feup-thesis-rt-drl

Code for my dissertation at FEUP, related to optimizing task scheduling on a real-time system. Supervised by professors Pedro Souto and Pedro Diniz.

You can read the dissertation with a detailed explanation of the work here.

The methodology was improved for the conference paper draft, which can be read here.

Compilation

Make sure you have an updated Rust toolchain in your machine. Then, simply:

cargo build --release

Testing

cargo test --release

Executing

First, set the environment variables:

• NUMBER_RUNNABLES: number of (automotive) runnables to be generated in each task set
• TRAIN_INSTANTS: number of simulated seconds to train each model
• TEST_INSTANS: number of simulated seconds to test each model
• NUMBER_TEST_SIMULATIONS: number of test simulations for testing each model
• THREAD_POOL_SIZE: number of models to be trained simultaneously

Then, you can test the system for 1 task set by simply running the program:

cargo run --release

If you want to collect statistics for more task sets, use the bundled Python scripts.