The basic idea of this project was to develop a framework with which games or installations can present a continuous game world across many different screens. To achieve this, I wanted to realize a cost-effective solution that does not require expensive specialized graphics cards with multiple outputs.

My solution uses inexpensive Raspberry PI 2 devices for which I wrote a custom graphics engine in C++ (Piun a.k.a. PI Unleashed) to take full advantage of the Raspberry PI's hardware capabilities and guarantee constant high frame rates.

Additionally, there is an application to manage the individual RASPI/display units (called Units) and align them according to their position (MuMu Commander). I developed this application with POWIDL for Java.

The audience could experience a simple spaceship game, the framework rendered on 12 screens. Visitors could ask questions about the underlying technique and try out by themself how to adjust the position of screens using the commander application or add and arrange new objects within the game world.

---

In the context of this framework, I offered and supervised the following master's thesis:

• Brucker, Eva-Maria Automatische Einrichtung von Multi-Monitor-Systemen mithilfe von optischen Bezugsmarkern. Interactive Media; FH Oberosterreich – Fakultät für informatik, Kommunikation und Medien, 2016. (PDF)

This work shows different video sequences, each generated from a single image. I used the y-axis of the images as the time axis for the video sequences and extended the x-axis to the entire frame height to replace the missing y-dimension. A careful selection of input images combined with fitting music shows the potential and variety of this concept.

I got inspired by this work by a technique called tx-transform¹, developed and used by Martin Reinhart. This technique is used for videos and transposes the time axis and the space x-axis of video sequences with one another.

An alternative interpretation for the video sequences is the following. The videos show how one-dimensional beings might perceive two-dimensional structures (images) moving through their space. That interpretation is, of course, inspired by Edwin Abbott's Flatland²

ByteLight

I generated the video sequences in ByteLight, a graphics program I developed.

ByteLight is script-based rather than GUI centered, and I originally designed it to create textures for video games. However, it has established itself as a reliable piece in my toolbox, and I often use it to visualize and debug graphical problems and conduct creative experiments.

---

¹Martin Reinhart. tx-transform. https://www.tx-transform.com

²Edwin A. Abbott. (1884). Flatland: A Romance of Many Dimensions. Seeley & Co.

Since my main teaching domain was game development, artificial intelligence (AI) is one of my fields of interest. Besides the currently prevalent methods of Deep Learning, I have extensive experience with classical AI, e.g., path-finding, optimization problems, and solution strategies for board games. Still, my main interests revolved around machine learning, mainly reinforcement learning.

Reinforcement Learning

The most exciting aspect of reinforcement learning is that it does not require large amounts of labeled data sets. Simulation and games are excellent ways to give AI enough opportunities to learn unassisted.

I conducted numerous experiments to replace classical AI with reinforcement learning, especially Q-learning. For example, I wanted to mimic human errors and human gaming behavior.

The required infrastructure, e.g., Q-tables and state-quantizers, are integral to my game engines. I used the conclusions in lectures and for topics for bachelor and master theses.

Deep Q-Learning

The limitations of Q-learning become apparent when dealing with large state spaces, such as those found in complex physical environments. Deep Q-learning is a possible solution to this problem by combining Q-learning with neural networks.

Virtual Reality as a development tool accompanies me in almost all projects, whether video games, digital installations, or handcrafted physical exhibits.

Exhibition venues are usually difficult to access in advance, making testing interaction concepts for interactive installations and games for co-located settings difficult. However, a functioning interaction concept and the impact of proper textures, lighting, and sound are crucial for the success of an installation.

That's why I've been using VR as a test environment for such projects for a long time. But VR is also fascinating, apart from the ever-recurring hype around this technology.

Asset and Venue Testing

One of my favorite things is to create 3D models and texture them, and I have found that the requirements for 3D assets for VR applications are much higher than for traditional applications. Therefore, I like to test my work in VR environments. In the case of installations, I prefer to conduct this test in combination with estimated lighting conditions at the exhibition venue.

Virtual Reality as a development tool accompanies me in almost all projects, whether video games, digital installations, or handcrafted physical exhibits.

Exhibition venues are usually difficult to access in advance, making testing interaction concepts for interactive installations and games for co-located settings difficult. However, a functioning interaction concept and the impact of proper textures, lighting, and sound are crucial for the success of an installation.

That's why I've been using VR as a test environment for such projects for a long time. But VR is also fascinating, apart from the ever-recurring hype around this technology.

Asset and Venue Testing

One of my favorite things is to create 3D models and texture them, and I have found that the requirements for 3D assets for VR applications are much higher than for traditional applications. Therefore, I like to test my work in VR environments. In the case of installations, I prefer to conduct this test in combination with estimated lighting conditions at the exhibition venue.