pith. sign in

arxiv: 2605.08190 · v1 · submitted 2026-05-05 · 💻 cs.LG · cs.SY· eess.SY

Synergistic Simplex: Cooperative Runtime Assurance for Safety-Critical Autonomous Systems

Ayoosh Bansal , Mikael Yeghiazaryan , Artyom Khachatryan , Tianyi Zhu , Hunmin Kim , Naira Hovakimyan , Lui Sha This is my paper

Pith reviewed 2026-05-12 01:18 UTC · model grok-4.3

classification 💻 cs.LG cs.SYeess.SY
keywords runtime assurancesynergistic simplexautonomous vehiclesmachine learning safetyobstacle detectionformal safety guaranteesperception simplex
0
0 comments X

The pith

The Synergistic Simplex architecture allows safety monitors to safely use machine learning outputs, raising overall performance while retaining formal safety guarantees.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

Autonomous vehicles depend on machine learning for perception tasks like obstacle detection, yet these components produce occasional long-tail errors that prevent their use in safety-critical roles. Traditional runtime assurance architectures keep machine learning and safety monitors separate to maintain verifiable guarantees, which limits overall capability. This paper introduces bidirectional integration so that monitors can draw on machine learning outputs under explicitly derived conditions. The resulting design is analyzed and tested on an autonomous-vehicle obstacle-detection task, showing measurable gains without safety violations.

Core claim

The central claim is that the Synergistic Simplex architecture improves system performance by enabling bidirectional integration between ML components and safety monitors while preserving formal safety guarantees. The key innovation is allowing safety monitors to use ML outputs, which is typically prohibited in RTA systems. The authors formally derive conditions under which this integration preserves safety and demonstrate the performance benefits on AV obstacle detection.

What carries the argument

The Synergistic Simplex architecture that permits safety monitors to incorporate machine-learning outputs under formally derived safety-preserving conditions.

If this is right

  • Overall system performance rises for safety-critical tasks such as obstacle detection because monitors can now exploit machine-learning strengths.
  • Formal safety guarantees remain intact provided the derived integration conditions are satisfied.
  • The architecture applies directly to autonomous-vehicle perception pipelines without requiring changes to the underlying machine-learning models.
  • Verification effort focuses on the new integration conditions rather than on the machine-learning component itself.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same bidirectional pattern could be tested on other perception or control tasks where traditional monitors are currently too conservative.
  • If the conditions prove robust, certified systems could adopt more capable machine-learning models without lengthening verification timelines.
  • Longer deployments might reveal whether the derived conditions need tightening when sensor noise or environmental shifts exceed the modeled cases.

Load-bearing premise

The formally derived conditions under which safety monitors can safely use ML outputs will continue to hold for the long-tail faults that occur in real-world autonomous-vehicle obstacle detection.

What would settle it

An experiment in which a safety monitor that receives an ML output fails to prevent a collision or violation that the identical monitor would have caught when denied the ML output.

Figures

Figures reproduced from arXiv: 2605.08190 by Artyom Khachatryan, Ayoosh Bansal, Hunmin Kim, Lui Sha, Mikael Yeghiazaryan, Naira Hovakimyan, Tianyi Zhu.

