Logiciel Année : 2022

Packet Routing Simulator for Multi-Agent Reinforcement Learning (PRISMA) (Version v0.1)

Résumé

PRISMA (Packet Routing Simulator for Multi-Agent Reinforcement Learning) is a network simulation playground for developing and testing Multi-Agent Reinforcement Learning (MARL) solutions for dynamic packet routing (DPR). This framework is based on the OpenAI Gym toolkit and the ns-3 simulator. The OpenAI Gym is a toolkit for RL widely used in research. The network simulator ns–3 is a standard library, which may provide useful simulation tools. It generates discrete events and provides several protocol implementations. Moreover, the NetSim implementation is based on ns3-gym, which integrates OpenAI Gym and ns-3. The main contributions of this framework: 1) A RL framework designed for specifically the DPR problem, serving as a playground where the community can easily validate their own RL approaches and compare them. 2) A more realistic modelling based on: (i) the well-known ns-3 network simulator, and (ii) a multi-threaded implementation for each agent. 3) A modular code design, which allows a researcher to test their own RL algorithm for the DPR problem, without needing to work on the implementation of the environment.

HAL

Cite hal-03709948 Objet présenté à une conférence Redha A. Alliche, Tiago da Silva Barros, Ramon Aparicio-Pardo, Lucile Sassatelli. PRISMA: A Packet Routing Simulator for Multi-Agent Reinforcement Learning. 4th Intl Workshop on Network Intelligence collocated with IFIP Networking 2022, Jun 2022, Catania, Italy. ⟨10.23919/IFIPNetworking55013.2022.9829797⟩. ⟨hal-03709948⟩

Citer

Redha A. Alliche, Tiago da Silva Barros, Ramon Aparicio-Pardo, Lucile Sassatelli. Packet Routing Simulator for Multi-Agent Reinforcement Learning (PRISMA) (Version v0.1). 2022, ⟨swh:1:dir:53e77bdd5593e2ef85805c8520626cfddb113fa9;origin=https://hal.archives-ouvertes.fr/hal-03998842;visit=swh:1:snp:c4feb2a2b2254e41f7187021d2905e65d12d07a6;anchor=swh:1:rel:29f812793e523375296b6ff52ffc50ce68250a98;path=/⟩. ⟨hal-03998842⟩
380 Consultations
92 Téléchargements

Altmetric

Partager

  • More