Pith. sign in

REVIEW 3 major objections 4 minor 1 cited by

STCLocker: Deadlock Avoidance Testing for Autonomous Driving Systems

T0 review · 3 major / 4 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read STCLocker is a search-based testing technique that induces multi-vehicle circular waits — deadlocks — in autonomous driving systems, and reports generating more such scenarios than existing baselines.

desk verdict A genuinely new testing objective with an oracle that needs independent validation; the core idea is worth engaging with, but the headline numbers are softer than they look. read the letter →

arxiv 2506.23995 v1 pith:EURBZMM3 submitted 2025-06-30 cs.SE cs.AIcs.RO

Mingfei Cheng , Renzhi Wang , Xiaofei Xie , Yuan Zhou , Lei Ma This is my paper · ORCID
classification cs.SEcs.AIcs.RO
keywords autonomousdrivingtestingdeadlockscenariogenerationmulti-agentcooperationsearch-basedwait-forgraphconflict-guidedfuzzingCARLAsimulationspatio-temporalconflict
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

This paper argues that autonomous driving systems (ADSs) fail in ways that single-vehicle testing cannot see: when two or more self-driving vehicles wait for each other at a shared region, they can enter a circular wait, or deadlock, and stall traffic indefinitely. To expose such failures, the authors propose STCLocker, a search-based testing method that detects deadlock cycles from observable vehicle trajectories and then guides scenario mutation toward spatial conflict points with synchronized arrival times. In CARLA simulations with two ADSs, STCLocker reports generating more DeadLock Scenarios than random generation and an existing multi-vehicle testing baseline, and manual review confirms a substantial fraction of the detected cases. If the results hold, the technique gives a practical way to assess cooperative deadlock avoidance before deployment.

What carries the argument

The central object is the wait-for graph $G_t = (V, E_t)$, where each vertex is an AV and a directed edge $e_{i \to j}$ is added when $\Phi_{\mathrm{edge}} = \Phi_{\mathrm{stop}} \wedge \Phi_{\mathrm{intent}}$ holds: AV $i$ is stationary for a minimum window and a Kalman-filter prediction of AV $j$'s future path overlaps AV $i$'s intended path. A DeadLock Scenario is declared when any scene's wait-for graph contains a cycle. To generate such scenes, Conflict Feedback computes a spatial-conflict score from intersecting trajectory segments and a temporal-conflict score from arrival-time gaps at conflict regions, combined as $\phi_S = \alpha \phi_{\mathrm{spatial}} + (1-\alpha)\phi_{\mathrm{temporal}}$; Conflict-aware Scenario Generation then either shifts trigger times to synchronize arrivals at a conflict region or mutates routes and spawn points in a local search that minimizes the estimated spatial score.

What would settle it

Run STCLocker on OpenCDA without human filtering, log each oracle-flagged DeadLock Scenario, and let the simulation continue for another 30 seconds; if most flagged vehicles eventually resume and clear the intersection, the trajectory-overlap oracle is overcounting transient stops, and the reported advantage over baselines would shrink or disappear.

Watch

Extended reading notes

Core claim

STCLocker's central claim is that multi-AV deadlocks can be systematically induced and detected without any internal ADS state by building a wait-for graph from low-level observations: an AV is waiting for another when it is stopped and its predicted path overlaps the other vehicle's predicted path, and a deadlock scenario is any scene whose wait-for graph contains a cycle. On this basis, the paper reports that STCLocker generates on average 188.2 DeadLock Scenarios across Roach and OpenCDA, compared with 58.8 for the best-performing baseline, with 94.6 manually confirmed among the STCLocker cases (46.4 on Roach and 48.2 on OpenCDA). The paper further concludes that current ADSs have limited cooperative capability to avoid deadlocks, with the communication-enabled OpenCDA showing more deadlocks than the end-to-end Roach.

Load-bearing premise

The load-bearing premise is that the oracle's rule — a stopped AV whose predicted path overlaps another predicted path is 'waiting for' that AV — identifies genuine circular waits, even though the paper reports only 40% precision on OpenCDA and relies on human review to discard false cases.

Editorial extensions

If this is right

  • Multi-AV deadlocks can be found automatically from trajectory-level observations, so testing cooperative behavior does not require access to the ADS's internal planning or communication state.
  • The temporal-aware mutation is the strongest single ingredient: removing it drops detected DLSs from 65.4 to 37.4 on Roach and from 122.8 to 99.4 on OpenCDA.
  • End-to-end and modular cooperative ADSs both exhibit circular waits, and the paper's results indicate the communication-enabled modular system (OpenCDA) produces more deadlock scenarios than the end-to-end system (Roach).
  • The technique can serve as a benchmark for deadlock avoidance in CARLA-like simulators, with a logged wait-for-graph cycle as the failure criterion.

