def test_sarsa_agent():
env = RoRoDeck()
env.rows = 12
env.lanes = 12
env.reset()
# set all units to equal length
env.vehicle_data[4][1] = 2
env.vehicle_data[4][3] = 2
agent = SARSA(env, None)
assert len(agent.q_table.keys()) == 0
agent.number_of_episodes = 1
agent.train()
assert len(agent.q_table.keys()) == 61
def test_max_action_method_sarsa():
env = RoRoDeck()
env.rows = 12
env.lanes = 12
state = env.reset()
agent = SARSA(env, None)
agent.q_table[state.tobytes()] = np.zeros(4)
assert np.count_nonzero(agent.q_table[state.tobytes()]) == 0
agent.q_table[state.tobytes()][2] = 1
agent.q_table[state.tobytes()][3] = 2
assert agent.max_action(state, env.possible_actions) == 3
env.possible_actions = np.array([0, 1, 2])
assert agent.max_action(state, env.possible_actions) == 2