Assuring Learning-Enabled Components in Small Unmanned Aircraft Systems

Aviation has a remarkable safety record ensured by strict processes, rules, certifications, and regulations, in which formal methods have played a role in large companies developing commercial aerospace vehicles and related cyber-physical systems (CPS). This has not been the case for small Unmanned Aircraft Systems (UAS) that are still largely unregulated, uncertified, and not fully integrated into the national airspace. However, emerging UAS missions interact closely with the environment and utilize learning-enabled components (LECs), such as neural networks (NNs) for many tasks. Applying formal methods in this context will enable improved safety and ease the immersion of UASs into the national airspace. We develop UAS that interact closely with the environment, interact with human users, and require precise plans, navigation, and controllers. They also generally leverage LECs for perception and data collection. However, the impact of ML-based LECs on UAS performance is still an area of research. We have developed an advanced simulator incorporatingML-based perception in highly dynamic situations requiring advanced control strategies to study the impacts of ML-based perception on holistic UAS performance. In other work, we have developed a WebGME-based software framework called the Assurance-based Learning-enabled CPS (ALC) toolchain for designing CPS that incorporate LECs, including the Neural Network Verification (NNV) formal verification tool. In this paper, we present two key developments: 1) a quantification of the impact of ML-based perception on holistic (physical and cyber) UAS performance, and 2) a discussion of challenges in applying these methods in this environment to guarantee UAS performance under various Neural Net (NN) strategies, executed at various computational rates, and with vehicles moving at various speeds. We demonstrate that vehicle dynamics, rate of perception execution, the design of the controller, and the design of the NN all contributed to total vehicle performance.

[1]  Eric Feron.  Formal methods for aerospace applications , 2012, 2012 Formal Methods in Computer-Aided Design (FMCAD).

[2]  Henson Graves,et al.  Using formal methods with SysML in aerospace design and engineering , 2011, Annals of Mathematics and Artificial Intelligence.

[3]  Charles Hartsell,et al.  Model-based design for CPS with learning-enabled components , 2019, DESTION@CPSIoTWeek.

[4]  Weiming Xiang,et al.  Verification of Deep Convolutional Neural Networks Using ImageStars , 2020, CAV.

[5]  Sebastian G. Elbaum,et al.  Towards Aerial Recovery of Parachute-Deployed Payloads , 2018, 2018 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS).

[6]  Ashish Kapoor,et al.  AirSim: High-Fidelity Visual and Physical Simulation for Autonomous Vehicles , 2017, FSR.

[7]  Justin M. Bradley,et al.  Predicting Digging Success for Unmanned Aircraft System Sensor Emplacement , 2018, ISER.

[8]  Lui Sha,et al.  Real-Time Reachability for Verified Simplex Design , 2014, RTSS.

[9]  Lui Sha,et al.  Real-Time Reachability for Verified Simplex Design , 2014, 2014 IEEE Real-Time Systems Symposium.

[10]  ARCH-COMP20 Category Report: Artificial Intelligence and Neural Network Control Systems (AINNCS) for Continuous and Hybrid Systems Plants , 2020, ARCH.

[11]  Frédéric Boniol,et al.  Toward a wider use of formal methods for aerospace systems design and verification , 2009, International Journal on Software Tools for Technology Transfer.

[12]  Charles Hartsell,et al.  CPS Design with Learning-Enabled Components: A Case Study , 2019, RSP.

[13]  James Nightingale,et al.  A UAV-Cloud System for Disaster Sensing Applications , 2015, 2015 IEEE 81st Vehicular Technology Conference (VTC Spring).

[14]  Weiming Xiang,et al.  Star-Based Reachability Analysis of Deep Neural Networks , 2019, FM.

[15]  Weiming Xiang,et al.  NNV: The Neural Network Verification Tool for Deep Neural Networks and Learning-Enabled Cyber-Physical Systems , 2020, CAV.

[16]  Weiming Xiang,et al.  Parallelizable Reachability Analysis Algorithms for Feed-Forward Neural Networks , 2019, 2019 IEEE/ACM 7th International Conference on Formal Methods in Software Engineering (FormaliSE).

[17]  Taylor T. Johnson,et al.  Improved Geometric Path Enumeration for Verifying ReLU Neural Networks , 2020, CAV.

[18]  Darren D. Cofer.  Formal Methods in the Aerospace Industry: Follow the Money , 2012, ICFEM.

[19]  Taylor T. Johnson,et al.  Verification of Closed-loop Systems with Neural Network Controllers , 2019, ARCH@CPSIoTWeek.

[20]  Kaiming He,et al.  Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks , 2015, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[21]  Taylor T. Johnson,et al.  ARCH-COMP19 Category Report: Artificial Intelligence and Neural Network Control Systems (AINNCS) for Continuous and Hybrid Systems Plants , 2019, ARCH@CPSIoTWeek.

[22]  Xenofon D. Koutsoukos,et al.  Safety Verification of Cyber-Physical Systems with Reinforcement Learning Control , 2019, ACM Trans. Embed. Comput. Syst..