diff --git a/obstacle_detector.py b/obstacle_detector.py
new file mode 100644
index 0000000..d6a6df7
--- /dev/null
+++ b/obstacle_detector.py
@@ -0,0 +1,89 @@
+#!usr/bin/env python3
+"""
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+Monash Nova Rover Team
+
+Takes in pointcloud data from camera, 
+  outputs average distance to obstacle in metres as FLoat32.
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+NODE: obstacle_detector
+TOPICS:
+  - /D435/depth/color/points [sensor_msgs.msg.PointCloud2]
+  - /obstacle_proximity [std_msgs.msg.Float32]
+SERVICES:
+  - /service_name [Service Type]
+ACTIONS: None
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+PACKAGE: 	obstacle_detector
+AUTHOR(S):	Alexander Li
+CREATION:	03/10/2021
+EDITED:		03/10/2021
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+TODO:
+ - 
+ - parse data
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+"""
+
+
+import rclpy
+from rclpy.node import Node
+from sensor_msgs.msg import PointCloud2
+from std_msgs.msg import Float32
+import struct
+from bitarray import bitarray
+
+
+# use this function for the recruit induction task (or write your own! this seems like a very messy way of doing it)
+def floatify_bytes(z_bytes):
+    """
+    :param lst of four bytes as numbers between 0-255
+    :return: the little endian float representation
+    """
+    bits = bin(z_bytes[0])[2:].ljust(8, "0") + bin(z_bytes[1])[2:].ljust(8, "0") + bin(z_bytes[2])[2:].ljust(8, "0") + bin(z_bytes[3])[2:].ljust(8, "0")
+    arr = bitarray(bits[::-1])
+    return struct.unpack(">f", arr)[0]
+
+
+class Obstacle_detector(Node):
+    def __init__(self):
+        super().__init__("obstacle_detector")
+        self.subscription = self.create_subscription(PointCloud2, "/D435/depth/color/points", self.callback, 100) # queue size 100
+        self.publisher = self.create_publisher(Float32, "/obstacle_proximity", 10)
+
+    def callback(self, msg: PointCloud2) -> None:
+        """
+        Publishes the average z-value of points in the point cloud within a 30cm x 30cm window in the x-y plane. 
+        :param msg: sensor_msgs.msg.PointCloud2
+        :return: None
+        """
+        sampling_sparseness = 100 # take only one point out of every [this number] of points.
+        sum_z = 0.0 # aggregate of z values for qualified points
+        num_z = 0.0 # number of qualified points
+
+        for i in range(msg.width//sampling_sparseness):
+            start_pos = 20*sampling_sparseness*i # start position within the byte list.
+
+            x = floatify_bytes(msg.data[start_pos:start_pos+4])
+            y = floatify_bytes(msg.data[start_pos+4:start_pos+8])
+            z = floatify_bytes(msg.data[start_pos+8:start_pos+12])
+
+            # only points within a 30cm x 30cm window are accepted.
+            if abs(x) < 0.15 and abs(y) < 0.15 and z > 0.3 and z < 3.0:
+                sum_z += z
+                num_z += 1
+
+        # create and publish average z value as FLoat32
+        z_average = Float32()
+        z_average.data = sum_z/num_z
+        self.publisher.publish(z_average)
+
+
+def main():
+    rclpy.init(args=None)
+    obstacle_detector = Obstacle_detector()
+    rclpy.spin(obstacle_detector)
+
+
+if __name__ == "__main__":
+    main()
\ No newline at end of file