Reading between the lines

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

  • If the oracle is as noisy as the 40% precision on OpenCDA suggests, the reported #DLS counts are upper bounds on true deadlocks; a precision-aware variant would be needed before using them as a safety metric.
  • The same wait-for-graph oracle could be applied to other multi-agent systems, such as warehouse robots, intersection managers, or multi-robot fleets, where agents' intended paths are observable, making the method a general deadlock-fuzzing recipe.
  • The conflict-guided search is conceptually a coverage-guided fuzzer for concurrency: the spatial and temporal scores act like coverage signals, so the same scaffolding could target other coordination failures such as livelocks, priority inversions, or response-time outages.
  • A direct test of the paper's explanation for OpenCDA's fragility would be to run STCLocker against a centralized-coordination ADS; if centralized control removes most DLSs, the rule-based decentralized cooperation is the culprit.
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 / 4 minor

Summary. The paper presents STCLocker, a search-based testing method for generating deadlock scenarios (DLSs) involving multiple autonomous vehicles (AVs) controlled by an ADS under test. The method has three components: a Deadlock Oracle that builds a wait-for graph from low-level observations using a stop condition and a Kalman-filter-based trajectory-overlap check; a Conflict Feedback mechanism that scores scenarios by spatial trajectory intersections and temporal arrival proximity; and Conflict-aware Scenario Generation that mutates trigger times and spawn positions to increase spatial and temporal competition. The authors evaluate STCLocker on two ADSs (Roach and OpenCDA) in the CARLA simulator across four road types, comparing against Random and DoppelTest baselines. They report that STCLocker detects 188.2 DLSs total versus 58.8 for the best-performing baseline, and that human-validated counts (#DLS-Hum) also favor STCLocker. Ablation studies decompose the contributions of spatial and temporal mutations and feedback variants.

Significance. If the results hold, this would be one of the first systematic efforts to test multi-AV cooperative behavior specifically for deadlock avoidance, a genuine gap in ADS testing. The problem is well motivated, the method is clearly presented, and the evaluation covers two different ADS architectures, four driving contexts, and both raw and human-validated metrics. The ablation study is a useful strength because it demonstrates that each component contributes to the overall generation count. The main limitation is that the primary metric relies on an oracle that the paper itself shows to be only 40% precise on OpenCDA, and the generation feedback is based on a similar trajectory-overlap heuristic, so the raw #DLS comparison may be partially an artifact of the oracle. The human-validated numbers still point in the same direction, but they are not accompanied by a documented validation protocol or statistical tests.

