Add tensor input type for arbitrary dimensional observation (#125)

* Allow arbitrary dimensional observation (non vector or image)
* Added creating PlanarMapsObservationSpace to GymEnvironment when number of channels is not 1 or 3

rl_coach/spaces.py

@@ -183,7 +183,7 @@ class ObservationSpace(Space):
 class VectorObservationSpace(ObservationSpace):
     """
     An observation space which is defined as a vector of elements. This can be particularly useful for environments
-    which return measurements, such as in robotic environmnets.
+    which return measurements, such as in robotic environments.
     """
     def __init__(self, shape: int, low: Union[None, int, float, np.ndarray]=-np.inf,
                  high: Union[None, int, float, np.ndarray]=np.inf, measurements_names: List[str]=None):
@@ -197,6 +197,16 @@ class VectorObservationSpace(ObservationSpace):
         super().__init__(shape, low, high)
 
 
+class TensorObservationSpace(ObservationSpace):
+    """
+    An observation space which defines observations with arbitrary shape. This can be particularly useful for
+    environments with non image input.
+    """
+    def __init__(self, shape: np.ndarray, low: -np.inf,
+                 high: np.inf):
+        super().__init__(shape, low, high)
+
+
 class PlanarMapsObservationSpace(ObservationSpace):
     """
     An observation space which defines a stack of 2D observations. For example, an environment which returns