TY - GEN
T1 - Deep Multi-agent Reinforcement Learning in a Homogeneous Open Population
AU - Rădulescu, Roxana
AU - Legrand, Manon
AU - Efthymiadis, Kyriakos
AU - Roijers, Diederik M.
AU - Nowé, Ann
PY - 2019
Y1 - 2019
N2 - Advances in reinforcement learning research have recently produced agents that are competent, or sometimes exceed human performance, in complex tasks. Most interesting real world problems however, are not restricted to one agent, but instead deal with multiple agents acting in the same environment and have proven to be challenging tasks to solve. In this work we present a study on a homogeneous open population of agents modelled as a multi-agent reinforcement learning (MARL) system. We propose a centralised learning approach, with decentralised execution in which agents are given the same policy to execute individually. Using the SimuLane highway traffic simulator as a test-bed we show experimentally that using a single-agent learnt policy to initialise the multi-agent scenario, which we then fine-tune to the task, out-performs agents that learn in the multi-agent setting from scratch. Specifically we contribute an open population MARL configuration, how to transfer knowledge from single- to a multi-agent setting and a training procedure for a homogeneous open population of agents.
AB - Advances in reinforcement learning research have recently produced agents that are competent, or sometimes exceed human performance, in complex tasks. Most interesting real world problems however, are not restricted to one agent, but instead deal with multiple agents acting in the same environment and have proven to be challenging tasks to solve. In this work we present a study on a homogeneous open population of agents modelled as a multi-agent reinforcement learning (MARL) system. We propose a centralised learning approach, with decentralised execution in which agents are given the same policy to execute individually. Using the SimuLane highway traffic simulator as a test-bed we show experimentally that using a single-agent learnt policy to initialise the multi-agent scenario, which we then fine-tune to the task, out-performs agents that learn in the multi-agent setting from scratch. Specifically we contribute an open population MARL configuration, how to transfer knowledge from single- to a multi-agent setting and a training procedure for a homogeneous open population of agents.
KW - Highway traffic
KW - Multi-agent systems
KW - Open population
KW - Reinforcement learning
UR - http://www.scopus.com/inward/record.url?scp=85076104088&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85076104088&partnerID=8YFLogxK
U2 - 10.1007/978-3-030-31978-6_8
DO - 10.1007/978-3-030-31978-6_8
M3 - Conference contribution
AN - SCOPUS:85072663439
SN - 9783030319779
T3 - Communications in Computer and Information Science
SP - 90
EP - 105
BT - ARTIFICIAL INTELLIGENCE
A2 - Atzmueller, Martin
A2 - Duivesteijn, Wouter
PB - Springer
T2 - 30th Benelux Conference on Artificial Intelligence, BNAIC 2018
Y2 - 8 November 2018 through 9 November 2018
ER -