fixing learning issues

torch_training/training_navigation.py

@@ -125,6 +125,7 @@ def main(argv):
         # Build agent specific observations
         for a in range(env.get_num_agents()):
             agent_obs[a] = normalize_observation(obs[a], observation_radius=10)
+            agent_obs_buffer[a] = agent_obs[a].copy()
 
         # Reset score and done
         score = 0
@@ -136,11 +137,13 @@ def main(argv):
             for a in range(env.get_num_agents()):
                 if env.agents[a].speed_data['position_fraction'] < 0.001:
                     register_action_state[a] = True
+                    action = agent.act(agent_obs[a], eps=eps)
+                    action_prob[action] += 1
+                    if step == 0:
+                        agent_action_buffer[a] = action
                 else:
                     register_action_state[a] = False
-
-                action = agent.act(agent_obs[a], eps=eps)
-                action_prob[action] += 1
+                    action = 0
                 action_dict.update({a: action})
 
             # Environment step
@@ -150,6 +153,7 @@ def main(argv):
             for a in range(env.get_num_agents()):
                 agent_next_obs[a] = normalize_observation(next_obs[a], observation_radius=10)
                 cummulated_reward[a] += all_rewards[a]
+
             # Update replay buffer and train agent
             for a in range(env.get_num_agents()):
                 if done[a]: