University of Luxembourg

There is a lot of AI happening at the University of Luxembourg! And we’re going to experience some of it during our stay there with the AWAI! Here we can see what to expect on 11 March:

Time table
location: Salle Paul Feidert

10:00 Welcome Speech by Pierre Kelsen

10:30 Presentation by Martin Caminada (Dutch, Formele Argumentatie; het definiëren van non-monotone afleiding d.m.v. graaftheorie)

11:30 Introduction for the afternoon workshops (Brain/Computer Interfacing by François Beckius and RoboLab by the RoboLab team)

12:00 Lunch!

12:30-15:30 Workshops (Brain/Computer Interfacing and RoboLab)

13:30 Presentation by Yoann Pinge (Realistic Simulations of Communicating Vehicles for Road Traffic Management)

14:30 Presentation by Tim Muller (Semantics of Trust)

15:30 Presentation by Rohit Parikh of City University of New York (The Logic of Campaigning)

16:30 Presentation by Nicolas Troquard of the University of Essex (Logics of agents with small models)

The abstracts:

Martin Caminada – Formele Argumentatie; het definiëren van non-monotone afleiding d.m.v. graaftheorie (Presentation, in Dutch)

In deze presentatie zal ik proberen een overzicht te schetsen van het onderzoeksgebied van formele argumentatie, en in het bijzonder van de populaire aanpak die geïnitieerd is door Dung in zijn 1995 paper in Artificial Intelligence. Het idee is grofweg om formele systemen voor non-monotoon redeneren te definiëren op basis van graaftheorie. Hierbij gaat het om een gerichte graaf waarin de knopen worden gevormd door weerlegbare afleidingen (“argumenten”) en de pijlen worden gevormd door een relatie van weerlegging (“valt aan”). Het blijkt mogelijk de redelijke vormen van logische afleiding te definiëren door middel van het selecteren van argumenten van deze graaf, zonder te kijken naar de logische inhoud van deze argumenten, puur op basis van de topologische structuur van de graaf. In de presentatie zal gepoogd worden te schetsen hoe dit alles functioneert en wat de nog open problemen zijn.

François Beckius – Brain-computer interfaces (BCI) (Presentation + Workshop)

Presentation: Brain-computer interfaces (BCI) allow a user to control an external device merely by thinking about it, with no involvement of the body. They achieve this goal by observing the electrical potential generated by neurons firing within the user’s brain. This presentation introduces the concepts behind gathering brain activity data and about translating it into useful commands. In a second part, it will show what can already be achieved with the technology and what still causes problems.

Demo: For the demonstration, two commercial BCI headsets will be available to practically illustrate what has been shown in the previous presentation. A few students will have the chance to work directly with the devices to showcase what they can do. Unfortunately, not everyone will have the opportunity to get hands-on-practice with the devices, since the set-up and training time for each user takes around ten minutes. Demonstrated will be vendor software allowing the user to move and rotate an animated 3D block, and a program that lets the user move a cursor with his mind.

Link to the website of our equipment’s vendor: http://www.emotiv.com/

The RoboLab Team – RoboLab (Presentation + Workshop)

The interdisciplinary research group Individual and Collective Reasoning (ICR) at the University of Luxembourg consists not only of humans but also of humanoids. They go by the names of Jempi, Rosi, Lucy, Fonsi and Marvin. These five Nao robots (http://www.aldebaran-robotics.com/) lend us their friendly services for educational and research purposes.

In this workshop we will give you the opportunity to play around with these robots. We will offer a short tutorial on the visual programming environment called Choreographe. This is an easy and intuitive way to edit movement and interactive behavior, which will get you started in no time. In addition, you can learn about the more advanced ways in which the robots are programmed, and we will demonstrate what the robots are capable of.

Website: http://robolab.gforge.uni.lu/

Yoann Pingé – Realistic Simulations of Communicating Vehicles for Road Traffic Management (Presentation)

With the increasing vehicles traffic and its environmental impact, road safety and traffic management has become one of the biggest concerns of most national and regional transport departments. Solutions have emerged with the recent vehicular wireless communication capabilities, which constitute a key component of Intelligent Transportation Systems (ITS). Such Vehicular Ad-hoc Networks (VANET) have rapidly emerged and raised novel research challenges such as networking protocols adapted to high speed environments.
Due to logistic, environmental and economical reasons, simulation is a must in the validation process of any of such solution. However, the relevance of a simulation platform depends on its ability to accurately consider the road environment and the vehicles motion. This implies the development of realistic models for the signal physical characteristics (attenuation, reflection, diffraction, scattering), the road network (speed limits, number of lanes, traffic
light logic, etc.), the mobility at a macroscopic level (flows of vehicles, planned routes) or at a microscopic level (acceleration/deceleration, car following each other, slow down anticipations, etc.).
In this presentation we will introduce our ongoing research work with a platform for online VANET simulation and a vehicular mobility model that simulates the mobility of thousands of vehicles, based on real traffic data from the roads of Luxembourg.

Tim Muller – Semantics of Trust (Presentation)

The meaning assigned to the word `trust’ is diverse. A formalism for trust should, therefore, allow various interpretations of trust. To simplify matters, we assume a static environment for our analysis. To this end, we introduce terms that specify the observations of agents, called connections. Then we apply epistemic semantics to reason about the knowledge of agents. We allow specifications of interpretations of trustworthiness in terms of facts. Given a specific interpretation; an agent that knows the right facts, can know that an agent is trustworthy. If agents know that a target is trustworthy under an interpretation, that agent trusts the target.

Specific difficulties arrive when dealing with observations made by other agents; so called second hand observations. Agents can lie about their observations, even when we trust them. Second hand observations are, therefore, never as good as first hand observations. On the other hands, we may believe that specific agents generally speak the truth. Second hand observations made by reliable agents should not be discarded. We show a method to reduce the problem of reliability of second hand observations to the trust problem.

Rohit Parikh (City University of New York) – The Logic of Campaigning (Presentation)

There is a great deal of existing work on changes in belief states when a public announcement is made. How is this work relevant to campaigning? In the US, while a presidential election typically takes a single day, the processs of campaigning starts more than a year in advance, with various candidates trying to influence public opinion. We offer a simple propositional logic based model for this process. We assume that the various issues are represented by certain propositions which may become true or false as the result of a particular candidate becoming elected. We also assume that various groups of voters of different sizes have wishes as to whether they want a particular proposition to be true or false and who also have a degree of passion for each such proposition representing how much they care. The candidate has already made some statements indicating how she will act if elected but her positions on some issues may be unclear. How should she speak now in order to increase her chances of being elected?

Nicolas Troquard (University of Essex) – Logics of agents with small models (Presentation)

Possible worlds and relational semantics are commonly used to model multi-agent systems. However, describing a complex system in terms of possible worlds is often unpractical. For instance, in MOCHA (a model checker for ATL), reactive modules are used to overcome this difficulty: the powers of agents and coalitions are derived from the ability to control some variables. In this talk, I will provide evidences that interesting logics with small models are attainable for a variety of notions: coalitional ability, dynamicity and imperfect information.

Coalition Logic of Propositional Control (CL-PC) (van der Hoek & Wooldridge 2005), whose models are directly inspired by reactive modules, is an example of logics of agents with small models. I will also present Dynamic Logic of Propositional Assignments (DL-PA) that is evaluated wrt. simple propositional models. I will show how DL-PA can simulate CL-PC. Finally, time permitting, I will present one or two epistemic variants of CL-PC. Axiomatisations, and complexities of model checking and satisfiability checking will be discussed.

Contains joint work with Wiebe van der Hoek, Michael Wooldridge (epistemic CL-PC) and Andreas Herzig (DL-PA).