Deep Q Network in Mountain Car Environment

import numpy as np
import matplotlib.pyplot as plt

import gymnasium as gym

from rlforge.experiments import ExperimentRunner
from rlforge.agents.semi_gradient import DQNAgent

env = gym.make("MountainCar-v0")
agent = DQNAgent(learning_rate=0.001,
                 discount=0.99,
                 state_dim=2,
                 num_actions=env.action_space.n,
                 temperature=0.001,
                 network_architecture=[256],
                 target_network_update_steps=1,
                 num_replay=4,
                 experience_buffer_size=50000,
                 mini_batch_size=8)

runner = ExperimentRunner(env, agent)

results = runner.run_episodic(
    num_runs=10,
    num_episodes=50,
    max_steps_per_episode=10000
)

runner.summary(last_n=10)

============================================================
 Experiment Summary (Episodic)
============================================================
Runs: 10
Average runtime per run: 21.245 seconds
Episodes per run: 50
First episode mean reward: -2795.300
Last episode mean reward: -160.900
Overall mean reward: -418.594
Mean reward (last 10 episodes): -171.320
First episode mean steps: 2795.3
Last episode mean steps: 160.9
Overall mean steps: 418.6
============================================================

plt.plot(results['mean_rewards'])
plt.grid()

plt.plot(results['mean_steps'])
plt.grid()