MaaS meets Travel Demand Modeling This car is the I.D. Concept, unveiled by VW at this year’s Paris Motor Show It is all-electric with a range of 600km and has a retractable steering wheel for AV mode VW claim that this car will be released in 2020 MaaS meets Travel Demand Modeling Klaus Noekel Michael Oliver ptvgroup.com

How does mobility as a service affect your city? How many vehicles left on the road? How to co-ordinate MaaS with transit? How much parking will be freed up? How best to regulate MaaS companies? What additional infrastructure for pick- up/drop/off? Can the city profitably run its own mobility service? Can you forecast a ‘most-likely’ future that includes vehicle and ride-sharing? Do you know what travel behavior changes vehicle and ride-sharing will bring? Does your transit plan include autonomous on-demand buses? Have you considered how to integrate vehicle and ride-sharing in to your transit plan? Is your scheme still profitable under these conditions? Do you know how to simulate mixed AV and regular traffic? Are you confident that your planned transit line will be profitable in 20 years’ time? Have you considered the reduced parking need in your land-use forecast? Will the introduction of shared, autonomous and electric fleets allow my city to meet its safety and environmental objectives? Do we have sufficient charging infrastructure? CITY HALL www.ptvgroup.com Page 2

Fitting maas into the model architecture Trip Generation Destination choice Mode choice Car Assignment Bike Public Transport Skims

Fitting MAAS into the model architecture Trip Generation Destination choice Mode choice Car Assignment Bike Public Transport MaaS 1 MaaS 2 Skims

Flavors of shared mobility systems General principle Alternative forms of mobility that do not require exclusive access (or exclusive ownership) of a means of transport Vehicle sharing (cars, bikes) One vehicle is shared sequentially by several travellers. Each traveller has exclusive use of the vehicle for a certain time. Ride pooling One vehicle is shared simultaneously by several travellers. Travellers travel together in one vehicle.

Ridesharing: model input DATA INPUT BASE = STANDARD TDM Street network Transit networks Key city hubs and interchanges City travel demand Typical traveler behavior, e.g. mode choice Page 6 www.ptvgroup.com

Ridesharing: model input SELECT PACKAGE 1 ($$$ - $): CAPACITY WAITING TIME DEVI TIME AREA TYPE: DROP OFF ONLY PICK UP ONLY DROP OFF & PICK UP 8 2.0 20 0.2 10 7 DATA INPUT SERVICE SPECIFICATION Pre-booking time Departure time window Max. Detour time Fare Vehicle capacity Max. fleet size Boarding/alighting time Pick-up/drop-off points Geographical coverage Wie viel Zeit? Wie griß Page 7 www.ptvgroup.com

ridesharing: assignment Model steps Generate trip requests from OD demand by spatial and temporal disaggregation Solve dial-a-ride-problem (DARP)  set of schedules for vehicles and assignment of passengers (= trip requests) to vehicles Visualize optimization result: inspect tours Extract user cost components for feedback into mode choice Calculate operating KPIs (fleet size, veh-km, empty veh-km, …) from operator perspective  economic evaluation

ridesharing : map method to software tools Model steps Generate trip requests from OD demand by spatial and temporal disaggregation Solve dial-a-ride-problem (DARP)  set of schedules for vehicles and assignment of passengers (= trip requests) to vehicles Visualize optimization result: inspect tours Extract user cost components for feedback into mode choice Calculate operating KPIs (fleet size, veh-km, empty veh-km, …) from operator perspective  economic evaluation scripted Tour planning optimizer TDM software

Ridesharing: Dynamic dial-a-ride-Problem dt = small time slice (5-15 min) Tourplan= empty for t = t_start to t_end step dt TR = all trip requests with birthtime in [t; t+dt) Freeze each tour in Tourplan until first event after t Tourplan = TourPlanOptimizer(Tourplan, TR) Import Schedule into PTV Visum for post-analysis Formulate task as a vehicle routing problem with pickup and delivery and with time windows. Solve it by very large-scale neighborhood search. Basis: R.K. Ahuja, J.B. Orlin, D. Sharma: Very large-scale neighborhood search, Intl. Trans. in Op. Research 7 (2000) 301-317

Ridesharing: Example result

Ridesharing: Example result

Ridesharing: Example result

Ridesharing: Example result

Ridesharing: model output KPI‘s Passenger perspective  Service quality Waiting time Travel time Journey time Revenue Unserved demand Max. number of other passengers in vehicle during trip Page 15 www.ptvgroup.com

Ridesharing: model output KPI‘s Operator perspective  Efficiency Required fleet size Tour plan for each vehicle Operating time (loaded, idle) Drive time Board/alight time Vehicle wait time Operating cost Revenue Cost coverage Spart Chart to show the KPIs – different scenarios that we are comparing Page 16 www.ptvgroup.com

Status & next steps Completed Tour optimization module for ride pooling KPIs from user perspective, operator perspective Next objectives Estimate and apply extended mode choice model Complete embedding in travel demand model BUT: scarce data! Best solution for now: use scenario technique, systematically vary unknown parameters, look at range of outcomes Important: try out and improve – it can‘t wait! There is a real challenge facing everyone involved in transportation that mobility is changing faster than we can plan for it Traditional tools are not sufficient, and so we need to develop new tools and fat We have a vision of how our software components can build business analysis and technology frameworks that plan, implement and control shared fleets We have solutions now that support the planning process and through collaboration with all stakeholders involved we will launch these and together we can create an equitable future

www.ptvgroup.com