Exploring the system impact of

automated taxis via simulation

Sebastian Hörl

17 May 2021

at CentraleSupélec

The street in 1900

http://www.loc.gov/pictures/item/2016800172/

The street today

https://commons.wikimedia.org/wiki/File:Atlanta_75.85.jpg

The street of tomorrow?

Autonomous Mobility
Mobility as a Service
Mobility on Demand
Electrification
Aerial Mobility

Julius Bär / Farner

I. MATSim

Agent-based models

Senozon VIA

https://pixabay.com/en/traffic-jam-stop-and-go-rush-hour-143391/

MATSim

Senozon VIA

Flexible, extensible and well-tested open-source transport simulation framework
Used by many research groups and companies all over the world
Extensions for parking behaviour, signal control, location choice, freight, ...

http://www.matsim.org

matsim-org/matsim-libs

MATSim

Mobility simulation

Decision-making

Analysis

Scenario

Flexible, extensible and well-tested open-source transport simulation framework
Used by many research groups and companies all over the world
Extensions for parking behaviour, signal control, location choice, freight, ...

MATSim

Senozon VIA

MATSim

Home

Work

Shop

Home

until 8am

9am to 6pm

6:15m to 6:30pm

from 6:45pm

walk

public

transport

walk

Discrete mode choice

P(k) = \frac{\exp(U_k)}{\sum_i \exp(U_i)}

Discrete mode choice

https://pixabay.com/en/zurich-historic-center-churches-933732/

II. AMoD in Zurich

What do we know about automated vehicles?

Cost structures?

User preferences?

System impact?

Cost Calculator for automated mobility

Stated preference survey

MATSim simulation

AMoD Cost Calculator

Bösch, P.M., F. Becker, H. Becker and K.W. Axhausen (2018) Cost-based analysis of autonomous mobility services, Transport Policy, 64, 76-91

AMoD Survey

Felix Becker, Institute for Transport Planning and Systems, ETH Zurich.

AMoD Survey

Felix Becker, Institute for Transport Planning and Systems, ETH Zurich.

AMoD Survey

13 CHF/h

AMoD

Taxi

19 CHF/h

Conventional

Car

12 CHF/h

Public

Transport

VTTS

Car by Adrien Coquet from the Noun Project

Bus by Simon Farkas from the Noun Project

Wait by ibrandify from the Noun Project

AMoD

AMoD Survey

Car by Adrien Coquet from the Noun Project

Bus by Simon Farkas from the Noun Project

Wait by ibrandify from the Noun Project

13 CHF/h

AMoD

Taxi

19 CHF/h

Conventional

Car

12 CHF/h

Public

Transport

VTTS

21 CHF/h

32 CHF/h

AMoD

Calibration

Fleet sizing with dynamic demand

Model structure

Cost calculator

Plan modification

Discrete Mode Choice Extension

Mobility simulation

Prediction

Price

Trips

- Utilization

- Empty distance, ...

- Travel times

- Wait times, ...

Visualisation

Automated taxi

Pickup

Dropoff

Hörl, S., F. Becker and K.W. Axhausen (2020) Automated Mobility on Demand: A comprehensive simulation study of cost, behaviour and system impact for Zurich

13% reduction in vehicles

100% increase in VKT

Fleet control

Hörl, S., C. Ruch, F. Becker, E. Frazzoli and K.W. Axhausen (2019) Fleet operational policies for automated mobility: a simulation assessment for Zurich, Transportation Research: Part C, 102, 20-32.

Fleet control

Load-balancing heuristic

Simple heuristic, fast runtime

Gloal Euclidean Bipartite Matching

Standard algorithm in Operations Research

Assignment

Car by Adrien Coquet from the Noun Project

Hail by Bradley Avison from the Noun Project

Example algorithms

Load-balancing heuristic

Undersupply

More vehicles than requests
When request pops up
find closest vehicle

Oversupply

More requests than vehicles
When vehicle gets available
find closest request

able to serve remote demand

able to serve peek demand

Example algorithms

Global Bipartite Matching

Find matching of vehicles and requests that minimizes total Euclidean (or network-based) empty distance
Nice benchmark because sole objective is to minimize empty distance

Fleet control

Feedforward Fluidic Optimal Rebalancing Policy

Linear program matching a priori known trip rates

Adaptive Uniform Rebalancing Policy

Linear program distributing vehicles uniformly

Car by Adrien Coquet from the Noun Project

Hail by Bradley Avison from the Noun Project

Redistribution

Example algorithms

Feed-forward fluidic rebalancing strategy

Travel time

Rebalancing flows

Arrival rate

Transition probability

Example algorithms

Feed-forward fluidic rebalancing strategy

Travel time

Rebalancing flows

Arrival rate

Transition probability

Informed algorithm from historical data
Minimization of rebalancing time
Minimization of waiting time?

