Why?
Many algorithms use a seed before starting an experiment. And the lack of that feature causes to throw an error.
What?
Added seeding functionality for Madras env.
Testing:
import gym
import MADRaS
env_name = 'Madras-v0'
env = gym.make(env_name)
a = env.seed(445)
print(a)
env.reset()
for i in range(1000):
print("STER: {}".format(i))
env.step(env.action_space.sample()) # take a random action
print('#################OVER####################')
env.reset()
for i in range(1000):
print("STER: {}".format(i))
ob, r, done, info = env.step(env.action_space.sample()) # take a random
env.close()
NOTE:
As of now the seed is not used by torcs before starting a race but we can incorporate it later, in maybe starting race at random positions.
Why? Many algorithms use a seed before starting an experiment. And the lack of that feature causes to throw an error.
What? Added seeding functionality for Madras env.
Testing:
NOTE: As of now the seed is not used by torcs before starting a race but we can incorporate it later, in maybe starting race at random positions.