Added CLI flags & Model Outputs - Githubissues

geoffrey-anto / drones-attention-based-lstm-deep-q-network-rpp

A Realtime Path Planning AIRSIM Simulation Using Deep Q Learning With Attention Based LSTM

MIT License

1 stars 0 forks source link

Added CLI flags & Model Outputs #2

Closed geoffrey-anto closed 6 days ago

geoffrey-anto commented 6 days ago

This pull request includes significant updates to the main.py file to support different model types, new Makefile targets for training and testing, and various changes to configuration parameters and output files.

Main Application Changes:

main.py: Added support for selecting model types (DQNLSTM or DQNALSTM) via command-line arguments and updated the training and testing logic accordingly.

Configuration Updates:

src/Config.py: Adjusted TARGET_ARRIVAL_THRESHOLD to 0.9 km and increased NUM_OBSTACLES to 12. [1] [2]

Build and Deployment:

makefile: Added new targets for training and testing models (train, test-lstm, test-attention).

Output Files:

Updated output files to reflect new metrics and results from the simulations:
- outputs/control/A-9/output.txt
- outputs/control/C-10/output.txt
- outputs/control/D-11/output.txt
- outputs/control/E-12/output.txt
- outputs/current/A-9/output.txt
- outputs/current/B-10/output.txt
- outputs/current/C-11/output.txt
- outputs/current/D-12/output.txt

Code Logic:

src/DQNALSTM.py: Updated the act method to always use model predictions during testing by setting the exploration probability to 0.