added back 0 reward if off road to reproduce previous behaviour

Author: JingfeiPeng

examples/game_of_tag/tag_adapters.py

@@ -179,8 +179,12 @@ def predator_reward_adapter(observations, env_reward_signal):
     if events.off_road:
         rew -= global_rewards.offroad
 
-    # print(f"predator {ego.id.split('-')[0]} reward: {rew} distance {distance_to_target}, laneIndex: {ego.lane_index}")
-    return rew
+    # if no prey vehicle avaliable, have 0 reward instead
+    # TODO: Test to see if this is neccessary
+    prey_vehicles = list(filter(
+        lambda v: _is_vehicle_wanted(v.id, PREY_IDS), observations.neighborhood_vehicle_states,
+    ))
+    return rew if len(prey_vehicles) > 0 else 0
 
 
 def prey_reward_adapter(observations, env_reward_signal):
@@ -212,4 +216,9 @@ def prey_reward_adapter(observations, env_reward_signal):
     if events.off_road:
         rew -= global_rewards.offroad
 
-    return rew
+    # if no predator vehicle avaliable, have 0 reward instead
+    # TODO: Test to see if this is neccessary
+    predator_vehicles = list(filter(
+        lambda v: _is_vehicle_wanted(v.id, PREDATOR_IDS), observations.neighborhood_vehicle_states,
+    ))
+    return rew if len(predator_vehicles) > 0 else 0