Abstract Repairing a set of components as a batch is often cheaper than repairing each of them separately. A primary reason for this is that initiating a repair action and testing the system after performing a repair action often incurs non-negligible overhead. However, most troubleshooting algorithms proposed to date do not consider the option of performing batch repair actions. In this work we close this gap, and address the combinatorial problem of choosing which batch repair action to perform so as to minimize the overall repair costs. We call this problem the Batch Repair Problem (BRP) and formalize it. Then, we propose several approaches for solving it. The first seeks to choose to repair the set of components that are most likely to be faulty. The second estimates the cost wasted by repairing a given set of components, and tried to find the set of components that minimizes these costs. The third approach models BRP as a Stochastic Shortest Path Problem (SSP-MDP) [1] , and solves the resulting problem with a dedicated solver. Experimentally, we compare the pros and cons of the proposed BRP algorithms on a standard Boolean circuit benchmark and a novel benchmark from the Physiotherapy domain. Results show the clear benefit of performing batch repair actions with our BRP algorithms compared to repairing components one at a time.

中文翻译：

---

用于自动故障排除的批量修复操作

摘要 批量修复一组组件通常比单独修复每个组件便宜。造成这种情况的一个主要原因是，在执行修复操作后启动修复操作并测试系统通常会产生不可忽略的开销。然而，迄今为止提出的大多数故障排除算法都没有考虑执行批量修复操作的选项。在这项工作中，我们缩小了这一差距，并解决了选择执行哪个批次修复操作以最小化整体修复成本的组合问题。我们称这个问题为批量修复问题 (BRP) 并将其形式化。然后，我们提出了几种解决它的方法。第一个寻求选择修复最有可能出现故障的组件集。第二个估计修理一组给定的组件所浪费的成本，并试图找到使这些成本最小化的组件集。第三种方法将 BRP 建模为随机最短路径问题 (SSP-MDP) [1]，并使用专用求解器解决由此产生的问题。通过实验，我们在标准布尔电路基准和物理治疗领域的新基准上比较了所提出的 BRP 算法的优缺点。结果表明，与一次修复一个组件相比，使用我们的 BRP 算法执行批量修复操作具有明显的优势。我们在标准布尔电路基准和物理治疗领域的新基准上比较了所提出的 BRP 算法的优缺点。结果表明，与一次修复一个组件相比，使用我们的 BRP 算法执行批量修复操作具有明显的优势。我们在标准布尔电路基准和物理治疗领域的新基准上比较了所提出的 BRP 算法的优缺点。结果表明，与一次修复一个组件相比，使用我们的 BRP 算法执行批量修复操作具有明显的优势。