Robust modelling of transportation

simulation scenarios

Sebastian Hörl

Brown Bag

ETH Zürich, IVT

13 November 2017

Research Plan

Robust scenario generation

framework

Detailed scenario for

Switzerland and Zurich

Case Studies on

Autonomous Vehicles

MATSim

Scenario

Network Simulation

Scoring

Replanning

Population

Setting up a MATSim simulation

Raw Input Data

Scenario

Simulation

Census data, count data, model parameters, map data, ...

Road network, public transport schedule, population, plans, ...

Example: Zurich    PT Mode (R5)

Example: Zurich    PT Mode (R5)

Example: Zurich    Car Mode

Example: Zurich    Car Mode

Example: Zurich    Counts?

Example: Zurich    Better counts...

Example: Zurich    Better counts... but travel times?

Before counts

After counts

Example: Zurich    Pipeline (CH)

  • Downsample: Reduce input population
  • NetworkCleaner: Clean sinks and sources
  • FixLinkIds: Assign links to activities and facilities
  • MergeSecondaryFacilities: Convert Registererhebung to MATSim
  • FixFacilityActivityTypes: Add missing activity types
  • FixShopActivities: Rename "shopping" to "shop"
  • RemoveInvalidPlans: Remove plans with obvious problems
  • ShiftTimes: Dilute departure times
  • SecondaryLocationChoice: Create plans with secondary activity locations
  • TypicalDurationForActivityTypes: Enrich population with durations
  • TrafficCountPlanSelector: Select plan to fit traffic counts
  • UnselectedPlanRemoval: Remove irrelevant plans

Example: Zurich    Pipeline (ZH/ASTRA)

  • ScenarioCutter: Cut scenario from Switzerland
  • FacilityRemover: Remove unused facilities
  • InitialRouting: Route all legs to reduce startup time
  • ConfigGenerator: Adjust config for Zurich

Robust scenario generation framework

Raw Input Data

Scenario

Simulation

Modification

Robust scenario generation framework

Raw Input Data

Scenario

Simulation

Validation

Modification

Robust scenario generation framework

Raw Input Data

Scenario

Simulation

Validation

Modification

Unit Tests

for Scenarios

Robust scenario generation framework  

Open Questions

  • Interface and library design
  • Computing infrastructure
     
  • Calibration tools
     
  • Simulation
    • Convergence
    • Speed-up
    • Proxies

Scenario for Switzerland and Zurich   

Possible / ongoing tasks

  • (Secondary) Location Choice
    • Initial population (static)
    • Replanning (dynamic)
       
  • Travel times and network dynamics
    • Traffic lights
    • Link speed models
  • Mode Choice
    • Discrete choice models
  • Departure times
    • Randomization

       
  • Anonymization

Case studies on autonomous vehicles   

Possible / existing components

  • Operational constraints
    • Refueling / recharging
    • Parking
       
  • Spatial constraints
    • Parking
    • Pick-up / drop-off process
       
  • Multimodality
    • Integration with public transport
  • Dynamic pricing
    • Based on IVT cost calculator

       
  • Fleet operation
    • Dispatching
    • Rebalancing

Thank you!   

Questions?

Brown Bag

By Sebastian Hörl

Brown Bag

13 November 2017 at IVT, ETH Zürich

  • 249
Loading comments...

More from Sebastian Hörl