A group of collaborative embedded systems does not depend on a central authority to operate in uncertain environments. By engaging in various negotiation protocols, the participants assign roles, schedule tasks, and combine their world views for more resilient perception and planning. To verify functional correctness, critical components can be tested with simulation-based methods, but the possibility of runtime faults still necessitates online monitoring. In this work, we characterize and address the runtime verification problem in the context of collaborative embedded systems. We present a case study based on industrial transport robots and model the main operating procedure, a distributed bidding protocol. The key properties that must hold for functional correctness turn out to comprise multiple semantic concepts that cannot be jointly expressed with any single formalism. To address this issue, we identify three specification languages that are particularly suitable for monitoring of collaborative embedded systems: Certifying distributed algorithms, trace expressions, and real-valued temporal logic. We show how each of them can be used to capture a subset of the relevant properties and outline a way of integrating them into a common framework.

一组协作嵌入式系统不依赖于中央机构在不确定的环境中运行。通过参与各种谈判协议，参与者分配角色、安排任务并结合他们的世界观，以实现更具弹性的感知和规划。为了验证功能的正确性，可以使用基于仿真的方法来测试关键组件，但运行时故障的可能性仍然需要在线监控。在这项工作中，我们描述并解决了协作嵌入式系统背景下的运行时验证问题。我们提出了一个基于工业运输机器人的案例研究，并对主要操作程序（分布式投标协议）进行了建模。功能正确性必须保持的关键属性包含多个语义概念，这些概念不能用任何单一形式主义联合表达。为了解决这个问题，我们确定了三种特别适合监控协作嵌入式系统的规范语言：认证分布式算法、跟踪表达式和实值时序逻辑。我们展示了如何使用它们中的每一个来捕获相关属性的子集，并概述了将它们集成到通用框架中的方法。