Merge pull request #126 from una-auxme:115-feature-research-lidar-data

115 feature research lidar data

doc/03_research/02_perception/LIDAR_data.md

@@ -0,0 +1,55 @@
+# LIDAR-Data
+
+This File discusses where the LIDAR-Data comes from, how its processed and how we could possibly use it.
+
+## Origin
+
+LIDAR-Data comes in Pointclouds from a specific LIDAR-Topic.
+
+`rospy.Subscriber(rospy.get_param('~source_topic', "/carla/hero/LIDAR"),
+                         PointCloud2, self.callback)`
+
+Read more about the LIDAR-Sensor [here](https://github.com/una-auxme/paf23/blob/main/doc/06_perception/03_lidar_distance_utility.md)
+
+## Processing
+
+The goal is to identify Objects and their distance. Therefor we need to calculate distances from the pointcloud data.
+To do this the lidar-distance node first converts pointcloud data to an array, which contains cartesian coordinates.
+
+`paf23-agent-1            |  (76.12445   , -1.6572031e+01, 13.737187  , 0.7287409 )`
+
+`paf23-agent-1            |  (71.9434    , -1.8718828e+01, 13.107929  , 0.7393809 )`
+
+`paf23-agent-1            |  (-0.3482422 , -1.6367188e-02, -0.20128906, 0.99839103)`
+
+`paf23-agent-1            |  (-0.3486328 , -1.4062500e-02, -0.20152344, 0.99838954)`
+
+`paf23-agent-1            |  (-0.35070312, -2.3828126e-03, -0.2025    , 0.99838144)`
+
+The first three values of each row correspon to x, y, z.
+
+x - the X Cartesian coordinate of a point (float32)
+
+y - the Y Cartesian coordinate of a point (float32)
+
+z - the Z Cartesian coordinate of a point (float32)
+
+It wasn´t specified anywhere, what the 4th values represents. My best guess is some sort of intensity.
+
+## Distance Calculation
+
+The distance to a point is calculated by the euclidian distance to (0,0,0) for every point in the point cloud.
+
+`distances = np.array(
+            [np.linalg.norm(c - [0, 0, 0]) for c in coordinates_xyz])`
+
+They then publish the minimum and maximum distance.
+
+## Open questions
+
+1. Currently we have no specified unit.
+    **Is a relative distance enough?**
+
+2. How do we translate cartesian coordinates to pixel in the Image and vice-versa?
+
+   We should be able to set the bounding box of the LIDAR-PointCloud2 to the camera-calibration.