Background: Behavioral data of participants of an empirical study have been recorded in the context of an interdisciplinary project in cooperation with clinical psychologists. The participants were confronted with a virtual scenario, in which they had to cope with a fire in a parking lot. Thus, physiological as well as geometrical data of 40 participants has been acquired.

Goal: In this bachelor thesis the acquired data should be analyzed with regard to the question, if the inter-individual differences derived from the questionnaires (coping behavior and emotional reactivity) systematically influenced the behaioral data. As a first step, for example, only the recorded head movements could be analzed to see, if it is possible to automatically distinguish more from less fearful participants.

Background: In order to realize an AI-project dealing with the simulation of social emotions such as guilt and shame for virtual agents, the free game engine "Unity 3D" is being used. It is planned to control virtual agents by means of a planning component, which is to be connected to the game engine based on network protocols.

Goal: This project should enable virtual agents to perceive their virtual environment similarly to how real human players would be able to do. Thus, the dynamically changing "field-of-view" as well as the occlusion have to be taken into account, before the perception data is transmitted to the planning component in a suitable format. The algorithms should not only be useful for a single agent in a static environment, but in the best case they should enable multiple agents to perceive a dynamic environment.

Previous knowledge: Basic programming experience in at least one of the languages C#, JavaScript, or Boo would be advantageous, but this can also be gained during the term of the project. In addition, expertise in the fields of "3D Computer Graphics", "Game Design" and "Network Programming" would be beneficial as well as basic knowledge of a 3D modelling tool (e.g., 3DSMax, Maya, Blender).

In robotics the workspace for a manipulator is a subset of 6-DoF space. It is implicitly given by the robot's kinematics. Knowledge about the workspace is necessary when reasoning about a robot's actions as in planning. Usually it can be queried by computing inverse kinematics. Alternatively capability maps precompute a manipulator's workspace and store the 6-Dof structure efficiently.

The goal of this work is to implement such capability maps. This consists of the algorithm for computing the space efficient 6-Dof representation and the querying of capability maps. Extensions like combining different capability maps can be added. The implementation will be done in C++.

Qualitative Constraint Network can be described as a set of qualitative constraints set on elements (such as "The french lesson is before the Math lesson" or "The station is North-East from the Town Hall"). Solving them consists in finding whether or not there exists a real-life instantiation that can satisfy all the constraints.

DaS is a prototype for solving Qualitative Constraint Network. Unlike most of its counterparts, DaS is based on propositional logic instead of CSP. Currently, it does not implement any advanced computation technique.

A linear pseudo-Boolean constraint (LPB) is an expression of the form a_1 l_1+...+a_m l_m ≥ d, where each l_i is a literal (it assumes the value 1 or 0 depending on whether a propositional variable x_i is true or false) and the a_1,...,a_m,d are natural numbers. The formalism can be viewed as a generalisation of a propositional clause. LPBs can be used to represent Boolean functions more compactly than the well-known conjunctive or disjunctive normal forms. E.g., the LPB 2x_1+¯x_2+x_3+x_4≥2 corresponds to the DNF x_1\/(¬ x_2/\ x_3)\/(¬ x_2/\ x_4)\/(x_3/\ x_4). Therefore, in the literature one finds several approaches of generalising techniques from SAT solving to LPBs. All these approahes assume that the LPBs arise naturally from some problem domain.

However, one might ask if it is interesting to transform arbitrary propositional formulae into a set of LPBs, so that one can then benefit from the compact representation during the SAT solving. There is an algorithm solving this problem partially: given a propositional formula representable as a single LPB, the algorithm finds this LPB.

The goal of the project is to implement this algorithm and to conduct some experiments to judge how difficult the transformation is in practice.