Latest Developments - Deep in the Code

Fuzzy Logic & Hybrid Algorithms

I've refactored the Fuzzy controllers and the environment mapping algorithms from the Evo3 project (2005). I've managed to extract all the necessary parameters from the current terrain model, so that feeding the new environment's information to the refactored systems should be no problem.

All of the old functionality seems to still be ticking along properly - thankfully I had the forethought to build generic Gnu code for these parts and didn't rely on any of the Microsoft VC++ and DirectX8 libraries that were at the time in vogue.

Next I'll be building the old fuzzy controllers into the navigation systems for the autonomous vehicles into the new simulation. Hopefully they will marry up without too much hammering and tweaking. This should be impressive. I'll give each one the hybrid long-distance path planner module as well, which will allow me to publish some new research, but in the long term I'm thinking about separating these two algorithms into different agents.

New Agent 'Force' Society Model
I'm working on a complete specification for a new society of agents model, which I think has potential to comprise a kind of 'holy grail' research for my PhD. Hopefully I can use the simulation engine to build a practical implementation of this theory and draw out some decent results.

Explosions

I've tinkered with explosions using particle visualisation technology. I have a working prototype that looks okay, but I need to construct a gravity affector, and perhaps design a completely new emitter to improve the particle timing.

Big Changes

No more missiles
I'm dumping the projectile model, because there is not enough data available to simulate this accurately, and it doesn't add much quality to presentation. This is going to be replaced with a probabilistic model, which is going to save a lot of development time and produce more realistic results.

New Research
It appears that our Hybrid Fuzzy A* intelligent navigation algorithm is actually still uncontested in the simulation field, and represents some novel research. I'm going to concentrate the main thrust of development on implementing this in the simulation engine, and making any improvements, for international publication. This should be the first research output for '08.

In addition to this, an Intelligent Agent research is emerging from our existing vehicle simulation work, which may turn into something interesting soon...

Craters Improved
I've managed to double the resolution of the landscape and improve the system performance. Still not as detailed as I would like, but a huge improvement over everybody else (including the market leaders) working in this field. Since the last post I've also incorporated a land-scorching textural effect into the blasts.

Physics and Terrain - Shell Crater experiment

This is my first experiment combining projectiles/terrain intersection and the creation of craters on projectile impact. I've fluffed around with it enough so that the dirt heaps up around the crater hole. If projectiles continue to hit the same spot then there is an accumulative effect, which I think might be a bit too dramatic (not a realistic dispersion of energy over terrain volume) . The other problem is that I couldn't make the splash resolution any smaller, which means the craters are too big for the projectiles - maybe I can either make all the terrain bigger or fiddle with some sort of internal details.

It's a little bit hard to see what's happening in the screen shots because of the lack of shading. The yellow dots are 'dead' (spent) projectiles that have already impacted with the terrain and caused cratering.