def add_customer(agent_id, time, belly_mean):
start_pos = random.choice(range(5))
belly_size = max([np.random.normal(loc=belly_mean, scale=2) / 2, 0.001])
entrance_occupied, _ = check_if_pos_empty(belly_size,
parkEntrances[start_pos], np.nan)
if entrance_occupied:
return agent_id
first_target = random.choice(attractions)
customers[agent_id] = Agent(index=agent_id,
entrance=parkEntrances[start_pos],
entrancesStr=parkEntrancesStr[start_pos],
entranceTime=time,
firstTarget=first_target,
mapSize=mapSize,
belly=belly_size)
path = ParkMap.get_path_to_next_pos(customers[agent_id])
customers[agent_id].path = path
customersInPark.append(agent_id)
ParkMap.agentsLocation[agent_id] = customers[agent_id].pos
return agent_id + 1
init_holding = 1
initialcash = 20000
theta = 75
gene_length = 64
risk_coef = 0.5
num_strategies = 80
max_stockprice = 200
min_stockprice = 0.01
min_excess = 0.005
eta = 0.005
specialist_iterations = 10
the_market = Stock(num_shares, init_price, dividend_startvalue, rho, noise_sd,
d_bar, interest_rate)
agents = [
Agent(init_holding, initialcash, num_strategies, theta, gene_length,
risk_coef, num_shares, init_price, dividend_startvalue, rho,
noise_sd, d_bar, interest_rate) for _ in range(100)
]
the_specialist = Specialist(max_stockprice, min_stockprice, min_excess, eta,
specialist_iterations, num_shares, init_price,
dividend_startvalue, rho, noise_sd, d_bar,
interest_rate)
for t in range(1000):
the_market.advance_arprocess()
the_specialist.clear_market(agents)
ma = the_market.calculate_ma(query='dividend', period=50)
def main():
f = open("parameters2.txt", "r")
flines = list(map(lambda x: x.replace('\n', '').split(), f.readlines()))
n = int(flines[0][1])
deadLine = float(flines[1][1])
cDeadLine = deadLine
graph = {}
j = 2
totalNumOfPpl = 0
for i in range(2, 2 + n):
ppl = int(flines[i][1][1:]) if len(flines[i]) > 1 else 0
graph[flines[i][0][1:]] = {"p": ppl, "e": []}
j = j + 1
totalNumOfPpl += ppl
for i in range(j + 1, len(flines)):
v1 = "V" + flines[i][1]
v2 = "V" + flines[i][2]
w = int(flines[i][3][1:])
graph[v1]["e"].append({"v": v2, "w": w, "blocked": False})
graph[v2]["e"].append({"v": v1, "w": w, "blocked": False})
print("please enter number of agents:")
num_of_agents = int(input())
print("please enter agents details: ")
agents_details = input().split(',')
agentDetails = list(map(lambda x: x.split(' '), agents_details))
agentsList = list()
for i in range(0, num_of_agents):
agentsList.append(
Agent(agentDetails[i][0], agentDetails[i][1], len(graph), L))
# main loop
print("Agents: \n", agentsList)
print("Graph: \n", graph)
while deadLine > 0 and totalNumOfPpl > 0 and not allTerminated(agentsList):
for i in agentsList:
if i.terminated:
continue
elif i.stepsLeft > 0:
i.stepsLeft -= 1
elif i.calcTime > 0:
i.calcTime -= 1
elif i.type == 's':
saboAct(i, graph)
else:
# we reached vertice, evacuating ppl
verPpl = graph[i.currentPosition]["p"]
totalNumOfPpl -= verPpl
i.peopleEvacuated += verPpl
graph[i.currentPosition]["p"] = 0
if totalNumOfPpl == 0:
break
#finding next vertice to travel
prevVer = i.currentPosition
i.currentPosition = getNextStep(i, graph)
print("Next step: ", i.currentPosition)
if not i.currentPosition:
i.terminated = True
else:
i.stepsLeft = getEdgeWeight(graph, prevVer,
i.currentPosition)
i.numOfActions += 1
deadLine -= 1
feedback = "Well Done!!! Agent: "
if agentsList[0].peopleEvacuated == 0:
feedback = "Too Bad.. You could've done better, "
print(feedback, agentsList[0].type, "\nevacuated ",
agentsList[0].peopleEvacuated, " people! And it took ",
cDeadLine - deadLine, " rounds.")
if __name__ == '__main__':
environment_name = 'CarRacing-v0'
env = environment(environment_name, img_dim, num_stack, num_actions,
render, lr)
num_states = img_dim
print(env.env.action_space.shape)
action_dim = env.env.action_space.shape[0]
assert action_list.shape[
1] == action_dim, "length of Env action space does not match action buffer"
num_actions = action_list.shape[0]
# Setting random seeds with respect to python inbuilt random and numpy random
random.seed(901)
np.random.seed(1)
agent = Agent(num_states, num_actions, img_dim, model_path)
randomAgent = RandomAgent(num_actions)
print(test_mode, train_mode)
try:
#Train agent
if test_mode:
if train_mode:
print("Initialization with random agent. Fill memory")
while randomAgent.exp < memory_size:
env.run(randomAgent)
print(randomAgent.exp, "/", memory_size)
agent.memory = randomAgent.memory
randomAgent = None
# the action space
action_size = brain.vector_action_space_size
print('Number of actions:', action_size)
# the state space
state = env_info.vector_observations[0]
state_size = len(state)
print('States have length:', state_size)
#######################################
########### Agent Setting ###########
agent = Agent(state_size,
action_size,
seed=0,
hidden_layers=hidden_layers,
drop_p=drop_p,
method=method,
Dueling=Dueling)
print('-------- Model structure --------')
print('method :', method)
print(agent.qnetwork_local)
print('---------------------------------')
#######################################
scores = [] # list containing scores from each episode
scores_window = deque(maxlen=100) # last 100 scores
eps = eps_start # initialize epsilon
isFirst = True
def wallet_to_agents(self, wallets):
return [Agent(wallet) for wallet in wallets]