Retrain the robot with a reward that penalizes quick action changes.

ll7 / robot_sf_ll7

robot_sf in the ll7 namespace.

GNU General Public License v3.0

1 stars 1 forks source link

Retrain the robot with a reward that penalizes quick action changes. #48

Closed ll7 closed 1 month ago

ll7 commented 2 months ago

Currently, the robot creates many small actions. This is unrealistic and energy consuming. A new reward should be designed to punish many changes in the picked action.

ll7 commented 2 months ago

A new model is trained, but the performance is not optimal: I don't know why the collision rates aren't better.