for ii, p in enumerate(generate_buyer_prices_paper(20)):
buyers.append(LinearBlackBoxBuyer('b' + str(ii), p, noisy=True))
agents = sellers + buyers
for ag in agents:
ag.new_game()
ROUNDS_PER_GAME = 10
market_env = MarketEnvironment(
sellers=sellers,
buyers=buyers,
max_time=10, ## not the same as rounds per game!!
matcher=RandomMatcher(reward_on_reference=True))
init_observation = market_env.reset()
round_avg = 0.
offer_avg = 0.
time_avg = 0.
records['demands'].append([])
records['rewards'].append([])
records['prices'].append(seller_agent.reservation_price)
for n_round in range(10):
init_observation = market_env.reset()
observations = {k.agent_id: None for k in agents}
done = {k.agent_id: False for k in agents}
reward_hist = []
agent.coefs = np.array([0.05, 0.95] + [0]*(size_coefs - 2))
# Reset agents' rewards and observations
for agent in sellers:
agent.reward = 0.0
agent.observations = {}
for agent in buyers:
agent.reward = 0.0
agent.observations = {}
# Loop over rounds
for r in range(n_round):
print("ROUND", r, '-----------------------------------------------')
# Reset market environment
market_env.reset()
# Initial offers are generated
current_offers = {}
for agent in sellers:
current_offers[agent.agent_id] = np.random.normal(200, 5)
for agent in buyers:
current_offers[agent.agent_id] = np.random.normal(100, 5)
# Loop over time steps
i = 0
while market_env.if_round_done is False:
print(i, '-------')
i += 1
# Environment calculates what happens
market_env.step(current_offers)