AgentSmith is currently a work in progress bot designed for the 1v1 rumble competition.
You hear that Mr. Anderson?... That is the sound of inevitability... It is the sound of your death... Goodbye, Mr. Anderson...
|Targeting||Virtual Guns with Linear, Circular and Head On Targeting|
AgentSmith has been a bot name I have used for a couple of bots that I have never released over the years. They have always had one thing in common: They used machine learning techniques to try to improve their score. I have used both reinforcement learning and genetic algorithms in various guises. However the bots have never been terribly competitive because the base movement and targeting techniques implemented were never up to scratch and trying to improve the parameters of rubbish techniques doesn't improve the technique, only the effectiveness of the technique.
This version of the AgentSmith is going to be different however. Its still in progress but I plan on making the bot an effective bot before trying to add machine learning techniques to it. Its currently work in progress but I aim to release it on the rumble incrementally and see how my changes get on in a live environment.
- How does it move?
I've recently implemented a new technique I've currently called DangerPrediction. Its still in development so I'm not going to write too much about it but it doesn't surf any waves. It generates possible routes it could take, evaluates danger along those routes and chooses the route with the least danger. Its currently 99% effective versus most linear, circular and head on targeting strategies such as the Wave Surfing challenge bots, but it falls down at the moment versus Guess Factor, DC and pattern matching guns.
- How does it fire?
I've just got a simple Virtual Gun array at the moment choosing the best gun between various guns including Linear, Circular and Head On and now with a DC gun.
- How does it dodge bullets?
See Movement above!
- Where did you get the name?
Agent Smith from the Matrix trilogy, a program that is relentless but can only learn within the bounds of its programming,
Agent Smith now uses Skilgannon's KD-Tree to do its Dynamic Clustering.
Everyone who contributes to the RoboWiki. It's an awesome resource, be proud of yourselves!