Pith. sign in

REVIEW 3 major objections 50 references

EvoEye: Self-Evolving Runtime Monitoring for Autonomous Driving Systems

T0 review · 3 major / 0 minor · reviewed 2026-07-12 · grok-4.5

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read A self-evolving runtime monitor for autonomous driving finds its own blind spots, then trains on them.

desk verdict Solid closed-loop ADS monitor with real Apollo/CARLA numbers; the 13.2 Hard-Set gain is real inside the ODD but not yet proof of transfer. read the letter →

arxiv 2607.03755 v2 pith:H2VUIEFR submitted 2026-07-04 cs.SE

Mingfei Cheng , Lionel Briand , Xiaofei Xie This is my paper
classification cs.SE
keywords runtimemonitoringautonomousdrivingsystemsself-evolutioncollisionpredictionscenarioacquisitioncross-modulefusionApolloCARLA
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

Autonomous driving systems can fail in ways that fixed safety rules and one-shot learning monitors never learn to catch. EvoEye closes that gap with a loop: a monitor that reads the full stack of perception, prediction, planning, and control signals over time, then a search process that turns the monitor's current mistakes into new simulation scenarios and retrains. The paper argues that this self-evolution is more effective than either random sampling or hunting only for collisions. On industrial Apollo software in highway cut-in and unprotected-intersection settings, the approach raises frame-level recall by tens of percentage points at a controlled false-positive rate, gives several seconds of warning, and, under a fixed simulation budget, recovers substantially more of the unsafe situations the initial monitor had missed.

What carries the argument

The closed self-evolution loop of FusionMonitor plus BlindSpotEvolver: FusionMonitor encodes and fuses perception, prediction, planning, and control messages over a short temporal window to score collision risk; BlindSpotEvolver treats average prediction error on an executed scenario as search feedback and uses density-aware mutation to decide between local exploitation and global exploration before updating the monitor.

What would settle it

Under the same simulation budget on held-out runs from the same or expanded logical scenario spaces, if BlindSpotEvolver no longer beats uniform sampling and violation-oriented search on F1 for previously missed unsafe frames (or if FusionMonitor loses its reported recall lead at FPR 0.05 with usable warning lead time), the self-evolution claim fails.

Watch

Extended reading notes

Core claim

EvoEye shows that a learning-based collision-risk monitor that fuses multi-module temporal runtime signals can be made substantially more effective by iteratively acquiring simulations that expose its current prediction errors, rather than by fixed rules or by collecting more data without regard to remaining blind spots. Under the same evolution budget, this monitor-guided, density-aware search outperforms uniform and violation-oriented sampling on both overall test performance and on previously missed unsafe contexts.

Load-bearing premise

The method assumes that a monitor's average prediction error on past runs, plus how densely similar scenarios have already been sampled, is a good enough guide for which new simulations will most improve future monitoring—and that gains on two CARLA/Apollo scenario families transfer to real monitoring value.

Editorial extensions

If this is right

  • Runtime monitors for modular ADSs need not stay frozen after initial training; they can improve by turning their own mistakes into new test data.
  • Searching for collisions alone is not the best use of simulation budget when the goal is better monitoring: error-guided, density-aware acquisition can recover more missed hazards.
  • Cross-module temporal fusion of internal ADS messages can separate pre-collision contexts from safe ones better than single-indicator rules or reconstruction of isolated signals, at millisecond latency.
  • If the loop generalizes, operators can allocate limited simulation time to closing known monitoring gaps rather than exhaustively exploring the full scenario space.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The same error-as-feedback loop could be applied to non-collision hazards (near-misses, traffic-rule violations) if those labels are available, testing whether the machinery is collision-specific or general.
  • Because the monitor depends on extractable internal module messages, transferring it to end-to-end driving stacks would require a different signal interface and would test how much of the gain is architecture-specific.
  • If density-aware mutation mainly helps in high-dimensional parameter spaces, simpler domains might not need it; comparing against pure error-guided selection would isolate that component's value.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 0 minor

