Project Overview
This project aims to develop an AI agent capable of protecting villagers by eliminating randomly spawning zombies, and minimizing villager casualties. Using reinforcement learning, the agent learns to identify and prioritize threats, optimizing its decision-making to neutralize zombies as efficiently as possible.
Project Goal
The objective is to maximize villager survival by training the agent to execute strategic actions, such as navigating the environment, locating threats, and attacking zombies effectively.
Input Data
Map environment: Defines the terrain where the agent, villagers, and zombies navigate. Entity positions: Real-time tracking of all objects, including obstacles. Decision-making algorithm: Guides the agent in selecting optimal actions such as movement and attack strategies.
Output Metrics
Possible Outcomes
The agent is eliminated by zombies.
All villagers are killed.
The agent successfully protects villagers within a set time frame (e.g., 2 minutes).
Deep Reinforcement Learning (DQN/Double DQN): Enables the agent to continuously refine its strategy through experience. Policy Optimization: Adjusts decision-making to prioritize high-value targets efficiently.
Key Metric: Number of villagers saved within a defined time limit (e.g., 2 minutes).
Baseline Performance:
The agent reacts without prioritization, focusing only on killing zombies rather than maximizing villager survival. Uses basic strategies (e.g., protecting a single villager while ignoring a larger threat, relying on a single weapon, or aimlessly running around).
Proposed Improvements:
Visualization & Behavior Analysis: