Neural networks (NNs) playing the role of controllers have demonstrated impressive empirical performances on challenging control problems. However, the potential adoption of NN controllers in real-life applications also gives rise to a growing concern over the safety of these neural-network controlled systems (NNCSs), especially when used in safety-critical applications. In this work, we present POLAR-Express, an efficient and precise formal reachability analysis tool for verifying the safety of NNCSs. POLAR-Express uses Taylor model arithmetic to propagate Taylor models (TMs) across a neural network layer-by-layer to compute an overapproximation of the neural-network function. It can be applied to analyze any feed-forward neural network with continuous activation functions. We also present a novel approach to propagate TMs more efficiently and precisely across ReLU activation functions. In addition, POLAR-Express provides parallel computation support for the layer-by-layer propagation of TMs, thus significantly improving the efficiency and scalability over its earlier prototype POLAR. Across the comparison with six other state-of-the-art tools on a diverse set of benchmarks, POLAR-Express achieves the best verification efficiency and tightness in the reachable set analysis.
翻译:担任控制器角色的神经网络在挑战性控制问题上展现出了令人印象深刻的性能。然而,神经网络控制系统(NNCS)在实际应用中的潜在采用也引起了人们对这些系统安全性的日益关注,尤其是在安全关键应用中的使用。在本文中,我们介绍了 POLAR-Express,一种用于验证 NNCS 安全性的高效精确的形式可达性分析工具。POLAR-Express 使用 Taylor 模型算法逐层传播 Taylor 模型(TMs),以计算神经网络函数的一种近似值。它可应用于分析任何具有连续激活函数的前馈神经网络。我们还提出了一种新的方法,在 ReLU 激活函数中更有效地和更精确地传播 TMs。此外,POLAR-Express 提供了针对逐层传播 TMs 的并行计算支持,极大地提高了其效率和可伸缩性,超过了其早期原型 POLAR。在与其他六种最先进的工具在多个基准测试上的比较中,POLAR-Express 在可达性分析中取得了最佳的验证效率和紧确性。