major comments (3)
  1. [Section 4.2, Section 4.3, Section 5.1.2, Table 1] The primary metric #DLS is computed by the Deadlock Oracle, whose wait-for edge condition Phi_edge = Phi_stop AND Phi_intent (Section 4.2) only checks that a stopped AV's predicted path overlaps another AV's predicted path in space and time. It does not verify that the other AV actually blocks the stopped AV, that the stop is caused by that AV, or that the circular wait persists beyond the 5-second Delta_t window. The paper's own threat-to-validity section (Section 5.4) concedes that 'premature classification of deadlocks may occur if ego vehicles eventually resume movement.' The oracle precision is reported as 68% on Roach and only 40% on OpenCDA, meaning 60% of OpenCDA DLS detections are false positives. Because the Conflict Feedback in Section 4.3 explicitly rewards spatial trajectory intersection and temporal coincidence, STCLocker is likely to generate more configurations that satisfy this trajectory-overlap proxy than Random or DoppelTest, so the raw #DLS gap (122.8 vs. 52.6 on OpenCDA) may overstate the true generation advantage. The manually validated counts in Table 1 still favor STCLocker (48.2 vs. 29.2 on OpenCDA; 46.4 vs. 5.4 on Roach), but the paper does not describe a blind, independent, or criteria-based human validation protocol, and it does not report how many raw detections were excluded per method. I would like to see oracle precision and false-positive counts reported separately for each method, and either a human-validated primary metric or a systematic bias analysis showing that the generation gap is not driven by oracle overfitting.
  2. [Section 5.1, Tables 1-2, Figures 6-7] All effectiveness claims are based on averages over five repeated runs, but the paper reports no standard deviations, confidence intervals, or significance tests. The RQ1 answer states that 'STCLocker significantly outperforms existing methods,' yet no statistical evidence is provided. The raw and human-validated counts in Table 1 are point estimates, and Figures 6-7 show nontrivial run-to-run variation. Please report the per-run values and apply an appropriate statistical test (for example, a Mann-Whitney U test or paired bootstrap over the five runs per scenario), and report effect sizes for both the raw #DLS and the human-validated #DLS-Hum metrics.
  3. [Section 5.1.1, Discussion] The Discussion draws a conclusion about the cooperative capability of the two ADSs from the number of DLSs generated by STCLocker, stating that 'Roach ... surprisingly demonstrates better cooperative behavior by inducing fewer DLSs than OpenCDA (65.4 vs. 122.8).' This comparison is confounded: the number of detected DLSs depends on the interaction between the search method and the ADS's behavior, not solely on the ADS's deadlock-avoidance competence. A less predictable or less controllable ADS could produce fewer detections for reasons unrelated to cooperative ability. This interpretive claim should be removed or re-framed, or supported by a metric that controls for search effectiveness, such as a detection rate per scenario family or a fixed set of scenarios applied to both ADSs.
minor comments (4)
  1. [Abstract and Section 1] The abstract reports '3.39x more DLS than the best-performing baseline,' but the numbers in Table 1 and Section 1 (188.2 vs. 58.8) give a ratio of approximately 3.20x; please reconcile these figures.
  2. [Section 5.1.2] The precision comparison against the 'naive time-based counter' would be clearer if the threshold and duration used by the naive oracle were specified; without that, the comparison is hard to reproduce.
  3. [Section 4.3.2] The normalization factor is written as 'N_ti' in the equation for phi_temporal and as 'N_temporal' in the text; please unify the notation.
  4. [Section 4.1, Algorithm 1] The description of seed corpus updates states that a new scenario is retained when 'phi_S' < 'phi_S'; given that lower scores are better, the direction is correct, but a brief intuition for why lower temporal and spatial conflict scores indicate higher deadlock potential would improve readability for readers unfamiliar with the scoring scheme.

Circularity Check

1 steps flagged · score 4.0 of 10

Partial circularity: STCLocker's conflict guidance optimizes the same spatio-temporal overlap condition that the Deadlock Oracle uses to count DLSs; raw #DLS advantage is partly built-in, but human-validated counts give independent support.

  1. self definitional [Sections 4.2 and 4.3; Algorithm 2 (Temporal-aware Mutation)]
    "Φintent(AVi, AVj) = (True, if Conflict(τpre_i, τpred_j) = True) ... Here, Conflict(τpre_i, τpred_j) returns True if the two sets of trajectories overlap in both space and time, suggesting a potential interaction that may have contributed to AVi's decision to stop. ... The vehicles should approach the conflict region simultaneously or within a short time window to increase the likelihood of triggering a deadlock."

    The Deadlock Oracle labels an edge only when a stopped AV's Kalman-predicted trajectory overlaps another AV's predicted trajectory in space and time (Φintent). The Conflict Feedback score φ_S = α·φ_spatial + (1-α)·φ_temporal is minimized by increasing trajectory intersections and decreasing arrival-time gaps at those intersections, and the Temporal-aware Mutation explicitly shifts trigger times to make AVs reach shared regions simultaneously. Thus the search objective is the same overlap-and-synchronization condition the oracle uses to count DLSs; scenarios generated by STCLocker are more likely to satisfy the oracle's edge condition by construction than scenarios from Random or DoppelTest, which do not optimize this proxy.

full rationale

The central empirical claim is that STCLocker generates more DLSs than the best baseline, measured by #DLS. The main circularity risk is that the Deadlock Oracle's edge condition and STCLocker's feedback objective are both built from spatio-temporal trajectory overlap: the oracle fires an edge when a stopped AV's predicted trajectory overlaps another AV's predicted trajectory in space and time (Section 4.2), while the feedback score is minimized by increasing trajectory intersections and simultaneous arrivals (Section 4.3). Therefore the raw #DLS advantage is partly self-fulfilling, since the guidance explicitly optimizes the oracle's firing condition. However, the oracle is applied uniformly to all methods, the generator does not directly force the stop condition or the cycle, and the paper's human-validated #DLS-Hum counts still show an advantage (e.g., 48.2 vs. 29.2 on OpenCDA). No load-bearing self-citation, no imported uniqueness theorem, and no fitted parameter renamed as a prediction were found. The score is moderate because the primary raw metric is coupled to the guidance objective, but the central result is not fully reducible to that coupling.

