CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G Transport Modelling CIVL5502 Transportation Engineering The University of Western Australia Lecturer: Dr. Chao Sun 1 CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G Part 1: Transport Supply & Demand 2 CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G Why do we need transport models? • Designs are often based on traffic volumes but how do we know the future traffic, especially when you make changes to the network? • Traffic changes over time because of changes in many factors, e.g. social & economic activities (fuel price increase), individual preferences, government policy (congestion charge, parking fees), land use, real estate price • Traffic responds to changes in supply • How do we deal with all these uncertainties? 3 CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G What’s a model • Examples of a model A model is ‘a simplified representation of a part of the real world – the system of interest – which concentrates on certain elements considered important for its analysis from a particular point of view’ (Ortuzar & Willumsen 1995, p2). 4 CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G What’s a good model • Does it capture the essence of the system that you want to model? • Does it replicate the behaviour of system? • • Not everything. Only the things that you care about! Not 100% accurate – be careful of overfitting! • Because it’s a model, not the real thing! • Does it have the right level of abstraction? • Predictive power - can it predict the system’s behaviour? 5 CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G What are transport models used for? • Transport infrastructure planning • Assessment of land use development • Assessment transport policy options • Design of transport facilities 6 CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G Different Classes of Model 7 https://slideplayer.com/slide/13003088/79/images/7/Level+of+Effort+ Graphic+by+Daiheng+Ni.jpg CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G Barcelo & Perarnau 2005 https://www.researchgate.net/ publication/242154775_METHO DOLOGICAL_NOTES_ON_CO MBINING_MACRO_MESO_AN D_MICRO_MODELS_FOR_TRA NSPORTATION_ANALYSIS 8 CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G https://atap.gov.au/tools-techniques/travel9 demand-modelling/2-overview.aspx CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G Transport Demand & Supply P Transport Cost (generalised cost) Inverse relationship btw Cost & Demand Cost changes quantity demanded along the curve Shifters of demand: 1. Tastes & preferences 2. Num of customers 3. Price of related goods (substitutes & complements) 4. Income 5. Expectations Change in Demand (decrease) Ref: https://www.youtube.com/watch?v=LwLh6ax0zTE Change in Demand (increase) Demand Transport Demand (numbers, types, composition of users) Q 10 CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G • Transport Demand Transport demand is derived from social/economic activities and their spatial distributions in a region, and is characterised by the regional demographics. • The demand includes the numbers, types (i.e., passenger and freight), compositions of potential transport users, and the transport cost determines how much such potential become eventuated. Be careful of induced demand 11 CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G Generalised cost • All outlays perceived by the traveller for a given trip, which can be expressed in monetary units Most people perceive waiting & walking time is perceived much longer than the real time. What does this mean for PT? Ref: Wikipedia 12 CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G Transport Demand & Supply P Transport Cost (generalised cost) Change in Supply (decrease) Cost changes quantity supplied along the curve Positive relationship btw Cost & Supply Change in Supply (increase) Supply (Performance function) Ref: https://www.youtube.com/watch?v=ewPNugIqCUM&t=s Transport Demand Q (numbers, types, composition of users) 13 CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G Transport Demand & Supply Shifters of supply: 1.Price of resources 2.Num of producers 3.Technology 4.Taxes & Subsidies 5.Expectations Ref: https://www.youtube.com/watch?v=ewPNugIqCUM&t=s 16 CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G Transport Demand & Supply They interact through P (costs). Driven by scarcity Equilibrium Q Transport demand (numbers, types, compositions of transport users) P Transport costs Q0 Transport demand eventuated P0 Transport costs eventuated I Transport infrastructure R Transport regulation E Economic activities, including their spatial distributions D Demographics f Demand functional form g Supply functional form The numbers, types, compositions of transport users and the corresponding costs in the equilibrium state are those that are going to happen in reality (by theory). 17 https://www.youtube.com/watch?v=ewPNugIqCUM&t=s CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G Transport Demand & Supply Shifting Demand & Supply Demand curve describes how price changes demand; Supply curve describes how demand changes price. Q https://www.youtube.com/wa tch?v=V0tIOqU7mc&list=PLD5BC727C84E254E5 18 CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G Transport Demand & Supply Beware of induced demand: diverted demand & new demand https://www.youtube.com/wa tch?v=QzgviOpWi74 Q 19 CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G Travel Demand Management (TDM) Enhance the attractiveness of alternatives to SOV Please read ‘Travel Demand Management’ in course material 20 CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G Social Marginal Cost By DavidLevinson - Own work, CC BY-SA 3.0, https://commons.wikimedia.org/w/index.php?curid=5004157 • http://www.traffic-simulation.de/ 21 CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G The Perth model results The regulations can include measures such as congestion charge that the government adopts to internalise some externalities of a transport system. 