STREET: Simulating Transportation for Realistic Engineering Education and Training

1. STREET: Simulating Transportation for Realistic Engineering Education and Training David Levinson April 7, 2006

2. Objectives Fellows will • learn about the strengths and weaknesses of alternative simulation tools for teaching transportation planning and engineering, • engage in active learning using simulators, • provide informed feedback about said simulators in transportation.

3. Outline

4. Ramp Merge Model by Treiber http://vwisb7.vkw.tu-dresden.de/~treiber/MicroApplet/

5. Traffic Light System by Fu-Kwun Hwang http://www.phy.ntnu.edu.tw/ntnujava/viewtopic.php?t=226

6. CA Model by Helbing http://rcswww.urz.tu-dresden.de/~helbing/RoadApplet/

7. Transims by LANL http://public.lanl.gov/bwb/Java/lanl/bwbush/transims/simulate.html

8. SONG: Simulator of Network Growth http://www.ce.umn.edu/~levinson/Song/Dynamics.html

9. Transportation Planning Model • Trip-Based Approach Focused on transportation behavior (trip generation, trip distribution, mode split, route assignment) • Four Step Method (e.g. Ortuzar and Willumsen, 2001) • Coherent instead of sequential method (Boyce 2002) • Distribution & Assignment Combined Model • Activity-Based Approach Focusing on individual decision-making • Activity Chain • Transportation Simulation • Mobility simulation (lane choice, acceleration, route choice) Micro-Simulation Model: AIMSUN, VISSIM… Macro-simulation, Queue model (Gawron, 1998) • Tactical simulation (departure time, mode, route choice) • Strategic simulation (Residence & workplace, long term) Planning Model: URBANSIM…

10. Agent-based Model • Multi-Agent Simulation (Nagel 2004) • Separating strategy from mobility simulation (distributed design) • Considering individuals • Including Travel habit (Time scope) • Simulating learning and feedback (knowledge spreading) • Computational feasible • Agent-based Demand and Assignment Model (Zhang and Levinson 2004) • Demand and assignment integrated • Heuristic learning rather than Shortest path algorithm • Individual decision-making rule • Agent & Nodes interaction (barbershop model)

11. Model Implementation • Coded in JAVA Applet • Use Sioux Falls network • Network converges within 10 iterations with low and median loads (maximum link flow change < 1%) • Finds shortest path effectively • Unstable when network is very congested

12. User Interface • Objective: • Easy to learn • Transportation phenomena • Project selection process • Network Type • Global Variables • Traffic generation • Distribution • Mode Choice • Assignment • On Screen Editing • Add and delete • Attributes modification • Information inspection • Tracing Traveler • MOE Outputs • On screen • Statistical table

13. Assignment Design • Expand network to improve network performance • Test different scenarios under budget constraints • Select network expansion projects and justify the choice • Three assignments aiming at improving students understanding of transportation planning and project evaluation

14. CE 3201 Lab 1 - Assignment 1 Assigned Week: 1/30/2006 Due Week: 2/6/2006 Objective: This assignment helps students to understand the concept of travel demand modeling using ADAM. The exercises will familiarize students with ADAM and facilitate the completion of subsequent assignments and enable better understanding transportation planning models. Instructions: The agent based travel demand model (ADAM) can be accessed online at the course website. Opening the html file you can see the graphical user interface which has several major parts: network editing, global variables, the network map, and the legend. The panel of inputs contains several choice boxes and scrollbars, which enable the selection of different networks and different values for global variables. The buttons under the editing option help you create your own network and change the properties of links and nodes. You can save and reload the network you created by clicking the corresponding option. Clicking the restore button will change all options back to their default value. The legend illustrates the meaning of the colors on the map after the model is run. You can also trace a particular vehicle. The statistics option will list all the important characteristics of the current network, including all global variables, the OD matrix generated, as well as important measures of effectiveness. You should save this result in a text file and include it as an appendix to your lab report. Tasks: 1. Test different options of the model to understand the meaning of different options. 2. Modify the Sioux Falls network and create your own network. 3. Use the evolve option and see the result generated. Choose one particular node and illustrate the relation between original workers and final trip generated. 4. Understand the meaning of OD matrix and see how it is obtained. 5. Write up results (~1 page) describing what you did and what it means. Include an appendix with statistics and picture of the network. Email your final report to Feng Xie at xiex0055@umn.edu . Assignment Week 1

