cumsum posture reward

zyb20230403 · Jun 7, 2022 · 84498f4 · 84498f4
1 parent 3dccc06
commit 84498f4
Showing 1 changed file with 9 additions and 6 deletions.
diff --git a/envs/JSBSim/reward_functions/posture_reward.py b/envs/JSBSim/reward_functions/posture_reward.py
@@ -1,4 +1,5 @@
 import numpy as np
+from wandb import agent
 from .reward_function_base import BaseRewardFunction
 from ..utils.utils import get_AO_TA_R
 
@@ -33,15 +34,17 @@ def get_reward(self, task, env, agent_id):
  Returns:
  (float): reward
  """
+ new_reward = 0
  # feature: (north, east, down, vn, ve, vd)
  ego_feature = np.hstack([env.agents[agent_id].get_position(),
  env.agents[agent_id].get_velocity()])
- enm_feature = np.hstack([env.agents[agent_id].enemies[0].get_position(),
- env.agents[agent_id].enemies[0].get_velocity()])
- AO, TA, R = get_AO_TA_R(ego_feature, enm_feature)
- orientation_reward = self.orientation_fn(AO, TA)
- range_reward = self.range_fn(R / 1000)
- new_reward = orientation_reward * range_reward
+ for enm in env.agents[agent_id].enemies:
+ enm_feature = np.hstack([enm.get_position(),
+ enm.get_velocity()])
+ AO, TA, R = get_AO_TA_R(ego_feature, enm_feature)
+ orientation_reward = self.orientation_fn(AO, TA)
+ range_reward = self.range_fn(R / 1000)
+ new_reward += orientation_reward * range_reward
  return self._process(new_reward, agent_id, (orientation_reward, range_reward))
 
  def get_orientation_function(self, version):