pith. sign in

arxiv: 1910.01465 · v2 · pith:KH6SMM4Xnew · submitted 2019-10-03 · 💻 cs.LG · cs.AI· cs.MA· stat.ML

Reducing Overestimation Bias in Multi-Agent Domains Using Double Centralized Critics

classification 💻 cs.LG cs.AIcs.MAstat.ML
keywords methodsmulti-agenttasksbiasapproachcentralizedcriticscurrent
0
0 comments X
read the original abstract

Many real world tasks require multiple agents to work together. Multi-agent reinforcement learning (RL) methods have been proposed in recent years to solve these tasks, but current methods often fail to efficiently learn policies. We thus investigate the presence of a common weakness in single-agent RL, namely value function overestimation bias, in the multi-agent setting. Based on our findings, we propose an approach that reduces this bias by using double centralized critics. We evaluate it on six mixed cooperative-competitive tasks, showing a significant advantage over current methods. Finally, we investigate the application of multi-agent methods to high-dimensional robotic tasks and show that our approach can be used to learn decentralized policies in this domain.

This paper has not been read by Pith yet.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 4 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Individual Control Barrier Functions-Guided Diffusion Model for Safe Offline Multi-Agent Reinforcement Learning

    cs.LG 2026-06 unverdicted novelty 7.0

    Proposes a diffusion model guided by individual control barrier functions for safe offline multi-agent RL, recovering policies through inverse dynamics and showing safety gains on benchmarks.

  2. PC3D: Zero-Shot Cooperation Across Variable Rosters via Personalized Context Distillation

    cs.LG 2026-05 unverdicted novelty 6.0

    PC3D trains decentralized policies to recover and use personalized coordination context from local histories, enabling higher returns than baselines on variable-roster cooperative MARL tasks with both seen and unseen ...

  3. TRIDENT: Breaking the Hybrid-Safety-Physics Coupling for Provably Safe Multi-Agent Reinforcement Learning

    cs.LG 2026-06 unverdicted novelty 5.0

    TRIDENT is a MARL framework using Richardson-Romberg gradient correction, Lyapunov-constrained trust-region updates, and a physics-informed residual critic that claims O(1/sqrt(K)) convergence to constrained Nash equi...

  4. ERPPO: Entropy Regularization-based Proximal Policy Optimization

    cs.LG 2026-05 unverdicted novelty 5.0

    ERPPO adds a DSA-based ambiguity estimator to MAPPO and switches between L1 and L2 entropy regularization to improve exploration and stability in non-stationary multi-dimensional observations.