Assumptions & free parameters 6 free parameters · 6 assumptions · 0 invented entities

The central result depends on several hand-set parameters (alpha, Delta_t, epsilon, N_A, N_temporal, N_local), on the validity of the oracle's intent inference, and on the reliability of manual validation. These are assumptions the paper states or leaves implicit.

free parameters (6)
  • alpha = 0.5
    Weight between spatial and temporal conflict scores in the overall feedback; set from preliminary observations with no sensitivity analysis.
  • oracle detection window Delta_t = 5 seconds
    Minimum stationary duration required for the stop condition; chosen based on preliminary observations and directly affects deadlock detection.
  • velocity threshold epsilon = 0.01 m/s
    Threshold below which a vehicle is considered stationary; hand-chosen.
  • maximum AV capacity N_A = 6
    Limits the number of AVs in spatial mutation; set following prior work and affects scenario complexity.
  • temporal normalization factor N_temporal = not specified
    Used to normalize the temporal conflict score into [0,1]; the paper does not state its value.
  • local search budget N_local = not specified
    Number of random candidate configurations tested in spatial mutation; unspecified, but affects how well the spatial score is optimized.
assumptions (6)
  • domain assumption A single-scene cycle in the wait-for graph is a deadlock
    Definition 3.2 defines a DLS as existence of a cycle in one scene, yet the text describes deadlocks as indefinite circular waits. The oracle does not require the cycle to persist, and Section 5.4 concedes premature classification may occur.
  • domain assumption Kalman-filter prediction of short trajectories captures wait-for intent
    Section 4.2 infers Phi_intent by checking overlap between the predicted pre-stop trajectory and the predicted future trajectory of another AV. Inaccurate prediction is listed as a threat in Section 5.4.
  • domain assumption Ground-truth perception from CARLA is a valid substitute for real perception
    Section 5, Environment, states perception errors are excluded to focus on decision-making. This matches prior work but narrows the generalization of the testing claim.
  • domain assumption Authors' manual review is reliable ground truth for deadlocks
    Sections 5.1.2 and 5.4 report that recordings were manually reviewed and false cases excluded. No inter-rater, blinded, or external protocol is described.
  • domain assumption Porting DoppelTest to CARLA preserves its intended testing logic
    Section 5 Baselines says DoppelTest's core algorithms were reused without modifying core testing logic; changes in execution environment could still alter its behavior.
  • standard math A* path planning predicts trajectories well enough for spatial mutation scoring
    Section 4.4.2 estimates future AV trajectories using A* when scoring candidate configurations without running the simulator. If A* differs from the ADS planner, the feedback signal may be miscalibrated.

how reviews work

0 comments
Cite this review

Pith. "Pith review of STCLocker: Deadlock Avoidance Testing for Autonomous Driving Systems." pith.science (2026). https://pith.science/paper/EURBZMM3

@misc{pith2026250623995,
  author       = {Pith},
  title        = {Pith review of: STCLocker: Deadlock Avoidance Testing for Autonomous Driving Systems},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/EURBZMM3}},
  note         = {Machine review of arXiv:2506.23995}
}
read the original abstract

Autonomous Driving System (ADS) testing is essential to ensure the safety and reliability of autonomous vehicles (AVs) before deployment. However, existing techniques primarily focus on evaluating ADS functionalities in single-AV settings. As ADSs are increasingly deployed in multi-AV traffic, it becomes crucial to assess their cooperative performance, particularly regarding deadlocks, a fundamental coordination failure in which multiple AVs enter a circular waiting state indefinitely, resulting in motion planning failures. Despite its importance, the cooperative capability of ADSs to prevent deadlocks remains insufficiently underexplored. To address this gap, we propose the first dedicated Spatio-Temporal Conflict-Guided Deadlock Avoidance Testing technique, STCLocker, for generating DeadLock Scenarios (DLSs), where a group of AVs controlled by the ADS under test are in a circular wait state. STCLocker consists of three key components: Deadlock Oracle, Conflict Feedback, and Conflict-aware Scenario Generation. Deadlock Oracle provides a reliable black-box mechanism for detecting deadlock cycles among multiple AVs within a given scenario. Conflict Feedback and Conflict-aware Scenario Generation collaborate to actively guide AVs into simultaneous competition over spatial conflict resources (i.e., shared passing regions) and temporal competitive behaviors (i.e., reaching the conflict region at the same time), thereby increasing the effectiveness of generating conflict-prone deadlocks. We evaluate STCLocker on two types of ADSs: Roach, an end-to-end ADS, and OpenCDA, a module-based ADS supporting cooperative communication. Experimental results show that, on average, STCLocker generates more DLS than the best-performing baseline.

Figures

Figures reproduced from arXiv: 2506.23995 by the authors.