Example algorithms

Adaptive Uniform Rebalancing Policy

Vehicles per zone

Current requests

Minimization of rebalancing time
Working on current information
Uniform distribution of vehicles

III. Paris Scenario

https://pixabay.com/en/paris-eiffel-tower-night-city-view-3296269/

Two components = eqasim

Population synthesis

Travel demand synthesis

Mobility simulation

(MATSim)

Framework to easily set up runnable MATSim (and other) simulations
Collection of open source tools and algorithms for transport simulation
Focus on process from raw data to final results for reproducible research
Increase interoperability of models

Data

Analysis

The case of Île-de-France

Census data

Récensement de la population

Icons on this and following slides: https://fontawesome.com // Background: Simunto VIA

The case of Île-de-France

Dispositif sur les revenus localisés sociaux et fiscaux
Income tax data

FiLo
SoFi

The case of Île-de-France

RP: Flux de mobilité
Commuting data

RP
Mob

FiLoSoFi

The case of Île-de-France

Enquête globale de transport
Household Travel Survey

EGT

FiLoSoFi

RP Mob

Enquête national transports et deplacements
Household Travel Survey

ENTD

The case of Île-de-France

Enquête globale de transport
Household Travel Survey

EGT

FiLoSoFi

RP Mob

Enquête national transports et deplacements
Household Travel Survey

ENTD

The case of Île-de-France

SIRENE
Enterprise census

FiLoSoFi

RP Mob

EGT

ENTD

BD-TOPO
Address database

The case of Île-de-France

FiLoSoFi

RP Mob

EGT

ENTD

SIRENE

Person ID

Age

Gender

Home (X,Y)

male

(65345, ...)

female

(65345, ...)

female

(65345, ...)

BD-TOPO

The case of Île-de-France

FiLoSoFi

RP Mob

EGT

ENTD

SIRENE

Person ID

Activity

Start

End

Loc.

523

home

08:00

(x,y)

523

work

08:55

18:12

(x,y)

523

shop

19:10

19:25

(x,y)

523

home

19:40

(x,y)

Person ID

Mode

Start

End

523

Public T.

08:55

523

Public T.

18:12

19:10

523

Walking

19:25

19:40

08:00

BD-TOPO

The case of Île-de-France

OpenStreetMap
Road network

OSM

FiLoSoFi

RP Mob

EGT

ENTD

IDFm GTFS
Public transport schedule

GTFS

SIRENE

OSM

GTFS

BD-TOPO

The case of Île-de-France

FiLoSoFi

RP Mob

EGT

ENTD

SIRENE

OSM

GTFS

BD-TOPO

The case of Île-de-France

FiLoSoFi

RP Mob

EGT

ENTD

BPE

OSM

GTFS

DCM

The case of Île-de-France

FiLoSoFi

RP Mob

EGT

ENTD

BPE

OSM

GTFS

DCM

The case of Île-de-France

FiLoSoFi

RP Mob

EGT

ENTD

BPE

OSM

GTFS

DCM

The case of Île-de-France

FiLoSoFi

RP Mob

EGT

ENTD

BPE

OSM

GTFS

DCM

The case of Île-de-France

FiLoSoFi

RP Mob

EGT

ENTD

BPE

OSM

GTFS

DCM

The case of Île-de-France

Open
Data

Open
Software

Reproducible research

Verifiable results

Integrated testing

Detailed documentation of whole open-source pipeline is under preparation

FiLoSoFi

RP Mob

EGT

ENTD

SIRENE

OSM

GTFS

BD-TOPO

IV. French use cases

The case of Île-de-France

CO2 Emmissions

The case of Île-de-France

Grand Paris Express

AMoD in Paris

Hörl, S., M. Balac and K.W. Axhausen (2019) Dynamic demand estimation for an AMoD system in Paris, paper presented at the 30th IEEE Intelligent Vehicles Symposium, June 2019, Paris, France.

Travel Behaviour

Zurich model, calibrated
for Paris population

Cost sturcture

Adapted from Berlin

AMoD Simulation

Load-balancing heuristic

AMoD in Paris

Hörl, S., M. Balac and K.W. Axhausen (2019) Dynamic demand estimation for an AMoD system in Paris, paper presented at the 30th IEEE Intelligent Vehicles Symposium, June 2019, Paris, France.

Travel Behaviour

Zurich model, calibrated
for Paris population

Cost sturcture

Adapted from Berlin

AMoD Simulation

Load-balancing heuristic

Maximum static demand: 2.3M trips

AMoD in Paris

Hörl, S., M. Balac and K.W. Axhausen (2019) Dynamic demand estimation for an AMoD system in Paris, paper presented at the 30th IEEE Intelligent Vehicles Symposium, June 2019, Paris, France.

Travel Behaviour

Zurich model, calibrated
for Paris population

Cost sturcture

Adapted from Berlin

AMoD Simulation

Load-balancing heuristic

Maximum static demand: 2.3M trips