First of all, let's make some changes to the current simulation so that we are able to work with perception better. We will create a new file named table.urdf in our workspace and add the following code into that file:

<robot name="simple_box"> 
  <link name="my_box"> 
    <inertial> 
      <origin xyz="0 0 0.0145"/> 
      <mass value="0.1" /> 
      <inertia  ixx="0.0001" ixy="0.0"  ixz="0.0"  iyy="0.0001"  iyz="0.0"  izz="0.0001" /> 
    </inertial> 
    <visual> 
      <origin xyz="-0.23 0 0.215"/> 
      <geometry> 
        <box size="0.47 0.46 1.3"/> 
      </geometry> 
    </visual> 
    <collision> 
      <origin xyz="-0.23 0 0.215"/> 
      <geometry> 
        <box size="0.47 0.46 1.3"/> 
      </geometry> 
    </collision> 
  </link> 
  <gazebo reference="my_box"> 
    <material>Gazebo/Wood</material> 
  </gazebo> 
  <gazebo> 
    <static>true</static> 
  </gazebo> 
</robot> 

Next, we will execute the following command in order to spawn an object right in front of the fetch robot:

$ rosrun gazebo_ros spawn_model -file /home/user/catkin_ws/src/table.urdf -urdf -x 1 -model my_object

The output is shown in the following diagram:

Later, we will execute the following command in order to move the fetch robot's head so that it points to the newly spawned object:

$ roslaunch fetch_gazebo_demo move_head.launch

We can also launch RViz and add the corresponding element in order to visualize the PointCloud of the camera, which is shown in the following screenshot: