We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

math.NA

Change to browse by:

References & Citations

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Mathematics > Numerical Analysis

Title: Neural Network Compression of ACAS Xu Early Prototype is Unsafe: Closed-Loop Verification through Quantized State Backreachability

Abstract: ACAS Xu is an air-to-air collision avoidance system designed for unmanned aircraft that issues horizontal turn advisories to avoid an intruder aircraft. Due the use of a large lookup table in the design, a neural network compression of the policy was proposed. Analysis of this system has spurred a significant body of research in the formal methods community on neural network verification. While many powerful methods have been developed, most work focuses on open-loop properties of the networks, rather than the main point of the system -- collision avoidance -- which requires closed-loop analysis.
In this work, we develop a technique to verify a closed-loop approximation of the system using state quantization and backreachability. We use favorable assumptions for the analysis -- perfect sensor information, instant following of advisories, ideal aircraft maneuvers and an intruder that only flies straight. When the method fails to prove the system is safe, we refine the quantization parameters until generating counterexamples where the original (non-quantized) system also has collisions.
Subjects: Numerical Analysis (math.NA); Artificial Intelligence (cs.AI); Logic in Computer Science (cs.LO)
Cite as: arXiv:2201.06626 [math.NA]
  (or arXiv:2201.06626v3 [math.NA] for this version)

Submission history

From: Stanley Bak [view email]
[v1] Mon, 17 Jan 2022 20:49:29 GMT (1085kb,D)
[v2] Wed, 26 Jan 2022 17:56:31 GMT (1085kb,D)
[v3] Sun, 27 Mar 2022 15:39:32 GMT (1088kb,D)

Link back to: arXiv, form interface, contact.