Summary. EvoEye proposes a self-evolving runtime monitor for full-stack modular ADSs. FusionMonitor encodes perception, prediction, planning, and control messages with module-specific encoders, fuses them per frame, aggregates a short temporal context, and predicts collision risk within horizon H. BlindSpotEvolver then closes a loop: it scores executed scenarios by average absolute prediction error (Eq. 10), mutates seeds with density-aware local/global sampling (Eqs. 11–15), and retrains the monitor under a fixed simulation budget. On Baidu Apollo with CARLA in two logical scenarios (highway cut-in, unprotected intersection), FusionMonitor reports large frame-level Recall gains over TTC, RSS, and a reconstruction baseline at fixed FPR, low CPU latency, and multi-second median lead times; under the same budget, BlindSpotEvolver improves F1 over Uniform and AVFuzzer, especially on a Hard Set of previously missed unsafe frames.

Significance. If the results hold under broader evaluation, the paper would be a clear contribution to ADS software engineering: it reframes runtime monitoring as an iterative, monitor-guided data-acquisition problem rather than a one-shot trained detector, and it targets full-stack inter-module messages instead of sensor-only or single-module signals. Strengths include a concrete industrial stack (Apollo+CARLA with the full perception-to-control pipeline), fixed-FPR operating points, AUROC/AUPRC, lead-time and latency reporting, ablations on L and module inputs, and head-to-head evolution comparisons under a shared budget. The self-evolution framing and the Hard-Set analysis are the main novelty claims and would matter for practical monitor maintenance if shown to transfer beyond the two logical ODDs studied.

major comments (3)
  1. §V-B and Table IV: The central self-evolution claim (up to 13.2 F1 on previously missed unsafe contexts) rests on the Hard Set, defined as unsafe frames missed by pre-evolution EvoEye* plus safe frames from the same executions, all drawn from the same two logical scenario spaces used for training and evolution. Because Eq. 10 and density-aware mutation optimize residual error inside this already-sampled ODD, large Hard-Set gains can reflect re-sampling residual pockets rather than acquiring broadly informative executions. Please either (i) report leave-one-logical-scenario-out or a third NHTSA-style logical scenario as a transfer test, or (ii) substantially qualify the claim so it is scoped to within-ODD residual-error reduction, and add multi-seed mean±std for Table IV (the text says three repeats but variance is not shown).
  2. §IV-C1, Eq. (10): Average absolute prediction error over a scenario’s contexts is treated as the acquisition objective without evidence that it correlates with learning value (e.g., expected reduction in validation loss or Hard-Set F1 after update). Scenarios with high average error may be noisy, label-boundary, or already well-covered after a few samples. Please justify this proxy (ablation vs. alternatives such as false-negative-only error, uncertainty, or min-distance criticality) or show that high-fr seeds systematically yield larger post-update gains than low-fr seeds under the same budget.
  3. §V, Logical Scenarios and Threats: External validity is load-bearing for a “first self-evolving runtime monitoring framework for full-stack ADSs.” Evaluation is confined to two CARLA logical scenarios and one ADS. The Threats section acknowledges this but does not bound how much of the reported Recall/F1 advantage depends on these two ODDs. At minimum, add sensitivity to H and δ (free parameters listed in the setup), and clarify which runtime message fields are required so that portability claims to other modular stacks are falsifiable rather than aspirational.

Circularity Check

0 steps flagged · score 0.0 of 10

No derivation circularity: self-evolution is an intentional feedback algorithm evaluated on held-out labels and independent baselines, not a claim that reduces to its inputs by construction.

full rationale

EvoEye is an engineering method paper, not a first-principles derivation. FusionMonitor is trained by minimizing BCE on labeled runtime contexts (Eq. 9) with ground-truth frame labels from simulator collisions within horizon H (Def. 1, Eq. 2), then scored on held-out test data against TTC, RSS, and AEM under fixed FPR thresholds. BlindSpotEvolver’s monitor-oriented feedback fr(x) (Eq. 10) and density-aware mutation (Eqs. 11–15) deliberately use current prediction error to prioritize new simulations; that is a closed-loop acquisition policy, not a circular proof that the evolved monitor must improve. Gains are measured by external metrics (Precision/Recall/F1/AUROC/AUPRC) on a held-out Test Set and against Uniform and AVFuzzer under a shared six-hour budget (Tables III–IV). Author self-citations appear only as related ADS testing work and do not load-bear uniqueness or force the central claims. The Hard Set (unsafe frames missed by pre-evolution EvoEye*) is a conditioned evaluation slice that can overstate absolute gains, but it is not a fitted parameter renamed as a prediction, nor does any equation reduce to its inputs by construction; relative comparisons on that set remain well-defined. No self-definitional identity, fitted-input-as-prediction, uniqueness import, or ansatz-via-citation step is present.

