# Excavator Notes

### Result: **Didn't** start training or training **interupted**
* Used all the RL features, PPO + LSTM + behavioral_cloning + GAIL + curiosity
* More than 2 training environments

### To start the training successfully
* Reduced number of RL features and reduced some hyperparameter values
* Reduced number objects(like stones) in the environment

### 9-10 (Fri) changes
* Limited the range of swing-rotation, so that random space reduces
* changed starting swing-angle to an angle between the container and the truck, so that it is faster to learn the important movements
* Reduced the number of stones and increased its size so that training can be faster and episode length is reduced
* Added a new Episode End condition such that moving only 1 stone is enough to get max reward
*