15. CE 3201 Lab 1 - Assignment 2 Assigned Week: Due Week: Objective: This assignment helps students to understand the role of the global variables and the concept of elasticity. Students will alter global variables on the model and explore the different outcomes that result from the exercise. Elasticity: measures the responsiveness of one variable to changes in another variable. Specifically it is the percentage change in one variable in response to a percentage change in another. The elasticity of Y due to X can be found by calculating: %Change in Y / % Change in X = [(Y2 - Y1)/Y1]/[(X2-X1)/X1] Instructions: We use the Sioux Falls network for this exercise. The global variables that we will be altering are the following: Trip production rate Trip attraction rate Travel length coefficient Peak hour rate Auto mode share Auto occupancy We will be altering one variable at a time. Record the output statistics under the default scenario and a new scenario created by changing one of the global variables above. You will then calculate the elasticity associated with each of the above global parameters. The measures of effectiveness (MOEs) for which you will calculate elasticities with respect to the global parameters are the following: Benefit Vehicle Hours Traveled Vehicle Kilometers Traveled Number of Trips Cost These can be found in the system statistics output. Submittal: Write a short report (2-3 pages) including a table for elasticities. Your report should illustrate your understanding of the concept of elasticity. Comment on your results. You can work in groups to do the simulations but you need to submit an individual report. Email your final report to Feng Xie at xiex0055@umn.edu. Assignment Week 2

16. Objective This assignment helps students to understand the effects of road construction on traffic and how to evaluate alternatives of network construction. Students will be policy makers and decide how to adapt the traffic of a road network to the development of a city. Instructions We use a new Sioux Falls network as the background. Imagine that three new regions emerge because of the growth of the city. Each of these will introduce new travel demand as well as new opportunities. New links have been added to serve the new nodes. However, severe traffic jams occurred due to the increased travel demand on the network. The existing road network may require improvement. As a planner of the Sioux Falls City, you are supposed to propose improvement on the network in order to adapt the traffic to the development of the city. The total budget allocated for this project is $12,000,000. Your objective is to maximize the benefit/cost ratio of your proposal subject to this budget. 1. Load the updated network A network file has been sent to your umn mailbox. Load the new Sioux Falls network. That is the network you are going to work on. Compared to the original Sioux Falls network, three new nodes have been added in this network. The coordinates, demands and opportunities for these nodes are as follows: Assignment Week 3 • Node 25 represents an emerging commercial district located in the center of the city and Node 26 and Node 27 are two residential regions emerging in the suburb. Twelve links have been added to connect the new nodes to the original network. You can examine the properties of new links on the screen. • Suppose all the global variables have been calibrated by the modeler and as a planner you are NOT allowed to make any change to them.

17. 2. Expand the existing network Suppose due to various constraints, you are NOT allowed to build any new road. Instead, you can only add new lanes to the existing roads to improve the network. Note that the expansion cost per lane-kilometer is $3,000,000, including land acquisition cost and construction cost. You have a total budget of $12,000,000, which is equivalent to a maximal capacity addition of 4 lane-kilometers. You can either invest all the money in one link or allocate your money to several links. You don’t have to use up the budget but make sure you spend no more than $12,000,000; otherwise your project proposal will be surely turned down by the city government. 3. Calculate the benefit /cost ratio of your proposed expansion Evolve the model after you expand the given Sioux Falls network. Examine the benefit and cost from proposed expansion in the statistics output. 3.1 Cost Suppose the life cycle of your project is ten years. You need to calculate all the costs of your project over its entire life. These costs include the initial expansion cost, which has been output in the statistics output, and a series of annual costs, including the costs of maintenance and operation. Costs that occur at different times may be placed on a comparable basis by reducing the future amounts to their present value. According to (15.3) of the textbook (Banks, 2002, p.409), the total present value of a project can be given by where I=the initial investment CA=the annual cost S=Salvage value after N years N=the life of the project In your project of road expansion, suppose: I= the expansion cost CA=($100,000/lane.kilometer)*added lane.kilometer S=0 r=0.05 N=10 Please calculate the life-cycle cost for your proposed expansion and record it as C. 3.2 Benefit To save your time, we have calculated the life cycle benefit of proposed expansion and list it in the statistics output for you. The life cycle benefit of a proposed expansion can be calculated as: B=K* CSi Where CSi =the the change in consumer surplus on link i due to your improvements. K= a coefficient that converts the vehicle travel time saving during peak hour to the present value of life cycle benefits of proposed expansion. It is calculated for this particular project as: K=2*Auto occupancy*VOT*life cycle length*days of year. Assignment 3 Continued

