As expected, the first meeting with my project supervisor highlighted that I had outlined a project more suitable for a PhD student and that I would need to refine things a bit for an undergraduate project. After a discussion, we agreed that I would look at what the fewest number of landmarks required for navigation are, as determined by research into insect navigation.
A high-level review of a wide range of research documents suggested that insects do not recognise individual landmarks, instead using scene matching through snapshots to guide them to a known area. Taking desert ants as an example, researchers identified two forms of navigation: visual, landmark-based navigation; and the insect equivalent of dead-reckoning, path integration. It was discovered that adding landmarks to a scene did not affect the visual navigation of the ants as, from their perspective, it blended into the overall scene.
Alongside the high-level research, it was also important to start looking at technology options. With scientific tools generally requiring a Linux platform, this meant installing VirtualBox on my laptop along with the Ubuntu operating system, allowing the first multi-agent simulation software to be tested: Flame.
The plan for the coming week is to:
- Refine research into biological systems to understand core features to model
- Identify previous studies where research has been implemented, to get further ideas
- Determine approaches to implementing navigation
- Determine technology options, such as agent-based modelling systems MASON, Flame, SeSam, Repast Simphony and SimSpark
The core concerns I have at present are that, a) I have yet to decide whether to go down the simulation or robotic route, and b) in both cases, the mathematical modelling required could be beyond my capabilities.