Our automated future of mobility

Challenges in prediction and control

Sebastian Hörl

Kyko 2018

Magdeburg, 26 May 2018

The "street" in 1900

The "street" today

Rochester, NY ...

... 1950 ...

... and today

The street of tomorrow?

  • Autonomous Mobility

  • Mobility as a Service

  • Mobility on Demand

  • Electrification

  • Aerial Mobility

The street of tomorrow?

  • Autonomous Mobility

  • Mobility as a Service

  • Mobility on Demand

  • Electrification

  • Aerial Mobility

How do we predict the impacts?

How do we control the system?

The impact of control?

The impact of control?

User optimum vs. system optimum

 20 min

5 min

Braess' Paradoxon

User optimum vs. system optimum

 20 min

5 min

Braess' Paradoxon

User optimum vs. system optimum

 16 min

7 min

9 min

Braess' Paradoxon

User optimum vs. system optimum

 16 min

7 min

9 min

Braess' Paradoxon

+ x EUR ?

Signal control

Flow Maximization

Flow Imbalance

Waiting times

Dedicated bus lanes

Signal control

"Green waves"

Scheduling

Mobility pricing: London

Mobility pricing: Stockholm

Problem: Real-world experiments

Solution: Simulations

MATSim

Home

Work

Shop

Home

until 8am

9am to 6pm

6:15m to 6:30pm

from 6:45pm

walk

public

transport

walk

MATSim

Network simulation

Scoring of the plans

Selection and modification

Scenario

Scenario calibration

Scenario calibration

Current research at IVT
Self-driving vehicles

Self-driving vehicles

Access to Mobility

Effective Capacities

Costs

Individualization

Empty distance

Travels

Net effect?

Self-driving vehicles

A_i = \sum_j o_j \cdot f(c_{ij})
Ai=jojf(cij)A_i = \sum_j o_j \cdot f(c_{ij})

Current research

  • Cost calculator

 

  • Model choice model

 

  • Simulation (MATSim)
    • Scenario
    • New components, integration

Cost Calculator

  • Fleets of autonomous vehicles
     
  • Private autonomous vehicles
     
  • Comparison with conventional vehicles

Cost Calculator

Cost Calculator

0.67 CHF

per passenger kilometer

Mode Choice Model

  • Stated preference survey
    • 400 respondents
       
  • Attitudes towards different AV services
    • Shared AV
    • Pooled AV
    • AV feed for public transport
       
  • Work done by Felix Becker

Mode Choice Model

P(m) = \frac{\exp(V_m)}{\sum_{m'} \exp(V_{m'})}
P(m)=exp(Vm)mexp(Vm)P(m) = \frac{\exp(V_m)}{\sum_{m'} \exp(V_{m'})}

Simulation of AVs

  • AV taxi services
     
  • Simple dispatching heuristic
    (Bischoff & Maciejewski)
     
  • Multiple operators
     
  • Dynamic demand responses

Idle

Pickup

Drive

Dropoff

Drive

Case study: dispatching strategies

  • Maximum demand case
     
  • Freeflow conditions
     
  • Four dispatching strategies
    • Load-balancing heuristic
    • Global Eucledian Bipartite Matching
    • Feedforward Fluidic Rebalancing
    • daptive Uniform Rebalancing

Case study: dispatching strategies

Case study: dispatching strategies

Case study: dispatching strategies

Case study: dispatching strategies

Case study 2: Dynamic demand simulations

Work in progress ...

  • Scenario analysis
    • Changes in accessibility
    • Induced Demand
    • Effects on public transit ridership
       
  • Missing elements

Spatial constraints

Operational constraints

Intermodality

Thanks!

Questions?

Contact: sebastian.hoerl@ivt.baug.ethz.ch

Our automated future of mobility: Challenges in prediction and control

By Sebastian Hörl

Our automated future of mobility: Challenges in prediction and control

Alumni Conference "Magdeburger Kybernetiker", 26 May 2018

  • 166
Loading comments...

More from Sebastian Hörl