Our research-driven approach has attracted top
cities to contract us for high-resolution traffic


Changes to public spaces, e.g. new bike lanes require detailed know-how
about what will happen to traffic demand and congestion.

Problem statement & data

Transport for London is the government
body for most of London’s transport

During the COVID-19 pandemics, London saw
changes in traffic performance between 2019 and
2020. In collaboration with TU Munich, we examined a
high-resolution data set of more than 11’000 loop
detectors and thousands of traffic cameras covering a
period of 2 years.

The image shows the vehicle
flow (every 3min) for a lane in
London. Our data analytics
revealed a decrease in flows
from 2019 to 2020 can be


Our automated pipeline generated a
digital road network and directly
matched the available data as well
as the changes in speed policy, lane
management, etc.

We then estimated average traffic demands for 2019 and 2020. Thanks to our endogenous demand estimation module we only required loop detector measurements to predict the origin-destination matrix.

Origin-Destination estimation
for different regions in the
greater London area. We can
estimate these relations using
loop detector counts only.


Our approach disentangled the main drivers
in the complex interplay between
transportation demand and supply.

This allowed us to control for the different changes in
the network and demand and discuss the traffic
performance in Greater London.


We evaluate the short- and long-term effects of construction sites.

Problem statement & data

According to several international rankings
the city of Zurich is at the forefront of
becoming a smart city.

Our ongoing project with the city consists of an analysis
for construction sites in the North of the city. The city has
roughly 1500 loop detectors which offered high
resolution insights over a 4-year period.

Visualization of our simulation


Our automated pipeline generated a
large-scale city-wide traffic
simulation including public
transportation, bicycles, as well as
car traffic.

We focus on an extended morning peak between 05:00-09:00. Transportation demand was estimated using our endogenous demand estimation which only requires loop detector measurements to predict the origin-destination matrix.

Comparison of the average of
empirically measured flows in
6 districts in the city of Zurich
and our simulation
(comparison every 3min).


Our simulation offers very high accuracy,
with less than 5% deviation in the measured
flows (over 15min periods), and 85% of all
monitored roads showed a very good value
of below 3.8 for the international standard
GEH statistic.

We are now working on the real-time prediction of
congestion, so that our simulation truly transforms into
a digital twin.