June 12th, 2018

First, let me start by saying I am a graduate in Computer Science and have been a PHP/MySQL developer for 12 years. I have not read or researched self-driving algorithms before. This algorithm is just an idea of how I would approach the problem of detecting objects of concern in a scene.

In development, we use version control systems. These systems generally depend on one command very heavily: diff

The diff of a file outputs all the changes to the file highlighting lines removed and lines added.

What you would need are 2 data sets to compare:

  1. A snapshot of all objects in the scene from history
  2. A snapshot of all objects in the scene in real-time

By running a diff on these datasets, you will return a set of objects that were not in the scene before but are now.  This subset of objects are the exact objects we need to compute trajectories for and detect collision with the car’s trajectory.

But what could happen is that by some strange coincidence, two objects appear in the exact same spot that they did previously.  Such as a car of the same make/model appearing 2 car lengths ahead in the same location as the original dataset.  These would raise false negatives and if that car suddenly stopped, it could ignore it.

To improve the dataset, you could run a diff for all snapshots of objects in a scene throughout history against each other.  This would reduce everything to a common subset of fixed elements.   Let U define the universal set of fixed elements.

When a vehicle enters a scene, every 30 seconds it would load the next 1 minute of scene data, a subset of U with respect to ~the next mile.  Let’s denote this set as A.

As it comes into an area we take a snapshot of S[lat,lng,elev] as close to a point in set A[lat,lng,elev] as possible.

Bash Code:
$ObjectsOfConcern = S[lat,lng,elev] diff A[lat,lng,elev]  | grep '>'

Posted In:

June 12th, 2018

ABOUT THE AUTHOR:
I am TSLA Long. Model 3 Owner. Brother of a Model 3 owner. Son of a Model S owner and Tesla enthusiast. Tesla Referral Code