18. The change in consumer surplus is equal to the area of the shaded trapezoid in the above figure. The area (CS) can be calculated as: CS = 0.5*(Q1+Q2)*(C1-C2) Where: C1 = Cost in terms of link travel time on any link i C2 = Cost in terms of link travel time on any link i on the improved network Q1 = Total number of trips passing link i on the original network Q2 = Total number of trips passing link i on the improved network We sum up the consumer surplus on all the existing roads to the total travel time (min*veh) saved on the network due to proposed expansion. 3.3 Benefit-Cost ratio Calculate the benefit-cost ratio of your proposed expansion as R=B/C 4. Repeat steps 2-3 to develop alternatives of road expansion and calculate their respective benefit-cost ratios. Tasks 1. Read Section 15.1.1 of Banks’ textbook for life cycle cost analysis; read Section 15. 1.2 for the calculation of consumer surplus and benefit-cost analysis; 2. Present at least three alternatives of road expansion on the given Sioux-Falls network and propose the project with the highest benefit-cost ratio. Develop a report of 3–5 pages with pictures and numbers explaining your proposal to the funding agency. 3. You may work in groups and turn in a report as a group. Assignment 3 Continued (3/3)

19. Survey information • Pre-assignment Survey and Post-assignment Survey • Assignment performance (not included yet) • Exam scores (not included yet)

20. Pre-assignment Survey • 1: Strongly Disagree 5: Strongly Agree

21. Post-assignment Survey • 1: Strongly Disagree 5: Strongly Agree • 0: Strongly Agree with option on left 9: Strong Agree with option on right

22. Post-assignment Survey • 1: Strongly Disagree 5: Strongly Agree

23. Cross-Analysis • 1: Strongly Disagree 5: Strongly Agree

24. Further Analysis • Incorporating scores in the assignments and exams to the analysis • Further investigate the correlation of learning style preferences and effectiveness of using ADAM • Build regression model to assess learning outcomes and teaching strategy

25. ADAM 2.0 • Less Myopic • Influencing radius • Using accessibility as Attractiveness of nodes • Imitate social network • Simulate recruiting process • New information spreading process considering social network • Test on more sophisticated network

26. Highway Design http://128.101.111.90/Lab_Mod/RoadDesign.html

27. Kronos

28. Intersection Design

29. Queueing Simulation

30. Questions and Comments?Feedback …Thank you!

31. Agents • Traveler • Status • Knowledge • Node • Travelers • Workers • Knowledge • Arc • Length • Capacity • Free Flow Speed • Costs • Flow http://www.ce.umn.edu/~levinson/Adam/Dynamics.html

32. Decision Rules • Destination Choice Rules • cost of traveling (C), opportunities on the network (B) and characteristics of travelers () • Learning rules • Experience (Omega) from traveling and from other travelers (N) • Information Storage and Spreading • Day-to-day learning • Agents Interaction rules • Job offering Vs job hunting? • Social network and activity choice

33. ADAM1.0 • Logit-like probability & BPR function • Shortest path information is pooled at nodes and spreads through travelers. • Opportunity information is only available at adjacent Nodes. • Link travel time is updated at the end of each iteration and travelers reenter the network until convergence reached. • MSA of link travel time is used between iterations.