Figure 1
Figure 1. Figure 1: Overview of the Synergistic Simplex (SS) architecture, extended from Perception Simplex [25]. As in traditional runtime assurance designs [16]–[19], the ML Layer executes system’s mission, and the Safety Layer enforces deterministic guardrails. Synergistic Simplex improves upon prior designs by leveraging bidirectional communication between the layers. A solution to this is the runtime assurance (RTA) desi… view at source ↗
Figure 2
Figure 2. Figure 2: Detectability region for LiDAR-based obstacle detec [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Simplified dependency graph of the autonomy stack, [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Instantiation of SS for AVs. The safety layer per￾forms verifiable LiDAR-based obstacle detection using the detectability model of [24] and a verified override policy [25]. Lane-detection outputs flow through the M2S link to refine obstacle criticality without weakening safety guarantees. 4) Downstream Functions: Downstream functions lie strictly after the Simplexed ML layer node DM in the de￾pendency grap… view at source ↗
Figure 5
Figure 5. Figure 5: Zoning policy guiding how the safety layer of [PITH_FULL_IMAGE:figures/full_fig_p008_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: ML fault regions and their relation to the safety [PITH_FULL_IMAGE:figures/full_fig_p008_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Visualization of the simulation scenarios used to [PITH_FULL_IMAGE:figures/full_fig_p009_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Example run from Exp. 2. The ML agent completes [PITH_FULL_IMAGE:figures/full_fig_p010_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Representative snapshots from CARLA (primary eval [PITH_FULL_IMAGE:figures/full_fig_p010_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Analysis of safety layer interventions in Exp. 2. The [PITH_FULL_IMAGE:figures/full_fig_p011_10.png] view at source ↗
read the original abstract

Autonomous systems increasingly rely on machine-learning (ML) components for safety-critical tasks such as perception and control in autonomous vehicles (AVs). While ML enables essential capabilities, it inevitably exhibits long-tail faults that make it unsuitable for safety-critical tasks. Runtime assurance (RTA) mitigates this issue by pairing ML components with verifiable safety monitors, e.g., Control Simplex and Perception Simplex architectures. However, the limited performance of safety monitors remains a major bottleneck. The Synergistic Simplex (SS) architecture improves system performance by enabling bidirectional integration between ML components and safety monitors while preserving formal safety guarantees. The key innovation here is allowing safety monitors to use ML outputs, which is typically prohibited in RTA systems. We formally derive conditions under which this integration preserves safety and demonstrate the performance benefits. We present the design, analysis, and evaluation of SS for AV obstacle detection.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The paper proposes the Synergistic Simplex (SS) architecture as an extension of runtime assurance (RTA) frameworks such as Control Simplex and Perception Simplex. It enables bidirectional integration between ML-based perception/control components and safety monitors for autonomous vehicle obstacle detection, formally derives conditions under which monitors may safely incorporate ML outputs (normally prohibited in RTA), and claims both preservation of safety invariants and measurable performance gains over unidirectional designs.

Significance. If the formal conditions are both sound and robust to the long-tail error distributions of deployed deep networks, the work would meaningfully relax a long-standing restriction in RTA literature, allowing monitors to exploit ML accuracy while retaining verifiable safety. This could reduce the performance penalty that currently limits RTA adoption in perception-heavy AV tasks.

major comments (2)
  1. [Formal derivation section (presumably §4 or §5)] The central claim rests on formally derived conditions that permit safety monitors to use ML outputs without violating invariants. The manuscript must state these conditions explicitly (including any assumptions on ML error models, bounded deviation, or detectable inconsistency) and prove that they remain valid under the long-tail fault distributions typical of real AV perception; without such a proof the bidirectional integration risks silently degrading the monitor.
  2. [Evaluation section] The evaluation for AV obstacle detection must include quantitative comparison against standard unidirectional RTA baselines on metrics that directly test safety preservation (e.g., false-negative rate on rare obstacle classes) rather than only aggregate performance; absent such data the performance-benefit claim cannot be assessed.
minor comments (2)
  1. [Abstract] The abstract asserts 'formal derivation' and 'demonstrate the performance benefits' yet supplies no equations, invariants, or numerical results; the introduction or abstract should at minimum reference the key theorem or table that supports these assertions.
  2. [Architecture description] Notation for the bidirectional interface (e.g., how ML outputs are filtered or validated before entering the monitor) should be introduced early and used consistently.

Simulated Author's Rebuttal

2 responses · 1 unresolved

We thank the referee for the constructive feedback on the formal conditions and evaluation requirements. We address each major comment below and outline the revisions we will undertake.

read point-by-point responses

  1. Referee: [Formal derivation section (presumably §4 or §5)] The central claim rests on formally derived conditions that permit safety monitors to use ML outputs without violating invariants. The manuscript must state these conditions explicitly (including any assumptions on ML error models, bounded deviation, or detectable inconsistency) and prove that they remain valid under the long-tail fault distributions typical of real AV perception; without such a proof the bidirectional integration risks silently degrading the monitor.

    Authors: We agree that the conditions and assumptions must be stated more explicitly. In the revised manuscript we will add a dedicated paragraph in the formal analysis section that enumerates the assumptions on the ML error model, including bounded deviation and detectable inconsistency. The derivation already shows that safety invariants are preserved whenever these assumptions hold and the monitor can verify consistency. A full proof of robustness specifically against arbitrary long-tail distributions of deployed deep networks is not provided in the current work, as it would require modeling network-specific error tails that lie outside the general theoretical framework; we will note this limitation explicitly. revision: partial

  2. Referee: [Evaluation section] The evaluation for AV obstacle detection must include quantitative comparison against standard unidirectional RTA baselines on metrics that directly test safety preservation (e.g., false-negative rate on rare obstacle classes) rather than only aggregate performance; absent such data the performance-benefit claim cannot be assessed.

    Authors: We will revise the evaluation section to include direct quantitative comparisons against unidirectional RTA baselines (Control Simplex and Perception Simplex) on safety-preservation metrics, specifically false-negative rates for rare obstacle classes, in addition to the existing aggregate performance results. These comparisons will be derived from re-analysis of the simulation data already collected for the AV obstacle-detection task. revision: yes

standing simulated objections not resolved
  • A complete proof that the derived conditions remain valid under arbitrary long-tail fault distributions of real deep-network perception errors.

Circularity Check

0 steps flagged

No circularity: formal derivation presented as independent of inputs

full rationale

The provided abstract and description state that conditions for safe bidirectional integration are formally derived while preserving safety guarantees, but contain no equations, parameter fits, self-citations, or ansatzes that reduce the claimed result to its own inputs by construction. No load-bearing steps are visible that match any of the enumerated circularity patterns; the central claim rests on an external formal analysis whose details are not shown to loop back on themselves.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review provides no explicit free parameters, axioms, or invented entities beyond the named architecture itself.

pith-pipeline@v0.9.0 · 5483 in / 994 out tokens · 33060 ms · 2026-05-12T01:18:24.419349+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

56 extracted references · 56 canonical work pages

  1. [1]

    Challenges of machine learning applied to safety-critical cyber-physical systems,

    A. Pereira and C. Thomas, “Challenges of machine learning applied to safety-critical cyber-physical systems,”Machine Learning and Knowl- edge Extraction, vol. 2, no. 4, pp. 579–602, 2020

    work page 2020

  2. [2]

    Assuring safety-critical machine learning enabled sys- tems: Challenges and promise,

    A. E. Goodloe, “Assuring safety-critical machine learning enabled sys- tems: Challenges and promise,” in2022 IEEE International Symposium on Software Reliability Engineering Workshops (ISSREW). IEEE, 2022, pp. 326–332

    work page 2022

  3. [3]

    Challenges in specifying safety- critical systems with ai-components,

    I. Malleswaran and S. Dinakaran, “Challenges in specifying safety- critical systems with ai-components,” 2023. [Online]. Available: https://odr.chalmers.se/items/d130c7f0-55a1-454c-acbc-1caf59a8e0e1

    work page 2023

  4. [4]

    Deep long-tailed learning: A survey,

    Y . Zhang, B. Kang, B. Hooi, S. Yan, and J. Feng, “Deep long-tailed learning: A survey,”IEEE transactions on pattern analysis and machine intelligence, vol. 45, no. 9, pp. 10 795–10 816, 2023

    work page 2023

  5. [5]

    Machine learning robustness: A primer,

    H. B. Braiek and F. Khomh, “Machine learning robustness: A primer,” inTrustworthy AI in Medical Imaging. Elsevier, 2025, pp. 37–71

    work page 2025

  6. [6]

    One pixel attack for fooling deep neural networks,

    J. Su, D. V . Vargas, and K. Sakurai, “One pixel attack for fooling deep neural networks,”IEEE Transactions on Evolutionary Computation, vol. 23, no. 5, pp. 828–841, 2019

    work page 2019

  7. [7]

    Identifying challenges to the certification of machine learning for safety critical systems,

    E. Jenn, A. Albore, F. Mamalet, G. Flandin, C. Gabreau, H. Delseny, A. Gauffriau, H. Bonnin, L. Alecu, J. Pirardet al., “Identifying challenges to the certification of machine learning for safety critical systems,” inEuropean congress on embedded real time systems (ERTS 2020), vol. 1, 2020

    work page 2020

  8. [8]

    Machine learning for reliability engineering and safety applications: Review of current status and future opportunities,

    Z. Xu and J. H. Saleh, “Machine learning for reliability engineering and safety applications: Review of current status and future opportunities,” Reliability Engineering & System Safety, vol. 211, p. 107530, 2021

    work page 2021

  9. [9]

    Taxonomy of machine learning safety: A survey and primer,

    S. Mohseni, H. Wang, C. Xiao, Z. Yu, Z. Wang, and J. Yadawa, “Taxonomy of machine learning safety: A survey and primer,”ACM Computing Surveys, vol. 55, no. 8, pp. 1–38, 2022

    work page 2022

  10. [10]

    The road to safety: A review of uncertainty and applications to autonomous driving perception,

    B. Araujo, J. F. Teixeira, J. Fonseca, R. Cerqueira, and S. C. Beco, “The road to safety: A review of uncertainty and applications to autonomous driving perception,”Entropy, vol. 26, no. 8, p. 634, 2024

    work page 2024

  11. [11]

    Safety assurance of machine learning for autonomous systems,

    C. Paterson, R. Hawkins, C. Picardi, Y . Jia, R. Calinescu, and I. Habli, “Safety assurance of machine learning for autonomous systems,”Reli- ability Engineering & System Safety, p. 111311, 2025

    work page 2025

  12. [12]

    Collision Between a Car Operating With Automated Vehicle Control Systems and a Tractor-Semitrailer Truck Near Williston, Florida, May 7, 2016,

    National Transportation Safety Board, “Collision Between a Car Operating With Automated Vehicle Control Systems and a Tractor-Semitrailer Truck Near Williston, Florida, May 7, 2016,” Accident Report NTSB/HAR-17/02 PB2017-102600, September 2017. [Online]. Available: https://www.ntsb.gov/investigations/accidentreports/ reports/har1702.pdf

    work page 2016

  13. [13]

    Collision Between Vehicle Controlled by Developmental Automated Driving System and Pedestrian Tempe, Arizona, March 18, 2018,

    ——, “Collision Between Vehicle Controlled by Developmental Automated Driving System and Pedestrian Tempe, Arizona, March 18, 2018,” Accident Report NTSB/HAR-19/03 PB2019- 101402, 2019. [Online]. Available: https://www.ntsb.gov/investigations/ AccidentReports/Reports/HAR1903.pdf

    work page 2018

  14. [14]

    Highway accident brief HWY19FH008,

    ——, “Highway accident brief HWY19FH008,” Highway Accident Brief, Accident Number: HWY19FH008, March 2019. [On- line]. Available: https://www.ntsb.gov/investigations/AccidentReports/ Reports/HAB2001.pdf

    work page 2019

  15. [15]

    Standing General Order on Crash Reporting for Automated Driving Systems and Level 2 Advanced Driver Assistance Systems,

    National Highway Traffic Safety Administration (NHTSA), “Standing General Order on Crash Reporting for Automated Driving Systems and Level 2 Advanced Driver Assistance Systems,” https://www.nhtsa.gov/ laws-regulations/standing-general-order-crash-reporting, 2021–Present, regularly updated reports tracking incidents involving Automated Driv- ing Systems

    work page 2021

  16. [16]

    Using simplicity to control complexity,

    L. Sha, “Using simplicity to control complexity,”IEEE Software, vol. 18, no. 4, pp. 20–28, 2001

    work page 2001

  17. [17]

    The simplex reference model: Limiting fault-propagation due to unreliable components in cyber-physical system architectures,

    T. L. Crenshaw, E. L. Gunter, C. L. Robinson, L. Sha, and P. R. Kumar, “The simplex reference model: Limiting fault-propagation due to unreliable components in cyber-physical system architectures,” in Proceedings of the 28th IEEE Real-Time Systems Symposium (RTSS 2007), 3-6 December 2007, Tucson, Arizona, USA, 2007, pp. 400–412. [Online]. Available: https...

    work page doi:10.1109/rtss.2007.34 2007

  18. [18]

    The system-level simplex architecture for improved real-time embedded system safety,

    S. Bak, D. K. Chivukula, O. Adekunle, M. Sun, M. Caccamo, and L. Sha, “The system-level simplex architecture for improved real-time embedded system safety,” in15th IEEE Real-Time and Embedded Technology and Applications Symposium, San Francisco, USA, 2009, pp. 99–107

    work page 2009

  19. [19]

    Run-time assurance: A rising technology,

    J. G. Fuller, “Run-time assurance: A rising technology,” in2020 AIAA/IEEE 39th Digital Avionics Systems Conference (DASC). IEEE, 2020, pp. 1–9

    work page 2020

  20. [20]

    Neural simplex architecture,

    D. T. Phan, R. Grosu, N. Jansen, N. Paoletti, S. A. Smolka, and S. D. Stoller, “Neural simplex architecture,” inNASA Formal Methods Symposium. Springer, 2020, pp. 97–114

    work page 2020

  21. [21]

    A component-based simplex architecture for high-assurance cyber-physical systems,

    D. Phan, J. Yang, M. Clark, R. Grosu, J. Schierman, S. Smolka, and S. Stoller, “A component-based simplex architecture for high-assurance cyber-physical systems,” in2017 17th International Conference on Application of Concurrency to System Design (ACSD). IEEE, 2017, pp. 49–58

    work page 2017

  22. [22]

    The black-box simplex architecture for runtime assurance of autonomous cps,

    U. Mehmood, S. Sheikhi, S. Bak, S. A. Smolka, and S. D. Stoller, “The black-box simplex architecture for runtime assurance of autonomous cps,” inNASA formal methods symposium. Springer, 2022, pp. 231– 250

    work page 2022

  23. [23]

    Soter: a runtime assurance framework for programming safe robotics systems,

    A. Desai, S. Ghosh, S. A. Seshia, N. Shankar, and A. Tiwari, “Soter: a runtime assurance framework for programming safe robotics systems,” in 2019 49th Annual IEEE/IFIP International Conference on Dependable Systems and Networks (DSN). IEEE, 2019, pp. 138–150

    work page 2019

  24. [24]

    Verifiable obstacle detection,

    A. Bansal, H. Kim, S. Yu, B. Li, N. Hovakimyan, M. Caccamo, and L. Sha, “Verifiable obstacle detection,” in2022 IEEE 33rd International Symposium on Software Reliability Engineering (ISSRE). IEEE, 2022, pp. 61–72

    work page 2022

  25. [25]

    Perception simplex: Verifiable collision avoidance in autonomous vehicles amidst obstacle detection faults,

    ——, “Perception simplex: Verifiable collision avoidance in autonomous vehicles amidst obstacle detection faults,”Software Testing, Verification and Reliability, p. e1879, 2024. [Online]. Available: https://doi.org/10. 1002/stvr.1879

    work page 2024

  26. [26]

    Road vehicles – Functional safety,

    ISO/TC 22/SC 32, “Road vehicles – Functional safety,” International Organization for Standardization, Geneva, Switzerland, Standard ISO 26262:2018, 2018

    work page 2018

  27. [27]

    DO-178C: Software Considerations in Airborne Systems and Equipment Certification,

    “DO-178C: Software Considerations in Airborne Systems and Equipment Certification,” RTCA, Inc., Washington, D.C., Standard RTCA/DO-178C, 2011

    work page 2011

  28. [28]

    Fault tolerable redundancy control,

    K. Ishii, A. Noguchi, and Y . Gotoh, “Fault tolerable redundancy control,” Apr. 15 1986, US Patent 4,583,224

    work page 1986

  29. [29]

    Architectural principles for safety-critical real-time applications,

    J. H. Lala and R. E. Harper, “Architectural principles for safety-critical real-time applications,”Proceedings of the IEEE, vol. 82, no. 1, pp. 25–40, 1994

    work page 1994

  30. [30]

    Blanke, M

    M. Blanke, M. Kinnaert, J. Lunze, M. Staroswiecki, and J. Schr ¨oder, Diagnosis and fault-tolerant control. Springer, 2006, vol. 2

    work page 2006

  31. [31]

    Towards a viable autonomous driving research platform,

    J. Wei, J. M. Snider, J. Kim, J. M. Dolan, R. Rajkumar, and B. Litkouhi, “Towards a viable autonomous driving research platform,” in2013 IEEE Intelligent Vehicles Symposium (IV), June 2013, pp. 763–770

    work page 2013

  32. [32]

    Safety!= security: On the resilience of asil-d certified microcontrollers against fault injection attacks,

    N. Wiersma and R. Pareja, “Safety!= security: On the resilience of asil-d certified microcontrollers against fault injection attacks,” in2017 Workshop on Fault Diagnosis and Tolerance in Cryptography (FDTC). IEEE, 2017, pp. 9–16

    work page 2017

  33. [33]

    Ml-based fault injection for autonomous vehicles: A case for bayesian fault injection,

    S. Jha, S. Banerjee, T. Tsai, S. K. Hari, M. B. Sullivan, Z. T. Kalbarczyk, S. W. Keckler, and R. K. Iyer, “Ml-based fault injection for autonomous vehicles: A case for bayesian fault injection,” in2019 49th Annual IEEE/IFIP International Conference on Dependable Systems and Net- works (DSN). IEEE, 2019, pp. 112–124

    work page 2019

  34. [34]

    Rsimplex: A robust control architecture for cyber and physical failures,

    X. Wang, N. Hovakimyan, and L. Sha, “Rsimplex: A robust control architecture for cyber and physical failures,”ACM Transactions on Cyber-Physical Systems, vol. 2, no. 4, pp. 1–26, 2018

    work page 2018

  35. [35]

    Sl1- simplex: Safe velocity regulation of self-driving vehicles in dynamic and unforeseen environments,

    Y . Mao, Y . Gu, N. Hovakimyan, L. Sha, and P. V oulgaris, “Sl1- simplex: Safe velocity regulation of self-driving vehicles in dynamic and unforeseen environments,”ACM Transactions on Cyber-Physical Systems, vol. 7, no. 1, pp. 1–24, 2023

    work page 2023

  36. [36]

    On using real-time reachability for the safety assurance of machine learning controllers,

    P. Musau, N. Hamilton, D. M. Lopez, P. Robinette, and T. T. Johnson, “On using real-time reachability for the safety assurance of machine learning controllers,” in2022 IEEE International Conference on Assured Autonomy (ICAA). IEEE, 2022, pp. 1–10

    work page 2022

  37. [37]

