This repository contains implementations of the methods presented in manuscript by Chinchali et al., which is currently under review.
-
overlaid belief distro:
- KL solid, H dashed
- overlay_learning_curve.sh
- python $CODE_DIR/overlaid_belief_distro.py --KL_results_dir $KL_RESULTS_DIR --H_results_dir $ENTROPY_RESULTS_DIR --config_file $CONF_FILE --output_results_dir $OUTPUT_RESULTS_DIR
-
overlaid learning curves
- overlay_learning_curve.sh
- python -i $CODE_DIR/overlaid_learning_curve.py --KL_results_dir $KL_RESULTS_DIR --entropy_results_dir $ENTROPY_RESULTS_DIR --output_results_dir $OUTPUT_RESULTS_DIR
-
cart drone images and speed plots
- drone_data/publication_plots.sh
- relies on pkls per agent of type:
- agent_239_scene_gates_video_1.pkl
- to generate these pkl files call:
- python idwithtasks/RL/drone_data/single_car_plot.py