minor bugfixes, still looking for infinite loop

examples/training_navigation.py

@@ -54,7 +54,8 @@ for trials in range(1, n_trials + 1):
 
     # Run episode
     for step in range(100):
-        #env_renderer.renderEnv(show=True)
+        if trials >= 114:
+            env_renderer.renderEnv(show=True)
 
         # Action
         for a in range(env.number_of_agents):

flatland/envs/rail_env.py

@@ -547,7 +547,7 @@ class RailEnv(Environment):
         global_reward = 1 * beta
 
         # Reset the step rewards
-        self.rewards_dict = {}
+        self.rewards_dict = dict()
         for handle in self.agents_handles:
             self.rewards_dict[handle] = 0