    Fast range image-based segmentation of sparse 3d laser scans for online operation,

    I. Bogoslavskyi and C. Stachniss, “Fast range image-based segmentation of sparse 3d laser scans for online operation,” in2016 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). IEEE, 2016, pp. 163–169

    work page 2016

  38. [38]

    Efficient online segmentation for sparse 3d laser scans,

    ——, “Efficient online segmentation for sparse 3d laser scans,”PFG– Journal of Photogrammetry, Remote Sensing and Geoinformation Sci- ence, vol. 85, no. 1, pp. 41–52, 2017

    work page 2017

  39. [39]

    Priority inheritance protocols: An approach to real-time synchronization,

    L. Sha, R. Rajkumar, and J. Lehoczky, “Priority inheritance protocols: An approach to real-time synchronization,”IEEE Transactions on Com- puters, vol. 39, no. 9, pp. 1175–1185, Sep. 1990

    work page 1990

  40. [40]

    On removing algorithmic priority inversion from mission-critical machine inference pipelines,

    S. Liu, S. Yao, X. Fu, R. Tabish, S. Yu, A. Bansal, H. Yun, L. Sha, and T. Abdelzaher, “On removing algorithmic priority inversion from mission-critical machine inference pipelines,” in2020 IEEE Real-Time Systems Symposium (RTSS). IEEE, 2020, pp. 319–332

    work page 2020

  41. [41]

    Real-time task scheduling for machine perception in intelligent cyber-physical systems,

    S. Liu, S. Yao, X. Fu, H. Shao, R. Tabish, S. Yu, A. Bansal, H. Yun, L. Sha, and T. Abdelzaher, “Real-time task scheduling for machine perception in intelligent cyber-physical systems,”IEEE Transactions on Computers, vol. 71, no. 8, pp. 1770–1783, 2021

    work page 2021

  42. [42]

    On exploring image resizing for optimizing criticality-based machine perception,

    Y . Hu, S. Liu, T. Abdelzaher, M. Wigness, and P. David, “On exploring image resizing for optimizing criticality-based machine perception,” in 2021 IEEE 27th International Conference on Embedded and Real-Time Computing Systems and Applications (RTCSA). IEEE, 2021, pp. 169– 178

    work page 2021

  43. [43]

    Real-time task scheduling with image resizing for criticality- based machine perception,

    ——, “Real-time task scheduling with image resizing for criticality- based machine perception,”Real-Time Systems, pp. 1–26, 2022

    work page 2022

  44. [44]

    Lidar cluster first and camera inference later: A new perspective towards autonomous driving,

    J. Chen, S. Yu, R. Tabish, A. Bansal, S. Liu, T. Abdelzaher, and L. Sha, “Lidar cluster first and camera inference later: A new perspective towards autonomous driving,”arXiv preprint arXiv:2111.09799, 2021

    work page arXiv 2021

  45. [45]

    Irrelevant pixels are everywhere: Find and exclude them for more efficient computer vision,

    C. Tung, A. Goel, X. Hu, N. Eliopoulos, E. S. Amobi, G. K. Thiru- vathukal, V . Chaudhary, and Y .-H. Lu, “Irrelevant pixels are everywhere: Find and exclude them for more efficient computer vision,” in2022 IEEE 4th International Conference on Artificial Intelligence Circuits and Systems (AICAS). IEEE, 2022, pp. 340–343

    work page 2022

  46. [46]

    Dnn-sam: Split-and-merge dnn execution for real- time object detection,

    W. Kang, S. Chung, J. Y . Kim, Y . Lee, K. Lee, J. Lee, K. G. Shin, and H. S. Chwa, “Dnn-sam: Split-and-merge dnn execution for real- time object detection,” in2022 IEEE 28th Real-Time and Embedded Technology and Applications Symposium (RTAS). IEEE, 2022, pp. 160–172

    work page 2022

  47. [47]

    Self-cueing real-time attention scheduling in criticality-aware visual machine perception,

    S. Liu, X. Fu, M. Wigness, P. David, S. Yao, L. Sha, and T. Abdelzaher, “Self-cueing real-time attention scheduling in criticality-aware visual machine perception,” in2022 IEEE 28th Real-Time and Embedded Technology and Applications Symposium (RTAS). IEEE, 2022, pp. 173–186

    work page 2022

  48. [48]

    What uncertainties do we need in bayesian deep learning for computer vision?

    A. Kendall and Y . Gal, “What uncertainties do we need in bayesian deep learning for computer vision?” inAdvances in Neural Information Processing Systems, I. Guyon, U. V . Luxburg, S. Bengio, H. Wallach, R. Fergus, S. Vishwanathan, and R. Garnett, Eds., vol. 30. Curran Associates, Inc., 2017. [Online]. Available: https://proceedings.neurips.cc/paper file...

    work page 2017

  49. [49]

    Simple and scalable predictive uncertainty estimation using deep ensembles,

    B. Lakshminarayanan, A. Pritzel, and C. Blundell, “Simple and scalable predictive uncertainty estimation using deep ensembles,” inAdvances in Neural Information Processing Systems, I. Guyon, U. V . Luxburg, S. Bengio, H. Wallach, R. Fergus, S. Vishwanathan, and R. Garnett, Eds., vol. 30. Curran Associates, Inc., 2017. [Online]. Available: https://proceedi...

    work page 2017

  50. [50]

    The road to safety: A review of uncertainty and applications to autonomous driving perception,

    B. Ara ´ujo, J. F. Teixeira, J. Fonseca, R. Cerqueira, and S. C. Beco, “The road to safety: A review of uncertainty and applications to autonomous driving perception,”Entropy, vol. 26, no. 8, 2024. [Online]. Available: https://www.mdpi.com/1099-4300/26/8/634

    work page 2024

  51. [51]

    Reluplex: An efficient smt solver for verifying deep neural networks,

    G. Katz, C. Barrett, D. L. Dill, K. Julian, and M. J. Kochenderfer, “Reluplex: An efficient smt solver for verifying deep neural networks,” inInternational conference on computer aided verification. Springer, 2017, pp. 97–117

    work page 2017

  52. [52]

    Safety verification of deep neural networks,

    X. Huang, M. Kwiatkowska, S. Wang, and M. Wu, “Safety verification of deep neural networks,” inInternational conference on computer aided verification. Springer, 2017, pp. 3–29

    work page 2017

  53. [53]

    Basic concepts and taxonomy of dependable and secure computing,

    A. Avizienis, J.-C. Laprie, B. Randell, and C. Landwehr, “Basic concepts and taxonomy of dependable and secure computing,”IEEE Transactions on Dependable and Secure Computing, vol. 1, no. 1, pp. 11–33, 2004

    work page 2004

  54. [54]

    Hidden biases of end-to-end driving models,

    B. Jaeger, K. Chitta, and A. Geiger, “Hidden biases of end-to-end driving models,” inProc. of the IEEE International Conf. on Computer Vision (ICCV), 2023

    work page 2023

  55. [55]

    Nuplan: A closed-loop ml- based planning benchmark for autonomous vehicles,

    H. Caesar, J. Kabzan, K. S. Tan, W. K. Fong, E. Wolff, A. Lang, L. Fletcher, O. Beijbom, and S. Omari, “Nuplan: A closed-loop ml- based planning benchmark for autonomous vehicles,” inCVPR ADP3 workshop, 2021

    work page 2021

  56. [56]

    CARLA: An open urban driving simulator,

    A. Dosovitskiy, G. Ros, F. Codevilla, A. Lopez, and V . Koltun, “CARLA: An open urban driving simulator,” inProceedings of the 1st Annual Conference on Robot Learning, 2017, pp. 1–16

    work page 2017