Deep Multi-Agent Reinforcement Learning for Decentralized Continuous Cooperative Control
By Christian Schröder de Witt, Bei Peng, Pierre-Alexandre Kamienny, Philip H. S. Torr, Wendelin Böhmer, Shimon Whiteson. 2020
100 citations recorded by the source corpus.
Subfields: Reinforcement Learning, Robotics & Control
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Reinforcement Learning
- Shimon Whiteson
- Chongjie Zhang
- Tonghan Wang
- Yaodong Yang
- Christian Schröder de Witt
- Philip H. S. Torr
- Heng Dong
- Jun Wang
In Robotics & Control
- Philip H. S. Torr
- Jonathan P. How
- Shimon Whiteson
- Matthew Riemer
- Miao Liu
- Christian Schröder de Witt
- J. Wang
- Viktor Makoviychuk
Citations in the corpus
- Is Independent Learning All You Need in the StarCraft Multi-Agent Challenge? (2020, methodology): “A particularly successful line of research uses value factorisation of the joint Q -value function in order to reduce the size of the joint action space [VDN, QMIX, FacMADDPG][25, 19, 7].”
- Off-Policy Multi-Agent Decomposed Policy Gradients (2020, background): “A concurrent work (de Witt et al., 2020) finds that a decomposed centralized critic in QMIX style can improve the performance of MADDPG for learning in continuous action spaces.”
- Trust Region Policy Optimisation in Multi-Agent Reinforcement Learning (2021, methodology): “Figure 3 demonstrate that, in all scenarios, HATRPO and HAPPO enjoy superior performance over those of parameter-sharing methods: IPPO and MAPPO, and also outperform non-parameter sharing MADDPG (Lowe et al., 2017b) both in terms of reward values and variance.”
- Multi-Agent Reinforcement Learning is a Sequence Modeling Problem (2022, methodology): “To evaluate if MAT meets our expectations, we test MAT on the StarCraftII Multi-Agent Challenge (SMAC) benchmark [31] where MAPPO with parameter sharing [46] has shown superior performance, and the Multi-Agent MuJoCo benchmark [7] where HAPPO [15] shows the current state-of-the-art performance.”
- A Policy Gradient Algorithm for Learning to Learn in Multiagent Reinforcement Learning (2020, methodology): “To answer this question, we experiment with a fully cooperative setting of the the 2-Agent HalfCheetah domain (de Witt et al., 2020), where the first and second agent control three joints of the back and front leg with continuous action spaces, respectively (see Figure 4c).”