Add Waypoints back to reward function

[Gizmotronn]

@wakeuplearn: "nothing i can think of right now. unless you can do a quick search on waypoints and reinforcement learning. i’ve found a lot of things online, but there still may be something I’m missing. waypoints actually do seem to help."

We're going to add waypoints back to the reward function...probably just 2 for now

We'll merge it with #11 & #10

[Gizmotronn]

I see that you added 2 more waypoints here: https://github.com/EXYNOS-999/AWS_JPL_OSR_DRL/commit/87ce1b2515036a17bc39ed8a246884463c5459bc#diff-007f71bedbc983e8560bc575cd52a5c4






Edwin M  2:58 AM
Yes , see if you can make more efficient . Namely , having specific coordinates for the waypoints can be considered bad if the course changes . Which they may do in the evaluation . 

Liam ARBUCKLE:slack_call:  2:59 AM
So just maybe fractions, that will be say 1/3, 2/3, etc of the distance to the checkpoint?

Edwin M  5:16 AM
yes,exactly.. that’s what I have just tried. Seems to work better.

Liam ARBUCKLE:slack_call:  9:26 AM
Okay, would you lile me to add some different sets of fractions? Could work...

Edwin M  9:26 AM
its worth a try

Liam ARBUCKLE:slack_call:  10:21 AM
So what sets have you done?

acordrobotics.slack.com