Assumptions & free parameters 6 free parameters · 5 assumptions · 4 invented entities

This is an empirical software-engineering systems paper. Load-bearing content is design choices and evaluation protocol rather than physical postulates. Free parameters are the labeling horizon, context length, search neighborhood, budgets, and FPR operating points that define both training labels and reported success. Domain axioms include modular ADS observability, collision-as-primary violation, and simulator fidelity. Invented entities are the named method components, which have no independent existence outside this design.

free parameters (6)
  • prediction horizon H
    Frame labels y_t depend on collision within H seconds (Eq. 2); set to 3s from prior intervention literature, not derived from first principles.
  • temporal context length L
    Chosen as L=5 after ablation; directly shapes model input and reported best performance.
  • neighborhood radius δ
    Set to 20% of each parameter range for density-aware mutation; controls local vs global exploration.
  • evolution budget T_budget and N_round
    T_budget=6 hours and N_round=20 define when the monitor updates; comparisons hold only under this budget.
  • FPR operating points 0.01 and 0.05
    Thresholds selected on validation to hit target FPRs; Precision/Recall/F1 claims are at these hand-chosen operating points.
  • scenario-level consecutive-frame parameter K
    K=1 and K=3 used to aggregate frame predictions to scenario detection; affects scenario-level metrics.
assumptions (5)
  • domain assumption Internal inter-module runtime messages from a modular ADS (perception, prediction, planning, control) are observable and sufficient to predict impending collisions without privileged simulator state at inference.
    Stated in §II–III and used throughout FusionMonitor design; monitor never uses O(x) at prediction time.
  • domain assumption Collisions within a short horizon are the primary safety violation to monitor; other hazards are out of scope.
    Definition 1 and threats (§V-D); labels and all metrics are collision-based.
  • domain assumption CARLA + full Apollo sensor-to-control pipeline is an adequate proxy for evaluating full-stack runtime monitoring under the chosen logical scenarios.
    Environment setup §V; authors note simulation cannot fully reproduce real sensing/traffic.
  • ad hoc to paper Average absolute prediction error over a scenario’s contexts is a useful acquisition objective for improving the monitor.
    Eq. 10 Monitor-Oriented Feedback; core of BlindSpotEvolver, justified empirically vs baselines rather than theoretically optimal.
  • standard math Standard supervised binary classification with BCE on labeled contexts is an appropriate training objective for risk scores.
    Eq. 9; conventional ML assumption.
invented entities (4)
  • FusionMonitor
    purpose: Hierarchical encoders + temporal fusion of multi-module ADS messages into a collision-risk score.
    Named architecture introduced in §IV-B; evaluated only inside this paper’s experiments.
  • BlindSpotEvolver
    purpose: Closed-loop scenario acquisition using monitor error feedback and density-aware mutation.
    Named search procedure §IV-C and Alg. 1; no external validation outside reported Apollo/CARLA runs.
  • Monitor-Oriented Feedback fr(x)
    purpose: Scalar scenario score equal to mean |ŷ−y| used to prioritize seeds/regions.
    Eq. 10; paper-specific guidance signal, not a pre-existing standard metric.
  • Density-Aware Mutation / regional exploit potential qδ(x)
    purpose: Switch between local mutation and global sampling based on neighborhood mean/std of feedback.
    Eqs. 11–15; design choice of this paper.

how reviews work

0 comments
Cite this review

Pith. "Pith review of EvoEye: Self-Evolving Runtime Monitoring for Autonomous Driving Systems." pith.science (2026). https://pith.science/paper/H2VUIEFR

@misc{pith2026260703755,
  author       = {Pith},
  title        = {Pith review of: EvoEye: Self-Evolving Runtime Monitoring for Autonomous Driving Systems},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/H2VUIEFR}},
  note         = {Machine review of arXiv:2607.03755}
}
read the original abstract

Runtime monitoring is essential for detecting impending hazards in autonomous driving systems (ADSs). However, existing ADS runtime monitors have fixed detection capabilities: rule-based monitors cover only manually specified hazards, while learning-based monitors depend heavily on their initial training data and may retain substantial prediction errors. We therefore propose EvoEye, which identifies the current monitor's errors, generates informative executions accordingly, and updates the monitor through self-evolution. To enable effective self-evolution, EvoEye combines a capable runtime monitor with targeted scenario acquisition. FusionMonitor learns cross-module temporal interactions for collision prediction, while BlindSpotEvolver converts current prediction errors into search guidance and uses density-aware mutation to acquire informative executions for subsequent monitor updates. We evaluate EvoEye on Baidu Apollo with CARLA in representative highway and urban scenarios. FusionMonitor improves frame-level Recall by up to 37.8 percentage points at a false positive rate of 0.05, with 2.49 ms latency and 2.8-4.2 seconds of median warning time. Under the same budget, BlindSpotEvolver outperforms uniform and violation-oriented sampling by up to 13.2 F1 points on previously missed unsafe contexts.

Figures

Figures reproduced from arXiv: 2607.03755 by the authors.

Figure 1
Figure 1. Overview of EvoEye. cost(D′ r ) denotes the simulation time required to acquire the new scenario records D′ r at evolution round r. Achieving this goal requires addressing two closely coupled subproblems. First, the runtime monitor must have sufficient representational and learning capacity to capture the complex temporal and cross-module dependencies among heterogeneous ADS runtime signals. This capability is a pre… view at source ↗
Figure 2
Figure 2. Architecture of FusionMonitor. Table I: Module-specific encoders used by FusionMonitor. Module Input Description Encoder Structure Perception States of detected obstacles {oˆt,i} nt i=1 Shared MLP + AttnPool −−−−−−−−−−−→ z perc t Prediction Predicted trajectories of obstacles {pt,i} nt i=1 Shared MLP + AttnPool −−−−−−−−−−−→ z pred t Planning Planned ego trajectory Pt Flatten + MLP −−−−−−−→ z plan t Control Action co… view at source ↗
Figure 3
Figure 3. Warning lead time distribution. consecutive positive frames are required, EvoEye remains the best-performing method, with F1-scores of 56.2% and 53.3% on S1 and 87.5% and 70.5% on S2. It still detects all S1 violations and 91.9% of S2 violations at the 0.05 target FPR, indicating that its warnings persist across consecutive hazardous frames rather than arising from isolated predictions. In contrast, the scenario-lev… view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Illustrative cases comparing EvoEye before and after self-evolution. On the Test Set, EvoEye consistently improves the initial monitor. Compared with EvoEye∗ , EvoEye increases F1 from 30.1% to 34.4% on S1 and from 67.8% to 71.5% on S2 at the 0.05 target FPR. Under the…

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

50 extracted references · 4 linked inside Pith

  1. [1]

    A survey of motion planning and control techniques for self-driving urban vehicles,

    B. Paden, M. ˇC´ap, S. Z. Yong, D. Yershov, and E. Frazzoli, “A survey of motion planning and control techniques for self-driving urban vehicles,” IEEE Transactions on intelligent vehicles, vol. 1, no. 1, pp. 33–55, 2016

    2016

  2. [2]

    Self- driving cars: A survey,

    C. Badue, R. Guidolini, R. V . Carneiro, P. Azevedo, V . B. Cardoso, A. Forechi, L. Jesus, R. Berriel, T. M. Paixao, F. Mutzet al., “Self- driving cars: A survey,”Expert systems with applications, vol. 165, p. 113816, 2021

    2021

  3. [3]

    Challenges in autonomous vehicle testing and validation,

    P. Koopman and M. Wagner, “Challenges in autonomous vehicle testing and validation,”SAE International journal of transportation safety, vol. 4, pp. 15–24, 2016

    2016

  4. [4]

    Testing of autonomous driving systems: where are we and where should we go?

    G. Lou, Y . Deng, X. Zheng, M. Zhang, and T. Zhang, “Testing of autonomous driving systems: where are we and where should we go?” inProceedings of the 30th ACM Joint European Software Engineering Conference and Symposium on the Foundations of Software Engineering, 2022, pp. 31–43

    2022

  5. [5]

    Runtime monitoring for safety of intelligent vehicles,

    K. Watanabe, E. Kang, C.-W. Lin, and S. Shiraishi, “Runtime monitoring for safety of intelligent vehicles,” inProceedings of the 55th annual design automation conference, 2018, pp. 1–6

    2018

  6. [6]

    Near miss determination through use of a scale of danger,

    J. C. Hayward, “Near miss determination through use of a scale of danger,”Highway Research Record, 1972

    1972

  7. [7]

    On a formal model of safe and scalable self-driving cars,

    S. Shalev-Shwartz, S. Shammah, and A. Shashua, “On a formal model of safe and scalable self-driving cars,”arXiv preprint arXiv:1708.06374, 2017

    arXiv 2017

  8. [8]

    Misbehaviour prediction for autonomous driving systems,

    A. Stocco, M. Weiss, M. Calzana, and P. Tonella, “Misbehaviour prediction for autonomous driving systems,” inProceedings of the ACM/IEEE 42nd international conference on software engineering, 2020, pp. 359–371

    2020

Show all 50 references
  1. [9]

    Thirdeye: Attention maps for safe autonomous driving systems,

    A. Stocco, P. J. Nunes, M. d’Amorim, and P. Tonella, “Thirdeye: Attention maps for safe autonomous driving systems,” inProceedings of the 37th IEEE/ACM international conference on automated software engineering, 2022, pp. 1–12

    2022

  2. [10]

    Baidu-apollo,

    Baidu, “Baidu-apollo,” 2023. [Online]. Available: https://www.apollo.auto

    2023

  3. [11]

    Carla: An open urban driving simulator,

    A. Dosovitskiy, G. Ros, F. Codevilla, A. Lopez, and V . Koltun, “Carla: An open urban driving simulator,” inConference on robot learning. PMLR, 2017, pp. 1–16

    2017

  4. [12]

    Pylot: A modular platform for exploring latency-accuracy tradeoffs in autonomous vehicles,

    I. Gog, S. Kalra, P. Schafhalter, M. A. Wright, J. E. Gonzalez, and I. Stoica, “Pylot: A modular platform for exploring latency-accuracy tradeoffs in autonomous vehicles,” in2021 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2021, pp. 8806–8813

    2021

  5. [13]

    Autoware on board: Enabling autonomous vehicles with embedded systems,

    S. Kato, S. Tokunaga, Y . Maruyama, S. Maeda, M. Hirabayashi, Y . Kitsukawa, A. Monrroy, T. Ando, Y . Fujii, and T. Azumi, “Autoware on board: Enabling autonomous vehicles with embedded systems,” in 2018 ACM/IEEE 9th International Conference on Cyber-Physical Systems (ICCPS). ...

    2018

  6. [14]

    End to end learning for self-driving cars,

    M. Bojarski, D. Del Testa, D. Dworakowski, B. Firner, B. Flepp, P. Goyal, L. D. Jackel, M. Monfort, U. Muller, J. Zhanget al., “End to end learning for self-driving cars,”arXiv preprint arXiv:1604.07316, 2016

    2016 arXiv

  7. [15]

    End-to-end driving via conditional imitation learning,

    F. Codevilla, M. M ¨uller, A. L ´opez, V . Koltun, and A. Dosovitskiy, “End-to-end driving via conditional imitation learning,” in2018 IEEE international conference on robotics and automation (ICRA). IEEE, 2018, pp. 4693–4700

    2018

  8. [16]

    openpilot,

    comma.ai, “openpilot,” 2025. [Online]. Available: https://comma.ai/ope npilot

    2025

  9. [17]

    End-to-end urban driving by imitating a reinforcement learning coach,

    Z. Zhang, A. Liniger, D. Dai, F. Yu, and L. Van Gool, “End-to-end urban driving by imitating a reinforcement learning coach,” inProceedings of the IEEE/CVF international conference on computer vision, 2021, pp. 15 222–15 232

    2021

  10. [18]

    Transfuser: Imitation with transformer-based sensor fusion for au- tonomous driving,

    K. Chitta, A. Prakash, B. Jaeger, Z. Yu, K. Renz, and A. Geiger, “Transfuser: Imitation with transformer-based sensor fusion for au- tonomous driving,”IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 45, no. 11, pp. 12 878–12 895, 2023

    2023

  11. [19]

    Trajectory-guided control prediction for end-to-end autonomous driving: A simple yet strong baseline,

    P. Wu, X. Jia, L. Chen, J. Yan, H. Li, and Y . Qiao, “Trajectory-guided control prediction for end-to-end autonomous driving: A simple yet strong baseline,”Advances in Neural Information Processing Systems, vol. 35, pp. 6119–6132, 2022

    2022

  12. [20]

    A framework for automated driving system testable cases and scenarios,

    E. Thorn, S. C. Kimmel, and M. Chaka, “A framework for automated driving system testable cases and scenarios,” National Highway Traffic Safety Administration, Tech. Rep., 2018

    2018

  13. [21]

    Scenarios for development, test and validation of automated vehicles,

    T. Menzel, G. Bagschik, and M. Maurer, “Scenarios for development, test and validation of automated vehicles,” in2018 IEEE intelligent vehicles symposium (IV). IEEE, 2018, pp. 1821–1827

    2018

  14. [22]

    Efficient online testing for dnn- enabled systems using surrogate-assisted and many-objective optimiza- tion,

    F. U. Haq, D. Shin, and L. Briand, “Efficient online testing for dnn- enabled systems using surrogate-assisted and many-objective optimiza- tion,” inProceedings of the 44th international conference on software engineering, 2022, pp. 811–822

    2022

  15. [23]

    Av-fuzzer: Finding safety violations in autonomous driving systems,

    G. Li, Y . Li, S. Jha, T. Tsai, M. Sullivan, S. K. S. Hari, Z. Kalbarczyk, and R. Iyer, “Av-fuzzer: Finding safety violations in autonomous driving systems,” in2020 IEEE 31st international symposium on software reliability engineering (ISSRE). IEEE, 2020, pp. 25–36

    2020

  16. [24]

    Identifying the hazard boundary of ml-enabled autonomous systems using cooperative coevolutionary search,

    S. Sharifi, D. Shin, L. C. Briand, and N. Aschbacher, “Identifying the hazard boundary of ml-enabled autonomous systems using cooperative coevolutionary search,”IEEE Transactions on Software Engineering, vol. 49, no. 12, pp. 5120–5138, 2023

    2023

  17. [25]

    Behavexplor: Behavior diversity guided testing for autonomous driving systems,

    M. Cheng, Y . Zhou, and X. Xie, “Behavexplor: Behavior diversity guided testing for autonomous driving systems,” inProceedings of the 32nd ACM SIGSOFT International Symposium on Software Testing and Analysis, 2023, pp. 488–500

    2023

  18. [26]

    Many-objective reinforcement learning for online testing of dnn-enabled systems,

    F. U. Haq, D. Shin, and L. C. Briand, “Many-objective reinforcement learning for online testing of dnn-enabled systems,” in2023 IEEE/ACM 45th International Conference on Software Engineering (ICSE). IEEE, 2023, pp. 1814–1826

    2023

  19. [27]

    Doppelg¨anger test generation for revealing bugs in autonomous driving software,

    Y . Huai, Y . Chen, S. Almanee, T. Ngo, X. Liao, Z. Wan, Q. A. Chen, and J. Garcia, “Doppelg¨anger test generation for revealing bugs in autonomous driving software,” in2023 IEEE/ACM 45th International Conference on Software Engineering (ICSE). IEEE, 2023, pp. 2591–2603

    2023

  20. [28]

    Pre-crash scenario typol- ogy for crash avoidance research,

    W. G. Najm, J. D. Smith, and M. Yanagisawa, “Pre-crash scenario typol- ogy for crash avoidance research,” U.S. Department of Transportation National Highway Traffic Safety Administration, Tech. Rep., 2007

    2007

  21. [29]

    Study on effectiveness of pre-crash active seatbelt using real time controlled simulation,

    J. K. Lee and K. W. Lee, “Study on effectiveness of pre-crash active seatbelt using real time controlled simulation,” in23rd International Technical Conference on the Enhanced Safety of Vehicles (ESV) National Highway Traffic Safety Administration, no. 13-0097, 2013

    2013

  22. [30]

    A review of essential technologies for collision avoidance assistance systems,

    Z. Zhao, L. Zhou, Q. Zhu, Y . Luo, and K. Li, “A review of essential technologies for collision avoidance assistance systems,”Advances in Mechanical Engineering, vol. 9, no. 10, p. 1687814017725246, 2017

    2017

  23. [31]

    Pytorch: An imperative style, high-performance deep learning library,

    A. Paszke, S. Gross, F. Massa, A. Lerer, J. Bradbury, G. Chanan, T. Killeen, Z. Lin, N. Gimelshein, L. Antigaet al., “Pytorch: An imperative style, high-performance deep learning library,”Advances in neural information processing systems, vol. 32, 2019

    2019

  24. [32]

    Preliminary evaluation of path-aware crossover operators for search-based test data generation for autonomous driving,

    S. Han, J. Kim, G. Kim, J. Cho, J. Kim, and S. Yoo, “Preliminary evaluation of path-aware crossover operators for search-based test data generation for autonomous driving,” in2021 IEEE/ACM 14th International Workshop on Search-Based Software Testing (SBST). IEEE, 2021, pp. 44–47

    2021

  25. [33]

    Neural network guided evolutionary fuzzing for finding traffic violations of autonomous vehicles,

    Z. Zhong, G. Kaiser, and B. Ray, “Neural network guided evolutionary fuzzing for finding traffic violations of autonomous vehicles,”IEEE Transactions on Software Engineering, vol. 49, no. 4, pp. 1860–1875, 2023

    2023

  26. [34]

    Systematic testing of autonomous driving systems using map topology-based scenario classification,

    Y . Tang, Y . Zhou, T. Zhang, F. Wu, Y . Liu, and G. Wang, “Systematic testing of autonomous driving systems using map topology-based scenario classification,” in2021 36th IEEE/ACM International Conference on Automated Software Engineering (ASE). IEEE, 2021, pp. 1342–1346

    2021

  27. [35]

    Specification-based autonomous driving system testing,

    Y . Zhou, Y . Sun, Y . Tang, Y . Chen, J. Sun, C. M. Poskitt, Y . Liu, and Z. Yang, “Specification-based autonomous driving system testing,”IEEE Transactions on Software Engineering, vol. 49, no. 6, pp. 3391–3410, 2023

    2023

  28. [36]

    Route coverage testing for autonomous vehicles via map modeling,

    Y . Tang, Y . Zhou, F. Wu, Y . Liu, J. Sun, W. Huang, and G. Wang, “Route coverage testing for autonomous vehicles via map modeling,” in2021 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2021, pp. 11 450–11 456

    2021

  29. [37]

    Collision avoidance testing for autonomous driving systems on complete maps,

    Y . Tang, Y . Zhou, Y . Liu, J. Sun, and G. Wang, “Collision avoidance testing for autonomous driving systems on complete maps,” in2021 IEEE Intelligent Vehicles Symposium (IV). IEEE, 2021, pp. 179–185

    2021

  30. [38]

    Moditector: Module- directed testing for autonomous driving systems,

    R. Wang, M. Cheng, X. Xie, Y . Zhou, and L. Ma, “Moditector: Module- directed testing for autonomous driving systems,”Proceedings of the ACM on Software Engineering, vol. 2, no. ISSTA, pp. 137–158, 2025

    2025

  31. [39]

    Using cooperative co- evolutionary search to generate metamorphic test cases for autonomous driving systems,

    H. Yousefizadeh, S. Gu, L. C. Briand, and A. Nasr, “Using cooperative co- evolutionary search to generate metamorphic test cases for autonomous driving systems,”IEEE Transactions on Software Engineering, 2025

    2025

  32. [40]

    Drivora: A unified and extensible infrastructure for search-based autonomous driving testing,

    M. Cheng, L. Briand, and Y . Zhou, “Drivora: A unified and extensible infrastructure for search-based autonomous driving testing,”arXiv preprint arXiv:2601.05685, 2026

    2026

  33. [41]

    Mdpfuzz: testing models solving markov decision processes,

    Q. Pang, Y . Yuan, and S. Wang, “Mdpfuzz: testing models solving markov decision processes,” inProceedings of the 31st ACM SIGSOFT International Symposium on Software Testing and Analysis, 2022, pp. 378–390

    2022

  34. [42]

    Stclocker: Deadlock avoidance testing for autonomous driving systems,

    M. Cheng, R. Wang, X. Xie, Y . Zhou, and L. Ma, “Stclocker: Deadlock avoidance testing for autonomous driving systems,”arXiv preprint arXiv:2506.23995, 2025

    2025 arXiv

  35. [43]

    Causality-aware safety testing for autonomous driving systems,

    W. Tang, M. Cheng, R. Wang, Y . Zhou, C. Liu, Y . Liu, and Z. Ding, “Causality-aware safety testing for autonomous driving systems,”IEEE Transactions on Software Engineering, 2026

    2026

  36. [44]

    Decictor: Towards evaluating the robustness of decision-making in autonomous driving systems,

    M. Cheng, X. Xie, Y . Zhou, J. Wang, G. Meng, and K. Yang, “Decictor: Towards evaluating the robustness of decision-making in autonomous driving systems,” in2025 IEEE/ACM 47th International Conference on Software Engineering (ICSE). IEEE, 2025, pp. 1–13

    2025

  37. [45]

    Automatically testing self-driving cars with search-based procedural content generation,

    A. Gambi, M. Mueller, and G. Fraser, “Automatically testing self-driving cars with search-based procedural content generation,” inProceedings of the 28th ACM SIGSOFT International Symposium on Software Testing and Analysis, 2019, pp. 318–328

    2019

  38. [46]

    Constrained co- evolutionary metamorphic differential testing for autonomous systems with an interpretability approach,

    H. Yousefizadeh, S. Gu, L. C. Briand, and A. Nasr, “Constrained co- evolutionary metamorphic differential testing for autonomous systems with an interpretability approach,”arXiv preprint arXiv:2509.16478, 2025

    2025 arXiv

  39. [47]

    Metamorphic fuzz testing of autonomous ve- hicles,

    J. C. Han and Z. Q. Zhou, “Metamorphic fuzz testing of autonomous ve- hicles,” inProceedings of the IEEE/ACM 42nd International Conference on Software Engineering Workshops, 2020, pp. 380–385

    2020

  40. [48]

    Generative model-based testing on decision-making policies,

    Z. Li, X. Wu, D. Zhu, M. Cheng, S. Chen, F. Zhang, X. Xie, L. Ma, and J. Zhao, “Generative model-based testing on decision-making policies,” in 2023 38th IEEE/ACM International Conference on Automated Software Engineering (ASE). IEEE, 2023, pp. 243–254

    2023

  41. [49]

    Dense reinforcement learning for safety validation of autonomous vehicles,

    S. Feng, H. Sun, X. Yan, H. Zhu, Z. Zou, S. Shen, and H. X. Liu, “Dense reinforcement learning for safety validation of autonomous vehicles,” Nature, vol. 615, no. 7953, pp. 620–627, 2023

    2023

  42. [50]

    Learning configurations of operating environment of autonomous vehicles to maximize their collisions,

    C. Lu, Y . Shi, H. Zhang, M. Zhang, T. Wang, T. Yue, and S. Ali, “Learning configurations of operating environment of autonomous vehicles to maximize their collisions,”IEEE Transactions on Software Engineering, vol. 49, no. 1, pp. 384–402, 2023

    2023

Pith tools

Reviewed July 12, 2026 · model on record in the stance chip above.