A modest conference hosted by the University of Essex in the UK, the IEEE Computer Science and Electronic Engineering Conference is an interesting approach to the conference formula. It’s small-scale and relatively broad, but carries an emphasis on quality. It carries itself well: providing a valid platform for the dissemination of strong research from scholars across the UK. A true testimony to the successes of a conference, in my humble opinion, is the opportunity for postgraduate students and doctoral candidates to disseminate their work. Having gone through the process, there is something to be said not only for the chance to publish your work prior to thesis writing, but also a chance to build a voice around your research profile. Sure, it’s a scary prospect: a sentiment echoed not only by students who I chatted with but my own experiences many years ago. But confidence comes with time and a platform that is local and relatively intimate in size can help foster that faster than large, immense conference venues.
This confidence does not apply solely to the dissemination of research, but also to the organisation of academic conferences. For those unfamiliar, the majority of organisation for academic conferences is conducted by the very researchers who would typically attend it. The management of the call for papers, reviews, proceedings, room bookings, coffee breaks (which I must remind you, are highly crucial) are all handled by academics. At CEEC, much of the organisation was passed down to the PhD students based at the University of Essex, while the overall coordination is managed by more senior academics (with head of department Professor Simon Lucas leading the charge). It’s a useful approach, given it allows for the necessary exposure and credibility of the conference, but gives the ‘next generation’, if you will, the experience of coordinating such an event.
Overall CEEC was an interesting and fun experience and this round-up is focussed largely on the small but fun computational intelligence and games segment that took place at the event. With all of the work published from student works.
Games @ CEEC 15
Before I speak largely of the work of students, I wanted to take a moment to consider something tangential albeit related in the keynotes. The third keynote of the conference delivered by Professor Sanaz Mostaghim from Otto Von Guericke Universität Magdeburg (Germany) giving an overview of the work she has pioneered in the field of swarm intelligence. Swarm intelligence is a rather exciting and fascinating sub-field of artificial intelligence with an emphasis on mimicking phenomena observed in nature. Specifically the creation of complex behaviours being generated from individual and de-centralised behaviours. Such behaviour is typically observed in the likes of ant colonies and flocks of birds, but has tremendous applications in the field of computing. As Professor Mostaghim explained, this can range from robotics to non-player character control in video games to the colour of hair dye!
Looking specifically at the Computational Intelligence and Games track of the event, chaired by Dr Diego Perez-Liebana of the University of Essex, we saw a small but interesting collection of games research taking place. This started first with Mohamed Abbadi from Ca’Foscari University in Italy whose work is focussed on the Casanova language. Casanova is a novel programming language seeking to optimise the development of games versus the traditional means used by popular game engines such as the use of C++ and C#. Casanova aims at achieving improved performance in a variety of engines, from MonoGame to Unity3D by rebuilding the paradigms of game development. In this publication, Mohamed is focussed on the issue of encapsulation in programming languages and how to allow for quick and easy manipulation of data.
This was followed by some interesting work by discussed by Xenija Neufeld from Otto Von Guericke Universität Magdeburg. Xenija’s work takes the rather popular General Video Game AI (GVG-AI) framework and procedurally generates levels within this engine. This is achieved through the use of Answer Set programming to establish rules parsed from the original Video Game Description Language (VGDL) interpretations in the GVG-AI framework. By establishing a collection of rules that could remain static and some with room for manipulation, an evolutionary algorithm would build new games and levels. The sample Monte-Carlo Tree Search (MCTS) and random agents provided in GVG-AI were subsequently used to assess the validity and quality of these levels. While certainly there are many challenges yet to be overcome by this work, it presents a valid (and interesting) method to procedurally generate games for the competition.
Third on the roster was my own dissertation student, Michele Vinciguerra from the University of Derby. While sadly I spoke in Michele’s stead, our talk discussed work on establishing a framework for the construction of robot agents for a video game that is constrained by rules of construction in addition to the use of resource budgets. This would allow for a range of robots to be constructed in-game whose resource is parametrised: permitting procedural generation that can be controlled as part of a games difficulty scale.
Lastly, we finished up with a presentation by Piers Williams of the University of Essex whose first body of work is exploring the challenges of 2-player cooperative games and whether current practices, notably Rolling Horizon Evolutionary Algorithms and Monte-Carlo Tree Search can solve these problems. While working in a small, albeit challenging problem space, there is still significant evidence of some algorithms (notably MCTS) performing better than others in these problems.
While the games track itself was rather small, the effect and indeed influence of games was evident throughout the conference, with research in other tracks adopting the likes of Unity3D for applications, to the discussions in two keynotes of AI applications in games. Furthermore, the final day ended with the closure of the 2015 General Video Game AI Competition, Naturally there is much to be said of the continued successes and indeed challenges remaining in GVG-AI field, but we hope to come back and speak about this specifically in a future article.
As we wrap up, a sincere thank you to all who made my visit to Essex an engaging, interesting and generally fun experience. A list of publications from the games track is listed below for those interested in reading more on these works.
- High Performance Encapsulation in Casanova 2.
- Mohamed Abbadi (Ca’Foscari University, Italy) et al.
- Procedural Level Generation with Answer Set Programming for General Video Game Playing.
- Xenija Neufeld (Otto-von-Guericke-University, Germany) et al.
- A Procedural Generation Framework for a Robot Construction Game.
- Michele Vinciguerra (University of Derby, UK) et al.
- Monte Carlo Tree Search Applied to Co-operative Problems.
- Piers Williams (University of Essex, UK) et al.
Note: We will aim to have links to publications if possible, but would like to ask for the blessing of CEEC 2015 coordinators first.