At a Glance

Title: Perception and Planning for Autonomous Driving
Dates: July 7-10, 2014
Who is welcome: Graduate students of all levels
How to apply: See the “Application” Section below

If you have any questions, please feel free to contact:



July 7th – July 10th, 2014

Research into fully automated driving is maturing rapidly with all major vehicle manufacturers targeting the rollout of some form of vehicle automation in the years ahead. This ambitious goal will see complicated robotic systems deployed widely. However, there is still a gap between the commercial systems now available—such as automated cruise control or parking assistance—and the functionality and robustness needed for truly automated driving. Providing this functionality using only a suite of sensors that is affordable for consumers is a great challenge that must be solved before fully automated vehicles make it into the hands of average consumers.

The V-Charge project will host a summer school on fully automated driving. We want to prepare a new generation of engineers for these challenges, expose them to the current state of the art, and highlight fascinating areas of future research. The summer school will give a compact introduction to the field of fully automated driving, offer technical sessions that will provide hands-on experience with state-of-the-art techniques, and present invited talks to frame the open research questions in the field.


  • Roland Siegwart, ETHZ – ASL
  • Paul Furgale, ETHZ – ASL
  • Christof Stiller, Karlsruhe Institute of Technology
  • Michel Parent, INRIA,
  • Uwe Franke, Daimler AG
  • Falko Dressler, University of Innsbruck
  • Steven Shladover, University of California, Berkeley
  • Ulrich Schwesinger, ETHZ – ASL
  • Marc Pollefeys, ETHZ – CVG
  • Torsten Sattler, ETHZ – CVG
  • Stefano Cattani, University of Parma
  • Julian Timpner, TU Braunschweig
  • Lina Paz, University of Oxford
  • Hugo Grimmett, University of Oxford
  • Alex Stewart, University of Oxford
  • Paolo Medici, University of Parma


Participation is limited to 50 students. Candidates will be selected based on the quality of their application. Please state on one single page (11pt font):

  • your scientific background and degree
  • your motivation to come to this school
  • the topic of your thesis
  • any prior knowledge in the field Please include a current academic transcript with your application (and optionally a CV with a list of publications) and send it to

EXTENDED deadline for the application is April 22th, 2014.

Successful candidates will be notified by April 29th and need to register and pay the course fee by May 13.



The cost for the summer school is

  • 300 CHF including accommodations
  • 250 CHF without accommodations

This will cover all course materials, as well as breakfast, coffee-breaks, and two dinners during the school. Lunch and dinner are available at the ETH-Cafeterias and in nearby venues from ~15 CHF (including a drink). For the whole week, we put the additional living expenses at around CHF 200.


We will provide basic accommodation for all participants in a remodeled former air-raid shelter in the basement of the Computer Science building (no windows). This is right next door to the course venue, and will feature dormitory style rooms with shared bathrooms. Pillows, blankets, and sheets will be provided. This accommodation will be free of charge for all participants. It will be open from Sunday evening to Friday morning, to accommodate travel that requires an additional night in Zurich. If more privacy and comfort is desired, students must organize alternative accommodation on their own responsibility and expense (double rooms in Zurich start at around 65 CHF per night and person).


The summer school is organized by the Autonomous Systems Lab ( and the V-Charge project ( in collaboration with:

  • Vislab at the University of Parma (
  • The Automotive Research Centre Niedersachsen at Technische Universität Braunschweig, Germany
  • The ETH Zurich Computer Vision and Geometry Group (
  • The Oxford Mobile Robotics Group (

This summer school is funded by the EU FP7 2007–2013 Program, Challenge 2, Cognitive Systems, Interaction, Robotics, under grant agreement No 269916, V-Charge.

Technical Sessions

Motion planning for on-lane driving – ETH Zurich Autonomous System Lab

This technical session will introduce students to the state-of-the-art in motion planning for fully automated cars. It will provide an overview of basic historical motion planning strategies, trajectory generation for sampling-based methods, cost function design as well as fast collision checking techniques for static and dynamic obstacles. During the exercise, students will implement their own planning algorithm and test it within a vehicle simulation framework.


Image-based localization — ETH Zurich Computer Vision and Geometry Group

Image-based localization is a key component for determining the position and orientation of a camera-equipped vehicle. This technical session will therefore introduce students to state-of-the-art approaches for image-based localization relative to a given 3D model of the scene. The session starts with a brief introduction of the principles of Structure-from-Motion techniques that reconstruct scenes from a set of images. It then explains how a novel query image can be localized relative to such a 3D model by establishing correspondences between 2D image positions and 3D point positions in the model, with a focus on efficient and effective techniques for large scale localization. The last part of the session finally gives an overview over 3D map construction and image-based localization for fully automated vehicles.
During the exercise, students will implement a simple image-based localization method, visualize their results, and extend their implementation to achieve state-of-the-art results.


An introduction to Delay/Disruption Tolerant Networks (DTN) — Automotive Research Centre Niedersachsen, Technische Universität Braunschweig

This technical session is about Delay or Disruption Tolerant Networks (DTNs). The students will learn about the history of this communication architecture and recent research topics such as disruption-tolerant routing. Information on the typical use cases of DTNs inside and outside of the project will be provided during the session.

