success check for high level policy

backends/kerasrl_learner.py

@@ -383,9 +383,10 @@ class DQNLearner(LearnerBase):
                             termination_reason_counter[termination_reason] += 1
                         else:
                             termination_reason_counter[termination_reason] = 1
-                    env.reset()
-                    if episode_reward >= success_reward_threshold:
+                    #TODO: remove below env-specific code
+                    if env.env.goal_achieved:
                         success_count += 1
+                    env.reset()
                     print("Episode {}: steps:{}, reward:{}".format(
                         n + 1, step, episode_reward))