22 CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G Congestion Charge Please watch: Economics of Land Transport in Singapore Managing Traffic Congestion in Singapore https://www.youtube.com/watch?v=gyJdfY4d 3eM By VK35 at English Wikipedia, CC BY 2.5, https://commons.wikimedia.org/w/index.php?curid=3797602 https://www.wbur.org/onpoint/2019/04/0 4/new-york-congestion-pricing-traffic A very interesting TED talk: Jonas Eliasson: How to solve traffic jams [pay attention to VDF in his talk] https://www.youtube.com/wa tch?v=CX_Krxq5eUI 23 CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G Brasses Paradox and Social Costs • In the 2019 class, I have played a game based on the following version of the Brasses Paradox https://www.youtube.com/watch?v=8mlH9bnvWVE . Students managed to replicate the same results. • Here’s a longer video describe the paradox https://www.youtube.com/watch?v=cALezV_Fwi0 and links it to the Nash Equilibrium, although I don’t necessarily agree with her statement of AVs will just solve the paradox. • Related problems are The Prisoner's Dilemma https://www.youtube.com/watch?v=t9Lo2fgxWHw & The Tragedy of the Commons https://www.youtube.com/watch?v=tLnA0AO2lXA • These problems tell us individuals could be entirely rational but still produce a lose-lose situation so the whole group is worse-off • Just as explained in the social marginal cost slide, the misalignment between the individual’s interests and societal interests is often the root cause of our transport problems. 25 CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G Part 2: 4-step models 26 CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G Representation of the Transport Network • Model scope can vary from multiple countries, several regions in a county, a whole city, or a local area in a city/town. • The geographic coverage and resolution need be represented accordingly so that the spatial distribution of transport demand and infrastructure can be modelled by mathematical expressions. 27 CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G Zone Centroids The whole area is divided into zones, each of which acts a single unit to generate and attract trips: • If a centroid acts as a point generating traffic, it is called a trip origin. • If a centroid acts as a point attracting traffic, it is called a trip destination. A zone is represented by its centroid. Source: Ortúzar, J. D. and Willumsen, L. G.: Modelling Transport, 3rd edition, p. 117, John Wiley & Sons, Ltd. 28 CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G Links & Nodes Links Each road section is represented a link, which has homogeneous characteristics, such as road width, pavement conditions and others. If a road sections has with heterogeneous characteristics along its length, it needs to be represented by two or more homogenous links, with node in between. Nodes When road links intersect or a road section is broken down into more links, the corresponding locations of these 29 intersections or changes are called nodes. CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G Zone Connectors A zone centroid is connected to the road network by one or more connectors. The connectors can be a real road or simply a nominal or virtual road 30 CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G A simple Perth model Image credit: Google Maps 31 CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G A simple Perth model External zone internal zone 32 CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G Simplified Transport User’s Decision Making Process • Should a journey (either a passenger trip or a freight shipment) be made or not? • Which destination should be chosen? • What transport mode should be used? • What route should be taken? • When should the journey start? 33 CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G Simplified Transport User’s Decision Making Process • In reality, these decisions could be made sequentially or simultaneously. • Or they are divided into subsets, with decisions being made simultaneously within a subset and sequentially between subsets • It depends on the specific type and purpose of a journey. • However, in most transport models these decisions are treated as a sequential process. 3 CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G 4 of 5 basic choice decisions are modelled, with departure time decision being either ignored or treated roughly by breaking a day into several modelling periods, such as morning peak, shoulder, off-peak and afternoon peak periods. 35 What’s the difference btw a journey and a trip CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G The 4-Step Model The model consists of a set of sub-models representing the 4 stages in transport decisions. Attention: you’re NOT doing this for your assignment! (Ortuzar & Willumsen 1994) 36 CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G 1. Trip-generation/attraction model This model produces the number of trips generated from an origin zone and the number of trips attracted to a destination zone: Origins could become destinations and vice versa Conservation equation to model a closed system (through the use of external zones) 38 CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G 1. Trip-generation/attraction model As indicated in Eq. 1 and 2, input into the tripgeneration/attraction model includes: • Population • Economic activities, such as retail floor areas, employments, students, GDP and so on • Car ownership • Household income • Others 39 CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G 1. Trip-generation/attraction model A linear regression model: Ti = b0 + biz1i + b2z2i + … + bkzki Ti : trips generated bk: coefficient for characteristic k, estimated from travel surveys zki: characteristic k (income, employment, number of household members, car ownership) Cross-classification models (look-up tables for trip generation rates) are also used. 40 CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G An trip attraction/generation example Why the difference? 