In the hands-on exercise, the students will set up the IBR-DTN software stack, an open source implementation of DTN, and experience the multi-hop features and disruption tolerance of the DTN by writing their own small application.


Semantic Mapping and Introspection for Classification — University of Oxford

Automated vehicles operating in urban environments can gain a lot of information about how to behave from a higher-level understanding of the objects around it. For instance, the knowledge that there is an upcoming traffic light or pedestrian crossing should be taken into account in the vehicle’s current motion plan and speed. Because many of these useful semantic cues are fixed in place, it makes sense to build up reusable semantic maps of the areas in which we want our automated cars to operate. Usually these maps are created manually, which is very time-consuming but offers guarantees that are currently not available when using unsupervised machine learning classifiers.

We are interested in creating semantic maps automatically (in an unsupervised, or semi-supervised manner), but we also need to have some idea of the confidence of a particular map. After all, if the vehicle is not totally sure about what is around it, it should drive more cautiously! There are classification algorithms which offer probabilities (like the Support Vector Machine (SVM) or LogitBoost, or the Gaussian Process (GP)), but can we trust these probabilities?

Introspection is a property of a classification framework to give appropriate probabilistic output. Are all classifiers equal in this regard? We will discuss the various merits of different frameworks, and give some theoretic insights about why all probabilities should not be considered equal.


The Oxford RobotCar: Autonomy on Offer — University of Oxford

The Oxford RobotCar Project aims to provide robust, long-term fully automated driving at a price point suitable for mass-market adoption. At its core, however, lies a philosophy fundamentally different to that fuelling more conventional automated driving research: autonomy when offered, as opposed to on demand. Initially, our vehicles will drive some of the people to some of the places some of the time. The vehicle will decide when it is safe to offer autonomy and for how long. And only over time – as the vehicle experiences the environment – coverage will increase. In this talk we will provide an overview of how this philosophy impacts the technology we develop.


Machine Learning for vehicle and pedestrian detection — University of Parma

Pattern recognition approaches have achieved considerable success in practical applications and the motivations to recognize pedestrians and vehicles in automotive environment are largely known. Embedded devices, that perform these tasks, based on image processing, are now being installed on commercial vehicles. Even considering the growth of computational power on embedded devices, not all algorithms, at present time, are suitable and ready to be used on devices in the coming years.This technical session aims to illustrate some promising algorithms,such as AdaBoostand SVM, the most common descriptor spaces, such asHaar, ICF and HOG, and give the bases for applying a classifier on an image.During the exercise, students will train several classifiers and apply them on images, understanding in such way potentials and problems behind those kinds of applications.

Slides from Talks and Technical Sessions


Steven Shladover — Road Vehicle Automation History, Automation, and Challenges (slides)

Cédric Pradalier — Environment Monitoring and Driving Assistance (slides)

Michel Parent — Cybercars: the first fully automated urban vehicles on the market (slides)

Cyrill Stachniss — Visual Localization Across Seasons using Image Sequences (slides)

Christoph Stiller — Map-Based Autonomous Driving (slides)

Marc Pollefeys — Visual Mapping and Pose Estimation for Self-Driving Cars (slides)

Uwe Franke — Autonomous Driving Reached Downtown (slides)

Alex Stewart — Shady Dealings: Techniques for Long-term Outdoor Visual Localization (slides)

Lina Maria Paz — The Oxford Robotic Car (slides)

Falko Dressler — Connected Cars (slides)

Technical Sessions

Hugo Grimmett — Introspective Classification for Robot Perception (slides)

Torsten Sattler — Image-Based Localization (slides)

Julian Timpner and Stephan Rottmann — An Introduction to Disruption Tolerant Networking (DTN) (slides)

Paulo Medici — Object Recognition (in Automotive Environments) (slides)

Ulrich Schwesinger — Motion Planning for Autonomous Cars (slides)


Monday, July 7 Tuesday, July 8 Wednesday, July 9 Thursday, July 10
8:30–10:00 Welcome and overview  

UOX: The Oxford RobotCarSemantic Mapping and Introspection


ETHZ-ASL: On-lane planning I TUB: An introduction to Delay/Disruption Tolerant Networks I
10:30–12:00 Invited Talks: 

Steven Shladover


UOX: Training and Evaluation of Classifiers 

Invited Talk:

Cedric Pradalier

ETHZ-ASL: On-lane planning II TUB: An introduction to Delay/Disruption Tolerant Networks I
13:30–15:00 Invited Talks: 

Cyrill Stachniss

Alex Stewart

Lina Paz

Invited Talks: 

Falko Dressler

UParma: Machine Learning for vehicle and pedestrian detection I ETHZ-CVG: Image-based localization I
15:30–17:00 Invited talks: 

Christof Stiller

Michel Parent

Invited Talks: 

Marc Pollefeys

Uwe Franke

UParma: Machine Learning for vehicle and pedestrian detection II ETHZ-CVG: Image-based localization II
17:00 + Rooftop Barbecue FIFA World Cup Football FIFA World Cup Football Wrap-up Banquet

One thought on “The V-Charge Summer School: Perception and Planning for Autonomous Driving

  1. [...] “The V-Charge Summer School: Perception and Planning for Autonomous” July 7th – July 10th, 2014 Swiss Federal Institute of Technology Zurich (ETH Zurich) [...]