# 多智能体RL安全滤波器蒸馏：Crazyflie五机零碰撞且20机零样本迁移

> 原文链接：https://robotics.hellobike.com/papers/2026-11-17-2611-19190-13700118
> 频道：papers | 更新日期：2026-11-17 | 标签：auto, arxiv
> 核心摘要：提出将CBF安全滤波器蒸馏至RL策略的方法，解决多机约束二次增长难题。1800次测试零碰撞，3机训练策略零样本迁移至20机团队，并在5架Crazyflie实机验证，实现无滤波器下的局部观测安全导航。

# 多智能体RL安全滤波器蒸馏：Crazyflie五机零碰撞且20机零样本迁移

::: v-pre
> 本文由 AI 根据公开内容整理，可能有误，请以原文为准：[Distilling Safety Filters for Filter-free Multi-agent RL](https://arxiv.org/abs/2611.19190)（arXiv 2611.19190）

## 核心内容

针对多机器人强化学习中控制障碍函数（CBF）安全滤波器导致的中心化优化与约束二次增长瓶颈，研究提出一种将滤波器安全性内化至RL策略的蒸馏方法。该方法通过隐式Q学习仅从执行的安全动作中学习，并在合成的近碰撞状态进行蒸馏，从理论上证明了蒸馏损失能界定策略的约束违反程度及碰撞概率。实验显示，在无滤波器且仅依赖局部观测的条件下，策略在1800个多智能体测试回合中实现零碰撞。

系统架构摒弃了运行时集中式安全滤波器，使智能体完全基于局部观测决策，保持单机计算量恒定。经3个智能体训练的策略展现出极强的规模泛化能力，可零样本迁移至20个智能体团队，且在8个智能体规模下仍保持无碰撞。实物验证环节采用Crazyflie四旋翼无人机平台，成功实现最多5架无人机同时飞行时的安全导航，证实了该算法在真实物理系统中的低算力部署可行性与鲁棒性。

## 关键看点

- **安全蒸馏机制**：利用隐式Q学习与合成近碰撞状态，将CBF安全性内化至策略，理论证明蒸馏损失可界定单步碰撞概率上界
- **零样本规模泛化**：3机训练策略以恒定单机算力零样本迁移至20机团队，8机以内实测无碰撞，突破多机约束二次增长限制
- **实机验证指标**：在Crazyflie四旋翼平台完成5机同飞测试，1800次仿真测试回合均实现零碰撞，验证纯局部观测下的安全性
- **去中心化部署**：推理阶段完全移除安全滤波器，智能体仅凭局部观测行动，消除中心化优化瓶颈，适配端侧实时控制需求

## 工程与产业观察

该工作为多机协同提供了“训练期强安全约束、部署期纯端到端策略”的工程范式。通过将CBF数学保证蒸馏进神经网络，既规避了在线优化的算力墙，又解决了RL探索期的安全风险。对于集群无人机或移动机器人编队，这种恒定单机算力的零样本扩展能力，是降低大规模部署边际成本的关键技术路径。

## 英文摘要 (Abstract)

Reinforcement learning (RL) can teach robot teams to navigate shared spaces, but trial and error causes collisions that real robots cannot afford. Control barrier function (CBF) safety filters prevent these collisions by correcting unsafe actions, yet in a team the filter becomes a centralized optimization whose number of constraints grows quadratically with the team size. We propose an approach that builds the filter's safety into the RL policy and examine how such a system behaves during filter-free deployment. However, the filter that keeps training safe also hides what the policy must learn: it masks the cost of unsafe actions, which actor-critic methods then overvalue, and it keeps near-collision states out of the training data. Our approach therefore learns only from executed safe actions with implicit Q-learning and distills the filter into the policy at synthetic near-collision states. We prove that filter distillation loss bounds the policy's constraint violation and hence how far agents can cross their safety margin, so errors never accumulate. When distillation states cover deployed states, the loss also bounds the collision probability at every step. Without a filter and acting only on local observations, our policies have no collisions in 1,800 test episodes with multiple agents. A policy trained with three agents transfers zero-shot to teams of twenty at constant per-agent computation and stays collision-free with up to eight agents. We test our policies on Crazyflie quadrotors with up to five flying at the same time.

作者：Srikar Babu Gadipudi, Somayeh Sojoudi, Claire J. Tomlin
:::

## 原始参考来源 (Sources)

- [arXiv 2611.19190：Distilling Safety Filters for Filter-free Multi-agent RL](https://arxiv.org/abs/2611.19190)