Abstract
In this case study, we have explored the use of a neural network model checker to analyze the safety characteristics of a neural network trained using reinforcement learning to compute collision avoidance flight plans for aircraft. We analyzed specific aircraft encounter geometries (e.g., head-on, overtake) and also examined robustness of the neural network. We verified the minimum horizontal separation property by identifying conditions where the neural network can potentially cause a transition from a safe state to an unsafe state. We show how the property verification problem is mathematically transformed and encoded as linear-constraints that can be analyzed by the Marabou model checker.
The authors wish to thank Aleksandar Zeljic for his help using Marabou. This work was funded by DARPA contract FA8750-18-C-0099. The views, opinions and/or findings expressed are those of the author and should not be interpreted as representing the official views or policies of the Department of Defense or the U.S. Government. Approved for Public Release, Distribution Unlimited.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Similar content being viewed by others
References
Cofer, D.: Unintended behavior in learning-enabled systems: detecting the unknown unknowns. In: 2021 IEEE/AIAA 40th Digital Avionics Systems Conference (DASC), pp. 1–7 (2021)
Cofer, D., et al.: Flight test of a collision avoidance neural network with run-time assurance. In: 2022 IEEE/AIAA 41st Digital Avionics Systems Conference (DASC), pp. 1–10 (2022)
Damour, M., et al.: Towards certification of a reduced footprint ACAS-Xu system: a hybrid ml-based solution. In: Habli, I., Sujan, M., Bitsch, F. (eds.) Computer Safety, Reliability, and Security, pp. 34–48. Springer International Publishing, Cham (2021)
Huang, C., Fan, J., Chen, X., Li, W., Zhu, Q.: Polar: a polynomial arithmetic framework for verifying neural-network controlled systems. In: Bouajjani, A., Holík, L., Wu, Z. (eds.) Automated Technology for Verification and Analysis, pp. 414–430. Springer International Publishing, Cham (2022)
Irfan, A., et al.: Towards verification of neural networks for small unmanned aircraft collision avoidance. In: 2020 AIAA/IEEE 39th Digital Avionics Systems Conference (DASC), pp. 1–10. IEEE (2020)
Katz, G., Barrett, C., Dill, D.L., Julian, K., Kochenderfer, M.J.: Reluplex: an efficient SMT solver for verifying deep neural networks. In: Majumdar, R., Kunčak, V. (eds.) Computer Aided Verification, pp. 97–117. Springer International Publishing, Cham (2017)
Katz, G., et al.: The marabou framework for verification and analysis of deep neural networks. In: International Conference on Computer Aided Verification, pp. 443–452 (2019)
Manzanas Lopez, D., Johnson, T.T., Bak, S., Tran, H.D., Hobbs, K.L.: Evaluation of neural network verification methods for air-to-air collision avoidance. J. Air Transp. 31(1), 1–17 (2023)
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2023 The Author(s), under exclusive license to Springer Nature Switzerland AG
About this paper
Cite this paper
Liu, C., Cofer, D., Osipychev, D. (2023). Verifying an Aircraft Collision Avoidance Neural Network with Marabou. In: Rozier, K.Y., Chaudhuri, S. (eds) NASA Formal Methods. NFM 2023. Lecture Notes in Computer Science, vol 13903. Springer, Cham. https://doi.org/10.1007/978-3-031-33170-1_5
Download citation
DOI: https://doi.org/10.1007/978-3-031-33170-1_5
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-031-33169-5
Online ISBN: 978-3-031-33170-1
eBook Packages: Computer ScienceComputer Science (R0)