search

ROBOTIS-GIT / turtlebot3_machine_learning

Apache License 2.0
121 stars 82 forks source link

An objection for argument 'distance_rate' in environment code #44

Open AssassinCrow opened 4 years ago

AssassinCrow commented 4 years ago

Hello, I'm working on ROS KInetic with Ubuntu 16.04.

Anyway, according to here, the distance_rate which corresponds to reward for distance is bigger than 2 in the case of the current goal-robot distance is shorter than absolute goal-distance.

But in environment code, the distance_rate argument is smaller in the upper case; in other words, I have a feeling that it should be modified like,

from distance_rate = 2 ** (current_distance / self.goal_distance) to distance_rate = 2 ** (self.goal_distance / current_distance)

I think the reward would get bigger and bigger when the robot gets farther from the goal point if it moves following the default code.

Would you happen to teach me why did you set the distance_rate like that?

Thanks in advance. :)

valderrama14 commented 4 years ago

Hi, was this issue ever answered in private? I'd actually like to know this, because the observation seems valid.