diff --git a/vmas/scenarios/navigation.py b/vmas/scenarios/navigation.py
index a40f0a0a..e8b71531 100644
--- a/vmas/scenarios/navigation.py
+++ b/vmas/scenarios/navigation.py
@@ -43,7 +43,8 @@ def make_world(self, batch_dim: int, device: torch.device, **kwargs):
         ScenarioUtils.check_kwargs_consumed(kwargs)
 
         self.min_distance_between_entities = self.agent_radius * 2 + 0.05
-        self.world_semidim = 1
+        self.world_semidim_x = 1 if self.x_semidim is None else self.x_semidim
+        self.world_semidim_y = 1 if self.y_semidim is None else self.y_semidim
         self.min_collision_distance = 0.005
 
         assert 1 <= self.agents_with_same_goal <= self.n_agents
@@ -135,8 +136,8 @@ def reset_world_at(self, env_index: int = None):
             self.world,
             env_index,
             self.min_distance_between_entities,
-            (-self.world_semidim, self.world_semidim),
-            (-self.world_semidim, self.world_semidim),
+            (-self.world_semidim_x, self.world_semidim_x),
+            (-self.world_semidim_y, self.world_semidim_y),
         )
 
         occupied_positions = torch.stack(
@@ -152,8 +153,8 @@ def reset_world_at(self, env_index: int = None):
                 env_index=env_index,
                 world=self.world,
                 min_dist_between_entities=self.min_distance_between_entities,
-                x_bounds=(-self.world_semidim, self.world_semidim),
-                y_bounds=(-self.world_semidim, self.world_semidim),
+                x_bounds=(-self.world_semidim_x, self.world_semidim_x),
+                y_bounds=(-self.world_semidim_y, self.world_semidim_y),
             )
             goal_poses.append(position.squeeze(1))
             occupied_positions = torch.cat([occupied_positions, position], dim=1)