41 CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G Trip generation manual Day care centre (7-9am) • • • Different land use types have different generate rates Heavy industries vs. light industries. Which has higher generates rates? What’s the average trip gen rates of Perth households? 42 CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G 1. Trip-generation/attraction model OD matrix Empty The output of the trip-generation/attraction model is the number of trips generated from and the number of trips to individual zones, which can be expressed by the above matrix format. Origin Generation Destination Attraction Total generation = total attraction 43 CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G 2: Trip distribution model It distributes the total number of trips generated from zone i to individual destination zone j, populating the empty cells in Table 1 into Table 2 in such way that 44 CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G 2: Trip distribution model A key issue in distributing the number of trips generated from an origin zone to individual destination zones is to choose a proper functional from which simulates: • the traveller’s destination choice behaviour or • the physical phenomenon of trip distribution over origin-destination pairs (OD pairs), which may not necessarily be based on any behavioural consideration. 45 CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G 2: Trip distribution model Why is this the highest? Ortúzar, J. D. and Willumsen, L. G.: Modelling Transport, 3rd edition, p. 173, John Wiley & Sons, Ltd. 46 CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G 2: Trip distribution (gravity) model Based on Newton's law of universal gravitation (Wikipedia By I, Dennis Nilsson, CC BY 3.0, https://commons.wikimedia.org/w/index.php?curid=3455682) Distance decay: the interaction between two locales declines as the distance between them increases (Wikipedia) but the friction function could be more complicated. Major Origins or Destinations have stronger “fields”, e.g. CBD has the largest catchment Normally non-linear Used for calibration 47 CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G 2: Trip distribution model Use all model parameters to reproduce the trip length distribution 48 CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G 2: Trip distribution model The point of lowest friction but why? After this point, F(c) declines with distance! 49 CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G 3. Modal split model It simulates the choice of transport modes by travellers, of which the modal choice behaviour has been the core for model development. Given a trip matrix, such as Table 2, and attributes or characteristics of various modes in question as input, the modal split model produces M modespecific trip matrices, each for a mode among the M modes available for travellers. 50 CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G 3. Modal split model Mathematically, this process can be expressed as follows: to develop a set of mode shares for each OD pair (Pijm). logit Modal split models with various forms, such probit, , nested logit, and HEV models have been developed over last thirty plus years. The fundamental of these models is the random utility theory pioneered by Daniel L. McFadden, who won He won the Nobel Prize in economics (year 2000) for his contribution to the development of the random utility theory for discrete choice models. 51 CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G Discrete Choice • Daniel L. McFadden (1937-) received the prize for “his development of theory and methods for analyzing discrete choice.” Before McFadden’s work, empirical economists … tended to assume that the variables … were continuous. But what if one is studying the demand for … people’s choice of travel modes for getting to work? ... In 1965, one of his graduate students at Berkeley was analyzing thesis data on the California state highway department’s choices on where to put freeways and asked for his help. Freeway placement is an example of a discrete, rather than a continuous, choice. • McFadden tested his model with data on people’s transportation choices before the Bay Area Rapid Transit (BART) system was built in the San Francisco Bay Area. While the official forecast was 15 percent, McFadden used his model to predict that only 6.3 percent of Bay Area travelers would use BART. The actual number turned out to be 6.2 percent. https://www.econlib.org/library/Enc/bios/McFadden.html 52 CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G 3. Modal split model The basic assumption of random utility theory in the context of a traveller will select a mode that generates the most utility. Utility represents a modal choices is that consumer's preferences. The problem then becomes one of developing an expression for the utility generated by various mode alternatives. Because it is unlikely that individual travellers’ utility functions can be specified with certainty, the unspecifiable portion is assumed to be random. 53 CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G 3. Modal split model Deterministic component Random component Random to the analysts but not to the travellers coefficients in the utility function (βm1, …, βmk ) can be estimated with data collected from traveller surveys 54 CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G A sample utility function Vcar = 0.25 – 1.21IVT – 2.5ACC – 0.3C/I + 1.1NCAR IVT: in-vehicle travel time ACC: access time C/I: cost/income NCAR: number of cars (per household) • Utility is derived from the characteristics of the • • alternative modes and those of the traveller Sensitivity: ACC has approximately twice the impact of a unit change on IVT and > 7 times the impact of C/I Mode/alternative specific constant represents unobserved or not modelled characteristics 55 CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G Two important characteristics of Utility • Utility combines the cost with the other features • Utility makes completely different alternatives comparable 56 CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G 3. Modal split model The probability that a traveller will choose some alternative mode m, is equal to the probability that the given alternative’s utility is greater than the utility of all other possible alternatives. The probabilistic component arises from the fact that the unspecifiable portion of the utility expression is not known to the analyst and is assumed to be a random variable. The basic probability statement is 57 CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G 3. Modal split model The function form of the choice probability depends on the distribution of the unobservable component εim. If it is assumed that εim independently (i.e., for i ≠ s) and identically follows a Gumbel distribution as shown in Figure 7, then the probability for individual i to choose mode alternative m is In general, error terms are normally distributed. However normal distributions are not bounded which creates a lot of mathematical complexities. Gumbel is a necessary simplification at that time to drive this closed form. Now we can use normal distribution because of modern computing 58 power (use Monte Carlo simulation). CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G 3. Modal split model 59 CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G Example 1 A simple work-mode-choice model is estimated from data in a small urban area to determine the proportions of individual travellers selecting various modes. The mode choices include automobile drive-alone (DL), automobile shared-ride (SR), and bus (B), and the utility functions are estimated as where cost is in dollars and time is in minutes. Between a residential area and an industrial complex, 4000 workers (generating vehicle-based trips) depart for work during the peak hour. For all workers, the cost of driving an automobile is $4.00 with a travel time of 20 minutes, and the bus fare is 50 cents with a travel time of 25 minutes. If the shared-ride option always consists of two travellers sharing costs equally, how many workers will take each mode? 60 CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G Example 1: Solution Note that the utility function coefficients logically indicate that as modal costs and travel times increase, modal utilities decline and, consequently, so do modal selection probabilities (see Eq. 15). Substitution of cost and travel time values into the utility expressions gives Mode specific Relative constants utility – the difference matters so UB can be 0 BTW, for those who drive, do you know what buses routes are 61 available to you? Perception = Reality CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G Example 1: Solution Substituting these values into Eq. 15 yields Multiplying these probabilities by 4000 (the total number of workers departing in the peak hour) gives 2564 workers driving alone, 944 workers sharing a ride, and 492 taking a bus. 62 CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G 4. Traffic Assignment Model • Number of trips between OD pairs are determined but which routes should they take? • To simulate the route choice of travellers • The end product is the “predicted” total number of vehicles on each road link of the network in question 63 CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G Deriving Vehicular OD Matrix Passenger OD Matrix (from Step 3) f – avg veh occupancy (passenger/veh) Vehicle OD Matrix A simple network CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G 14 Generic traffic assignment model CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G vehicle trips on route k between i & j Route Link Origin total vehicle trips between i&j Destination Key assumptions CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G • Trip maker chooses a route btw i & j , which is perceived to give the maximum utility • The maximisation of utility is normally associated with the shortest travel time, which depends on: • Link performance functions • Whether perfect information is assumed CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G Link performance functions • What’s the performance given the flow? • Constant link travel times • Flow has no impact so use free flow travel time • Variable link travel times VDF: Volume delay function 𝑽 𝜷 𝑻𝒍 = 𝑻𝟎 [𝟏 +∝∗ ] 𝑪 CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G Classification of assignment models CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G Perfect information • Perceived TT = actual TT • Perception = Reality Link TT CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G Perfect info & constant link TT shortest time route btw i & j if link l is on the shortest time route num of veh travelling along the shortest route btw i & j All-Or-Nothing (AON) assignment: • For an OD pair: All flows assigned to shortest path; • Others get nothing • A link might get 0 veh for one OD but might get from others CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G Perfect info & variable link TT For an OD