use cnn_to_mlp to allow training of racecar using (extremely simplified) ZED camera pixel data using OpenAI baselines.

add a red sphere to make training a bit easier for now.

data/sphere2red.urdf

@@ -0,0 +1,32 @@
+<?xml version="0.0" ?>
+<robot name="urdf_robot">
+  <link name="baseLink">
+  	<contact>
+      <rolling_friction value="0.03"/>
+      <spinning_friction value="0.03"/>
+    </contact>
+
+    <inertial>
+      <origin rpy="0 0 0" xyz="0 0 0"/>
+       <mass value="10.0"/>
+       <inertia ixx="1" ixy="0" ixz="0" iyy="1" iyz="0" izz="1"/>
+    </inertial>
+    <visual>
+      <origin rpy="0 0 0" xyz="0 0 0"/>
+      <geometry>
+				<mesh filename="sphere_smooth.obj" scale="0.5 0.5 0.5"/>      
+      </geometry>
+      <material name="red">
+        <color rgba="1 0.2 0.2 1"/>
+        <specular rgb="1 1 1"/>
+      </material>
+    </visual>
+    <collision>
+      <origin rpy="0 0 0" xyz="0 0 0"/>
+      <geometry>
+	 <sphere radius="0.5"/>
+      </geometry>
+    </collision>
+  </link>
+</robot>
+

data/sphere_smooth.mtl

@@ -0,0 +1,10 @@
+# Blender MTL File: 'None'
+# Material Count: 1
+
+newmtl None
+Ns 0
+Ka 0.000000 0.000000 0.000000
+Kd 1. 0.2 0.2
+Ks 0.8 0.8 0.8
+d 1
+illum 2