A third year paper by BSc Games Programming student Michael King has been accepted for inclusion in the 2017 AISB Convention. In this post he and his supervisor discuss Michael’s work.
Dr Mohammad Majid Al-Rifaie, Lecturer in Natural Computing
Throughout history, nature has been a source of inspiration to mathematicians, physicists and technologists. In computer science, for example, swarm intelligence is inspired by ant colonies, bird flocks and fish shoals, where interactions between individual members (or agents) create an ‘intelligent’ global behaviour. Evolutionary computation subjects algorithms to mutation and natural selection, resulting in ever-fitter generations of ‘child’ algorithms.
In my third year module Natural Computing, students learn about nature-inspired techniques, and then apply them to real-life challenges such as medical imaging, protein folding, statistical analysis, economics, art generation and analysis, and engineering.
Following Michael’s choice of Natural Computing as one of his third year modules, he decided to continue working on what he has learnt. Now as part of his final year project, he has prepared an elegant academic paper which has been accepted for presentation and publication for the AISB 2017 Convention in AI & Game Symposium.
Michael is likely to be the only undergraduate student who will be presenting his work amongst PhD students, researchers, and professors. The Society for the Study of Artificial Intelligence and Simulation of Behaviour (AISB), which hosts the event, is the oldest Artificial Intelligence society in the world, and the largest in the UK, so this is quite an achievement!
My paper represents the theoretical side of my third year Natural Computing project, a virtual reality drawing game. In this game, the user creates a drawing within a VR space – and then small spherical agents recreate it. For this to work, the game uses two algorithms, Dispersive Flies Optimisation and A* Pathfinding, to build organic-looking structures based on a simple input.
- Dispersive Flies Optimisation, developed by my supervisor Mohammad Majid Al-Rifaie, is an open source algorithm that mimics how flies swarm around food. I’ve used DFO to identify the next place for an agent to explore. [Download DFO]
- A* Pathfinding is an efficient and reliable algorithm for finding a path from one place (or ‘node’) to another. For my project, I modified the algorithm so that each node was actually an agent. This, coupled with DFO, allowed structures to be built in a rather organic and unique looking way.
In the two pictures above, the green spheres are agents that are still moving, while the red spheres are agents that are locked in place and form the structure. The larger yellow sphere is the target/food for the agents and the red cube is the ‘best’ position for the agents to start building from.
Dispersive Flies Optimisation is an open-source algorithm. Download DFO here