Your agent will need to follow an Agent Oriented architecture. It will have a number of methods that the environment will call. Most of these methods will simply pass parameters to the agent containing information about the environment (e.g. who was on the mission, and how many betrayed the mission). The agent should use these parameters to update their internal state.
Some of these methods will require the agent to perform an action (e.g. propose a mission or vote on a mission). The agent should use the information in their internal state to strategically select the best action.
This scenario is an imperfect information game. True resistance members do not know who the spies are.
Resistance members typically want to limit their uncertainty whilst spies (who know everything) want to maximise the uncertainty in their opponents. In the worst case there are hundreds of different possibilities to consider. Good agents will need to consider which scenarios are most likely given the observations they have seen so far. Good spies will need to consider what the resistance members consider most likely scenarios and exploit this. Statistical modelling will be important here. Bayes rule, Markov models and Monte Carlo tree search are all potential techniques to try.
Adventurous people may want to consider epistemic logic, logic programming and theorem proving for reasoning agents, or perhaps try genetic algorithms or neural nets to build optimised populations of agents.
You are given a lot of scope to choose your research question. For example, you may choose to focus on opponenet modelling (which would require modifying the test harness), social evolution where population norms may emerge, or identifying optimal strategies in perfect information version fo the games. To write a good resaerch report you should be clear on:
- What the research question is;
- Why the question is interesting;
- How you intend to answer the question;
- What data you collected; and
- What the data tells us about the question.
This is not the first time an AI resistance tournament has been held. This paper (./Resistance.pdf) describes a number of different techniques and how well they performed. Note that the rules of the tournament are different and opponenet modelling will be restricted. Use this paper as a starting point to look up extra references and do a thorough literature review. Make sure that you cite all resources you use in full.
EasyDue™ 支持PayPal, AliPay, WechatPay, Taobao等各种付款方式!
E-mail: firstname.lastname@example.org 微信:easydue