Figure 1
Figure 1. Illustration of a deadlock scenario. halt traffic flow, block other vehicles and entire road segments, and lead to broader negative impacts on transportation efficiency and societal well-being. Therefore, it is crucial to develop systematic testing techniques to evaluate the deadlock-handling capabilities of ADSs in multi-AV scenarios. To the best of our knowledge, there is limited research on evalu￾ating the ADS co… view at source ↗
Figure 2
Figure 2. Overview of STCLocker. driving commands, bypassing the need for explicit modular decom￾position. Other approaches adopt a hybrid paradigm by replacing only specific components within conventional ADSs. Methods such as PlanT [48] and Interfuser [50] map sensor data to planned tra￾jectories, while still relying on conventional rule-based controllers (e.g., PID) for low-level vehicle control. Roach [73], for instance, … view at source ↗
Figure 3
Figure 3. Illustration of spatial conflict regions. [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (4 more)
Figure 5
Figure 5. Figure 5: After temporal [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]
Figure 6
Figure 6. Figure 6: Statistical comparison of #DLS on Roach. [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]
Figure 7
Figure 7. Figure 7: Statistical comparison of #DLS on OpenCDA. [PITH_FULL_IMAGE:figures/full_fig_p007_7.png]
Figure 8
Figure 8. Figure 8: Precision of oracles. median and the green triangle indicates the average. Overall, we find that STCLocker outperforms all baselines in both #DLS and #DLS-Hum across different ADSs and scenarios. Roach. For the end-to-end ADS Roach, STCLocker achieves the best overall …

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. EvoEye: Self-Evolving Runtime Monitoring for Autonomous Driving Systems

    cs.SE 2026-07 conditional novelty 6.0 of 10

    A self-evolving ADS monitor that fuses multi-module runtime signals and uses prediction-error-guided, density-aware scenario search to raise collision-warning recall and hard-set F1 under a fixed budget.

Reference graph

Works this paper leans on

78 extracted references · 71 canonical work pages · cited by 1 Pith paper

  1. [1]

    Matthias Althoff and Sebastian Lutz. 2018. Automatic generation of safety-critical test scenarios for collision avoidance of road vehicles. In 2018 IEEE Intelligent Vehicles Symposium (IV). IEEE, Changshu, Suzhou, China, 1326–1333

    work page 2018

  2. [2]

    2020.Taxonomy and definitions for terms related to cooperative driving automation for on-road motor vehicles

    Cooperative Driving Automation. 2020.Taxonomy and definitions for terms related to cooperative driving automation for on-road motor vehicles . SAE International

    work page 2020

  3. [3]

    Baidu. 2019. Apollo: Open Source Autonomous Driving. https://github.com/ ApolloAuto/apollo

    work page 2019

  4. [4]

    Baidu. 2025. Apollo Go. https://www.apollogo.com/. Accessed: May 2025

    work page 2025

  5. [5]

    Rebecca Bellan. 2024. 6 Waymo robotaxis block traffic to San Francisco freeway on-ramp. TechCrunch (17 April 2024). https://techcrunch.com/2024/04/17/seven- waymo-robotaxis-block-traffic-to-san-francisco-freeway-on-ramp/ Accessed: May 31, 2025

    work page 2024

  6. [6]

    Na Chen, Bart van Arem, Tom Alkim, and Meng Wang. 2021. A Hierarchi- cal Model-Based Optimization Control Approach for Cooperative Merging by Connected Automated Vehicles. IEEE Transactions on Intelligent Transportation Systems 22, 12 (2021), 7712–7725. doi:10.1109/TITS.2020.3007647

    work page arXiv 2021

  7. [7]

    Mingfei Cheng, Xiaofei Xie, Yuan Zhou, Junjie Wang, Guozhu Meng, and Kairui Yang. 2025. Decictor: Towards Evaluating the Robustness of Decision-Making in Autonomous Driving Systems . In 2025 IEEE/ACM 47th International Conference on Software Engineering (ICSE) . 651–651

    work page 2025

  8. [8]

    Mingfei Cheng, Yuan Zhou, and Xiaofei Xie. 2023. BehAVExplor: Behavior Diversity Guided Testing for Autonomous Driving Systems. In Proceedings of the 32nd ACM SIGSOFT International Symposium on Software Testing and Analysis . 488–500

    work page 2023

Show all 78 references
  1. [9]

    comma.ai. 2022. OpenPilot: An open source driver assistance system. Retrieved Nov 7, 2022 from https://github.com/commaai/openpilot

    2022

  2. [10]

    Yao Deng, Xi Zheng, Mengshi Zhang, Guannan Lou, and Tianyi Zhang. 2022. