Offroad-Path-Planning

Datasets

All used datasets can be found at the following links with associated access procedures.

1. Massachusetts Road Dataset
2. DeepGlobe Dataset
3. CAVS Dataset

How to run Image Segmentation models

All image segmentation training and testing codes are written in python notebook files and can be found under Colab directory. The codes can be run through Google Colab (Online), Jupyter Notebook, or any suitable IDE with required data and library configuration.

Aerial Traversability Prediction Data

Pre-computed segmentation results can be downloaded from the following Google Drive links.

1. Massachusetts Road Dataset
2. DeepGlobe Dataset
3. CAVS Dataset

Each dataset comes with 4 subfolders:

• the input images
• binary traversability predictions
• binary traversability ground truths
• traversability probability matrices

Uncertainty-aware Replanning A* (URA*)

To run URA* in static mode (i.e. to generate the initial path) for the Massachusetts Road Dataset:

python3 main.py --im_folder datasets/sample_predictions_ensemble_massachuests/predictions  --sat_folder datasets/sample_predictions_ensemble_massachuests/original_image  --pred_matrix_folder datasets/sample_predictions_ensemble_massachuests/prediction_matrix/ --gt_folder datasets/sample_predictions_ensemble_massachuests/ground_truth --logged_points data/updated_mass_points.csv --output_image_path mass_output --csv_file_name mass_output/STATIC/results.csv --path_planners URA --static_or_dynamic STATIC

This command will save the path planning results to the output directory and save evaluation metrics in an output CSV file.

To run URA* in static mode (i.e. to generate the initial path) for the DeepGlobe Dataset:

python3 main.py --im_folder datasets/Filtered_Ensemble_DeepGlobe/Prediction  --sat_folder datasets/Filtered_Ensemble_DeepGlobe/Satellite\ Image/  --pred_matrix_folder datasets/Filtered_Ensemble_DeepGlobe/Prediction\ Matrix/ --gt_folder datasets/Filtered_Ensemble_DeepGlobe/Ground\ Truth/ --logged_points data/deepglobepoints.csv --output_image_path deepglobe_output --csv_file_name deepglobe_output/STATIC/results.csv --path_planners URA --static_or_dynamic STATIC    

To run URA*, A*, A**, RRT* for the CAVS Dataset:

python3 main.py --im_folder datasets/sample_predictions_Ensemble_CAVS_V3/predictions  --sat_folder datasets/sample_predictions_Ensemble_CAVS_V3/original_image/  --pred_matrix_folder datasets/sample_predictions_Ensemble_CAVS_V3/prediction_matrix/ --gt_folder datasets/sample_predictions_Ensemble_CAVS_V3/ground_truth/ --logged_points data/cavs_points.csv --output_image_path cavs_output --csv_file_name cavs_output/STATIC/results.csv --path_planners A ASTARTHRESHOLD RRTSTAR URA --static_or_dynamic STATIC

To run URD*, RRA*, D*-Lite for the Massachusetts Road Dataset:

python3 main.py --im_folder datasets/sample_predictions_ensemble_massachuests/predictions  --sat_folder datasets/sample_predictions_ensemble_massachuests/original_image  --pred_matrix_folder datasets/sample_predictions_ensemble_massachuests/prediction_matrix/ --gt_folder datasets/sample_predictions_ensemble_massachuests/ground_truth --logged_points data/updated_mass_points.csv --output_image_path mass_output --csv_file_name mass_output/DYNAMIC/results.csv --path_planners URD RRA DLITE --static_or_dynamic DYNAMIC

To run output the metrics from the csv file generated by main.py:

python3 calculate_metrics.py --static_or_dynamic STATIC (for static runs) or DYNAMIC (for dynamic runs) --csv_file_name --csv_file_name cavs_output/STATIC/results.csv