Practical 1 — River Crossing 🌉🔥

Overview ✨

Educational Gymnasium environment modeling the classic River Crossing puzzle, plus multiple agents to interact with it: random, predefined action list, and a simple reflex strategy that solves it in exactly 15 minutes.

The Puzzle 🧩

• Friends: Alberta (1m), Bernardo (2m), Carlos (5m), Diana (8m)
• Rules:
  • The bridge holds at most 2 people per crossing
  • There is a single torch that must be carried on every crossing
  • Someone must return with the torch when needed
  • Crossing time equals the slower person in the pair
• Goal: all four must cross in exactly 15 minutes

Project Structure 📁

• river_crossing_env.py: Gymnasium Env (observation/action spaces, transitions, reward, render)
• agent.py: abstract agent interface (next_action(obs))
• input_agent.py: reads actions from the console
• random_agent.py: generates random actions (validated by the env)
• action_list_agent.py: executes a predefined list of actions in order
• simple_reflex_agent.py: fixed policy that achieves the 15-minute solution
• river_crossing_utils.py: small helpers (e.g., finish)
• river_crossing.ipynb: guided notebook to explore and run agents

Action Format 🎮

Each action is a dictionary:

{"direction": 0|1, "person1": 0..3, "person2": 0..3}
# direction: 0 = left, 1 = right
# persons: 0=A, 1=B, 2=C, 3=D

How to Run ▶️

• Recommended: open river_crossing.ipynb and run cells in order.
  • Sections: “Input Agent”, “Random Agent”, “Action List Agent”, “Simple Reflex Agent”.

Requirements 📦

• Python 3.11+
• Gymnasium
• (Optional) Poetry

Quick install with pip:

pip install gymnasium

Included Agents 🤖

• Random Agent: explores by sampling actions at random.
• Action List Agent: consumes a predefined sequence of actions.
• Simple Reflex Agent: deterministic policy that reaches the 15-minute goal.

Minimal script example (similar to the notebook):

from river_crossing_env import RiverCrossingEnv
from simple_reflex_agent import SimpleReflexAgent

env = RiverCrossingEnv()
agent = SimpleReflexAgent(env)
obs = env.reset()
done = False

while not done:
    action = agent.next_action(obs)
    obs, reward, done, info = env.step(action)

print("reward:", reward, "total_time:", info.get("time"))

Recommended Reading 📚

Gym/Gymnasium basics: Official docs