Frans Groen
The ideas we study in this project are centered around the concept of landmark recognition. A user should be able to walk in an unknown environment, make a picture of a landmark/point of interest, after which our aplication returns on topic information. In a more particular situation a tourist walks across a large square in Amsterdam and sees an interesting church. He makes a picture of it and his mobile phone tells it’s the Nieuwe Kerk situated at the Dam square, it’s opened from 10:00 to 18:00, and he could visit a concert that evening.
Arnoud Visser
The goal of the RoboCup Project is to let a robot soccerteam win from a human team in 2050. For that reason every year the circumstances become more realistic. This year the circumstances for the Aibo League are nearly equal to the previous year, to stimulate the emergence of teamplay. It is the challenge for this year students to design and evaluate behaviors that make use of information from the teammates. The project will deliver a model of joint actions, and perform experiments to demonstrate the value of this modification (favorably as function of the reliability of underlying modules), for instance by playing 'fooling around'. 'Fooling around' is an extension of a passing game with an opponent that chases the ball.
Peter van Lith
A number of simulated animats implemented as agents live in a very simple world, where they are attracted to each other, are repelled or ignore each other. Influenced by rules that determine their behavior or by altering their collective behavior complex interactions must occur between the agents and their environment. This behavior is later on going to be modified by a number of sensors that thus allow this program to become an interactive work of moving art. It's main purpose is to allow students to experiment with a very simple agent world and artists to create interesting interactive computer-based objects. Develop a simulation environment in which a map is shown and on it the location of a number of sensors. An XML script simulates sensor values that also need to be shown on the map. In addition to the simulated sensors some real sensors are used in the simulator.
Bram Bakker
Last year we started a project concerned with intelligent traffic control. Using a simulator that models urban road traffic, we developed an improved traffic light controller based on measuring traffic congestion on the roads and reinforcement learning. This year an important focus will be on dealing with traffic accidents. In this student project we want to investigate a learning traffic control system that will detect accidents and handle them intelligently. It will learn to direct traffic around accident spots such that traffic delay is minimized.
Marinus Maris
The identification of persons in homes is an issue that hasn't been solved yet. The problem is that most current detection systems are either not reliable or difficult to work with. An interesting approach is to identify persons by means of analyzing their movements and using sensors for easy-to-detect characteristics (such as height and weight). In this project, a detection system must be created, based on a database with rules and a Bayesian network. The resulting reasoning system will have to identify persons by attaching probability values for their identification.