi Matt Crowder Introductions Who are we? - Members of the Network Analyst development team Who are you? - Network Analyst users? - ArcGIS Online

More information

Allocation of Transportation Resources. Presented by: Anteneh Yohannes

Allocation of Transportation Resources. Presented by: Anteneh Yohannes Allocation of Transportation Resources Presented by: Anteneh Yohannes Problem State DOTs must allocate a budget to given projects Budget is often limited Social Welfare Benefits Different Viewpoints (Two

More information

CPSC 531: System Modeling and Simulation. Carey Williamson Department of Computer Science University of Calgary Fall 2017

CPSC 531: System Modeling and Simulation. Carey Williamson Department of Computer Science University of Calgary Fall 2017 CPSC 531: System Modeling and Simulation Carey Williamson Department of Computer Science University of Calgary Fall 2017 Quote of the Day A person with one watch knows what time it is. A person with two

More information

On the Partitioning of Servers in Queueing Systems during Rush Hour

On the Partitioning of Servers in Queueing Systems during Rush Hour On the Partitioning of Servers in Queueing Systems during Rush Hour Bin Hu Saif Benjaafar Department of Operations and Management Science, Ross School of Business, University of Michigan at Ann Arbor,

More information

arxiv: v1 [cs.gt] 16 Jul 2012

arxiv: v1 [cs.gt] 16 Jul 2012 Road Pricing for Spreading Peak Travel: Modeling and Design arxiv:128.4589v1 [cs.gt] 16 Jul 212 Tichakorn Wongpiromsarn *, Nan Xiao, Keyou You, Kai Sim, Lihua Xie, Emilio Frazzoli, and Daniela Rus * Singapore-MIT

More information

Improving Search Space Exploration and Exploitation with the Cross-Entropy Method and the Evolutionary Particle Swarm Optimization

Improving Search Space Exploration and Exploitation with the Cross-Entropy Method and the Evolutionary Particle Swarm Optimization 1 Improving Search Space Exploration and Exploitation with the Cross-Entropy Method and the Evolutionary Particle Swarm Optimization Leonel Carvalho, Vladimiro Miranda, Armando Leite da Silva, Carolina

More information