Pubblicazioni

Autori:

Marzari, Luca; Corsi, Davide; Marchesini, Enrico; Farinelli, Alessandro

Titolo:

Curriculum learning for safe mapless navigation

Anno:

2022

Tipologia prodotto:

Poster

Tipologia ANVUR:

Poster

Lingua:

Inglese

Titolo del Convegno:

SAC '22: Proceedings of the 37th ACM/SIGAPP Symposium on Applied Computing

Luogo:

Virtual event

Periodo:

April 25 - 29, 2022

ISBN:

9781450387132

Intervallo pagine:

766-769

Parole chiave:

Deep Reinforcement Learning (DRL), Curriculum Learning (CL), Transfer of Learning (ToL), Fine-tuning, Mapless Navigation, Formal verification

Breve descrizione dei contenuti:

This work investigates the effects of Curriculum Learning (CL)-based approaches on the agent's performance. In particular, we focus on the safety aspect of robotic mapless navigation, comparing over a standard end-to-end (E2E) training strategy. To this end, we present a CL approach that leverages Transfer of Learning (ToL) and fine-tuning in a Unity-based simulation with the Robotnik Kairos as a robotic agent. For a fair comparison, our evaluation considers an equal computational demand for every learning approach (i.e., the same number of interactions and difficulty of the environments) and confirms that our CL-based method that uses ToL outperforms the E2E methodology. In particular, we improve the average success rate and the safety of the trained policy, resulting in 10% fewer collisions in unseen testing scenarios. To further confirm these results, we employ a formal verification tool to quantify the number of correct behaviors of Reinforcement Learning policies over desired specifications.

Id prodotto:

129718

Handle IRIS:

11562/1075086

ultima modifica:

25 ottobre 2024

Citazione bibliografica:

Marzari, Luca; Corsi, Davide; Marchesini, Enrico; Farinelli, Alessandro, Curriculum learning for safe mapless navigation  in - , Atti di "SAC '22: Proceedings of the 37th ACM/SIGAPP Symposium on Applied Computing" , Virtual event , April 25 - 29, 2022 , 2022 , pp. 766-769

Consulta la scheda completa presente nel repository istituzionale della Ricerca di Ateneo