# 想象中的约束多智能体强化学习

> 原文链接：https://robotics.hellobike.com/papers/2026-10-15-2610-21319
> 频道：papers | 更新日期：2026-10-15 | 标签：arxiv, cs.RO
> 核心摘要：针对约束多智能体强化学习的交互效率问题，CA-MAMBA用多智能体世界模型预测成本并在想象中更新，在SMAC和MAMuJoCo验证。

# 想象中的约束多智能体强化学习

::: v-pre
> 原文链接：[Constrained Multi-Agent Reinforcement Learning in Imagination](https://arxiv.org/abs/2610.21319)（arXiv 2610.21319 · 发布日期：2026-10-15）

**Constrained Multi-Agent Reinforcement Learning in Imagination**

针对约束多智能体强化学习的交互效率问题，CA-MAMBA用多智能体世界模型预测成本并在想象中更新，在SMAC和MAMuJoCo验证。

## 核心要点

- 多智能体约束学习中，强模型无关基线仍需大量交互，凸显交互效率问题。
- CA-MAMBA预测多智能体世界模型中的成本，将多数受约束策略更新转移到想象中。
- 在SMAC中位配置上，强模型无关基线需要CA-MAMBA的10到20倍交互。
- 等量预算下，CA-MAMBA在MAMuJoCo获得基线2.8到12倍的奖励。

**验证范围与局限**：验证范围包括SMAC中位配置、MAMuJoCo等量预算比较，以及每个基准一个任务的受控测试。

## 英文摘要 (Abstract)

Predicting cost in a multi-agent world model moves most constrained policy updates into imagination: the stronger model-free baseline needs $10$ to $20$ times CA-MAMBA's interactions on the median SMAC configuration, and CA-MAMBA earns $2.8$ to $12$ times the baselines' MAMuJoCo reward at equal budget. In controlled tests on one task per benchmark, pricing the model's cost advantages at parity more than halves a cost-blind learner's cost and meets a tight continuous-control limit the learned multiplier misses, suggesting the model's cost estimates are not the bottleneck.

作者：Raz Shmueli, Shahaf Shperberg
:::

## 原始参考来源 (Sources)

- [arXiv 2610.21319：Constrained Multi-Agent Reinforcement Learning in Imagination](https://arxiv.org/abs/2610.21319)