Scholarlink

Providing safety guarantees for autonomous systems is difficult as these systems operate in complex environments that require the use of learning-enabled components, such as deep neural networks (DNNs), for visual perception. DNNs are hard to formally verify due to their size (they can have billions of parameters), lack of formal specifications, and sensitivity to slight changes in the surrounding environment. Furthermore, the high-dimensional inputs to the DNNs come from sensors such as high-fidelity cameras that are themselves complex and hard to model – they bear complex relationships to the system states and are subject to random environmental perturbations. We present a survey of verification techniques that aim to provide quantitative or qualitative formal guarantees for such autonomous systems.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Formal Verification Techniques for Vision-Based Autonomous Systems – A Survey

  • 章节

  • 发表日期：

    2025年

查看全文
  • Sayan Mitra,
  • Corina Păsăreanu,
  • Pavithra Prabhakar,
  • Sanjit A. Seshia,
  • Ravi Mangal,
  • Yangge Li,
  • Christopher Watson,
  • Divya Gopinath,
  • Huafeng Yu

摘要

Providing safety guarantees for autonomous systems is difficult as these systems operate in complex environments that require the use of learning-enabled components, such as deep neural networks (DNNs), for visual perception. DNNs are hard to formally verify due to their size (they can have billions of parameters), lack of formal specifications, and sensitivity to slight changes in the surrounding environment. Furthermore, the high-dimensional inputs to the DNNs come from sensors such as high-fidelity cameras that are themselves complex and hard to model – they bear complex relationships to the system states and are subject to random environmental perturbations. We present a survey of verification techniques that aim to provide quantitative or qualitative formal guarantees for such autonomous systems.