Function reward_function

rl3/es_mujoco.py:141–169

(params, display=False)

def reward_function(params, display=False):
  model = ANN(D, M, K)
  model.set_params(params)

  env = gym.make(ENV_NAME)
  if display:
    env = gym.wrappers.Monitor(env, 'es_monitor')
  
  # play one episode and return the total reward
  episode_reward = 0
  episode_length = 0 # not sure if it will be used
  done = False
  state = env.reset()
  while not done:
    # display the env
    if display:
      env.render()

    # get the action
    action = model.sample_action(state)

    # perform the action
    state, reward, done, _ = env.step(action)

    # update total reward
    episode_reward += reward
    episode_length += 1

  return episode_reward


if __name__ == '__main__':