1. Passenger travel behavior model in railway network simulation Ting Li Eric van Heck Peter Vervest Jasper Voskuilen Dept. Of decision and information sciences RSM Erasmus University Freek Hofker Marketing Research and Advice Netherlands Railways Fred Jansma Incontrol Enterprise Dynamics

2. Overview Introduction Modeling Approach Simulation Framework Passenger Travel Behavior Model Conclusion

3. Introduction Public transportation is crowded during certain times. Transportation planners want to diffuse the concentration of the peak period travel:  Expand tracks, build more trains.  Scheduling.  Dynamic pricing. The objective is to evaluate the differentiated pricing impact on the passenger travel behavior, and further evaluate the overall network performance.

4. Introduction What is dynamic pricing?  Flexible pricing based on customers and others circumstance. Examples:  Online auction: eBay  Amazon DVD’s sale Dynamic pricing in transportation industry:  Airline ticket price varies depend on season

5. Introduction How to evaluate the effectiveness of the dynamic pricing:  How does demand change over time?  What is the capacity utilization?  Does revenue increase or decrease?

6. Introduction Reasons to use simulation:  This is a large problem that is beyond computational ability.  Simulation model can easily accommodate changes such as additional tracks, additional trains, or different pricings. How to simulate:  Modeling the supply  Stations, trains and tracks  Modeling the demand  Customers  Modeling the interactions between supply and demand

7. Modeling Approach Modeling supply:  Create a large network that may includes infrastructures such as stations, tracks, and trains  Ability to create time table, train schedules over the railway networks Modeling demand (customer travel behavior):  Ability to model a passenger with his/her attributes.  To model the decision of a passenger on time-shifting, mode choice, departure time choice, and route choice.  Track an individual passenger’s behavior changes under dynamic pricing Interaction between supply and demand:  Impacts of supply’s changes on demand (travel behavior.)  Impacts of behaviors on the operational supply.

8. Modeling Approach The supply simulation consists of 5,000 passenger trains on a network of 325,000 km railway. The crisscross network of 100 different train lines along nearly 400 stations. Model is designed and implemented by Incontrol Enterprise Dynamics using a simulation tool called SIMONE (Simulation MOdel of NEtwork.)

9. Modeling Approach A passenger is similar to an agent:  Autonomous (operate without human intervention)  Goal-oriented (act according to specific rules to accomplished a pre-defined goal)  Asynchronous (operate independently)  Reactivity (Perceive the environment and respond to changes)  Pro-activeness (have individual internal states and goals, acts through rules to meet its goals)

10. Simulation Framework

11. Demand Simulation Traveler is modeled as an individual agent, who has certain characteristics and travels according to his activity schedule. Traveler plans his trip based on a set of available options. These options include both public and private mode of travel. Traveler can respond to policy changes (e.g., price increase at a particular hour) based on pre-defined decision rules. Traveler makes decisions based on a utility maximization choice. Utility is the combination of passenger fare, maximum willingness-to-pay, ability to reschedule

12. Dynamic Pricing Strategy Time-based pricing (time-of-day): price varies according to different time-of-day (e.g., lower price before 7:00; higher price between 7:00-9:00). Time-based pricing (day-of-week): Weekday vs. weekend; Different day-of-week price. Spatial-based pricing (regional-based): a high density area vs. a not so high density area Spatial-based pricing (route-based): price varies according to the utilization of the specific route.

13. Dynamic Pricing Strategy Service-based pricing (class-based): 1st and 2nd class. Service-based pricing (different train types): price varies for international train, intercity and stop train. Profile-based pricing (travel in group): for example, travel in group is cheaper. Or the combination of different pricing policies.

14. Performance Metrics

15. Passenger Travel Behavior Model Travel choice process is one of the key components within the model  How does a passenger schedule his/her trip?  How does a passenger re-schedule his/her trip? The process includes two steps:  Generation of choices (schedule)  Making decision to pick a “best” choice

16. Passenger Travel Behavior Model

17. Passenger's characteristics  Age  Income  Education  Car ownership

18. Passenger Travel Behavior Model Activity Schedule:  Activity location  Timing  Duration  Purpose

19. Passenger Travel Behavior Model Decision Windows:  How does a passenger react to price change  Example: time-base price policy introduces an :afternoon peak price, meaning passengers who travel between 4PM:6PM will pay more  1. Time-shifting: shift their departure time to before 4PM or after 6PM in order to avoid the "afternoon-peak price“  2. Mode-change: leave railway and change to private transport (e.g., car)  3. No change, accept the price increase  4. Stop travel: it is more likely to happen for leisure travelers)

20. Passenger Travel Behavior Model Maximum willingness to pay: How much a passenger are willing to pay  Travel purpose  Income Passenger disutility:  auxiliary attributes

21. Passenger Travel Behavior Model Passenger choice set:  Possible schedule  A number of trip options are generated based on required departure time, arrival time, origin and destination...  Previous experience:  Crowdness of the train  Comfort of the trip  Delay and safety  Product and ticket type:  Discount for future trips

22. Passenger Travel Behavior Model Passenger decision:  Passenger follows his decision rule to make choices on departure time, route, mode and fare.  The decision rule is to minimize generalized cost of travel or minimize the time of travel and number of transfers.

23. Conclusion This paper proposes a Passenger Railway Network Simulation for policy evaluation (dynamic pricing.) The focus is on the design and modeling approach of the Travel Behavior Model. The project are currently in prototyping state.

24. Questions?

25. References Passenger travel behavior model in railway network simulation by Ting Li et al, Winter Simulation Conference 2006 Online Dynamic Pricing: Efficiency, Equity and the Future of E-commerce by Robert M. Weiss and Ajay K. Mehrotra, Virginia Journal of Law and Technology, 6 VA. J.L. & TECH. 11