🎮 Virtual World Simulator

---

Table of Contents

1. Introduction
2. Getting Started
   • Dependencies
3. Usage
   • Run Training
   • Run Testing
   • Run Fine Tuning
4. Project Structure
5. Model Results

Introduction

The Simulation Hypothesis proposes that the universe we perceive may not be a real, physical world, but rather a vast virtual simulation. This concept was introduced by philosopher Nick Bostrom in his 2003 paper "Are You Living In a Computer Simulation?", arguing that an advanced civilization may have created simulations indistinguishable from reality.

I found this hypothesis fascinating and began exploring whether a similar system could be implemented. Observing how economic systems develop in the real world, I noticed that with the introduction of currency and financial systems, markets emerged, leading to the formation and evolution of nations. If such principles can be applied to digital environments, then the creation of virtual economies and states should also be feasible. In fact, cryptocurrencies like Bitcoin have established borderless financial systems, and concepts of independent economies and virtual states are emerging within the metaverse.

Building on these ideas, I have defined an entity (a human) and constructed an interactive environment in which this entity performs specific actions, learns optimal survival strategies, and strives to live as efficiently and as long as possible. Through this approach, I aim to explore the potential for autonomous evolution within a simulation.

Getting Started

Dependencies

pip install -r requirements.txt

Usage

Run Training

# Single Entity
python train.py --env single

# multi Entity
python train.py --env multi --num_entities <num of entities>

Run Testing

python test.py --checkpoint <path_to_trained_model> --episodes <num of test episodes>

Run Fine Tuning

python fine_tune.py --checkpoint <path_to_trained_model>

Project Structure

Module

• module/Entity.py: Defines entities and their actions (mining, resting, leisure, religious activities, transactions).
• module/Account.py: Defines entity accounts and encryption.
• module/Bank.py: Defines a virtual bank and blockchain-based transactions between accounts.
• module/Transaction.py: Manages transaction records and encryption.
• module/StockMarket.py: Defines a virtual stock market.
• module/Invest/Stock.py: Defines virtual stocks.

Simulator

• simulator/SimulatorEnv.py: Implements a single-entity simulation environment.
• simulator/MultiEntityEnv.py: Implements a multi-entity simulation environment.

Training

• train.py: Agent training.
• fine_tune.py: Supports fine-tuning of the trained model.

Model Results

Results

Episode	Timestep	Total Mined	Balance	Age	Day	End Reason
0	73	17	850	20	3	Health depleted
1	27	6	300	20	1	Happiness depleted
2	46	11	550	20	1	Health depleted
3	50	9	450	20	2	Health depleted
4	200	17	850	20	8	Health depleted
5	81	10	500	20	3	Health depleted
...	...	...	...	...	...	...
26	26	13	650	20	1	Happiness depleted
27	700800	1966	98300	99	29199	Max episode length reached