ModelLogs.md

Model Registry

model1

Trained with with the improved reward function, DISTANCE_PENALTY=4, MINOR_SAFETY_PENALTY=1 and MAJOR_SAFETY_PENALTY=5. No noise during training.

model2

Same reward function. Trained with unbiased noise with standard deviation 0.1

model3

Same reward function. Trained with unbiased noise with standard deviation 1.0