pair: • Trips are spread over multiple routes • Each has equal TT as the result of different traffic vol assigned • Wardrop user equilibrium: The travel time between a specified origin and destination on all used routes is the same and is less than or equal to the travel time that would be experienced by a traveller on any unused route. Or: No user can save time by switching routes CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G Perfect info & variable link TT Equal TT for all used routes btw ij set of used routes btw i & j TT of route k is the sum of its link TT 0 or 1 Vol on all routes add up to total demand btw ij Vol link l = all traffic assigned to it by all OD pairs Deterministic User Equilibrium (DUE) assignment • Only assign trips to shortest (time) paths • Don’t get confused with Dynamic User Equilibrium A DUE example CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G Equal TT q < q* q > q* CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G DUE assignment R1: 10km @ 100km/hr +2 min for every 500veh 4,500 veh City Suburb R2: 5km @ 75km/hr TT increase with the square of 103 veh/hr Step 1: Determining free-flow travel times DUE assignment CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G R1: 10km @ 100km/hr +2 min for every 500veh City 4,500 veh Suburb R2: 5km @ 75km/hr TT increase with the square of the number of 103 veh/hr Step 2: link performance functions DUE assignment CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G If total traffic flow q < q* all traffic needs to be assigned to Route 2 under DUE q* can be obtained by solving the equation for route 2: q* > q so assign to both routes DUE assignment CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G DUE assignment It can also be solved using calculus! Imperfect info CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G Perception ≈ Reality Err term for route k btw i & j Independently & identically follows a Gambel distribution Model parameter Add up link TT to get route TT Allocating trips using a Logit choice model CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G Imperfect info & constant link TT Multiple-routes assignment model Typically applies to rural roads: • Most travellers chose actual shortest path but not all • When θ →∞, it converges to AON CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G Imperfect info & variable link TT Spreading of traffic due to: • Imperfect info Multiple-routes model doesn’t have these • Traffic level Stochastic User Equilibrium (SUE) • • • ∞ When θ → , it converges to DUE When traffic is light, it converges to the Multiple-Route traffic assignment model When both are true, it converges to AON CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G Classification of assignment models Zone 1 CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G 𝑶𝒊 = O1 = fo(X1) 𝑫𝒋 D1 = fd(X1) R1,3 2 Zone 2 O2 = fo (X2) D2 = fd(X2) R1,3 1 R1,3 3 O3 = fo (X3) D3 = fd(X3) Zone 3 CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G http://etc.ch/pfrH Route choice experiments 85 CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G Summary & discussions 93 CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G The four-step modelling framework (Ortuzar & Willumsen 1994) 94 CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G Use models for rational decision making (Ortuzar & Willumsen 1994) 95 CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G Modelling and systems thinking (Ortuzar & Willumsen 1994) 96 CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G Modelling and systems thinking (Ortuzar & Willumsen 1994) 97 CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G The 'land-use transport feedback cycle' This calls for integrated land use & transport models Wegener, M. 1996 https://www.researchgate.net /publication/265028805_Redu ction_of_CO2_Emissions_of_ Transport_by_Reorganisation _of_Urban_Activities/figures 98 CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G A case study First mile & last mile problem (Mungundan, Radhakrishnan, 2018: EVALUATING THE IMPACT OF INCREASING 99 ‘PARK AND RIDE’ AT PERTH TRAIN STATIONS) CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G A case study • How would increasing PnR supply change PT patronage? (Mungundan, Radhakrishnan, 2018: EVALUATING THE IMPACT OF INCREASING ‘PARK AND RIDE’ AT PERTH TRAIN STATIONS) 100 CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G The “optimal” model complexity (Ortuzar & Willumsen 1994) 101 CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G Debate about models and their “accuracy” The links below are about climate change but you can easily image people having the same argument about transport models • http://www.abc.net.au/news/2016-08- 16/professor-brian-cox-vs.-senator-malcolmroberts/7746576 Round 3 @ 2:30 @4:00 @4:41 • https://www.theguardian.com/australianews/2016/aug/16/qa-brian-cox-brings-graphsmalcolm-roberts • 102 CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G Use and Abuse of Models Inspired by the book Use and Abuse of Statistics https://www.questia.com/library/83108/use-and-abuse-of-statistics • How and in what circumstances they may be used • How they should not be used • Dangers of misinterpretation • Difficulties which beset the modellers path of investigation • Black art? CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G Is Newtonian physics correct? How about Theory of Relativity & quantum physics? All models are wrong but some are useful, so use them carefully!!! Where’s the biggest weakness of all models? 104 CIVL5502 - Transportation Engineering – UWA - CRICOS Provider Code: 00126G Weakness of models Assumptions • Why do models always have assumptions? Data: • • Your model can only be as good as your data “Garbage in garbage out” Verification & calibration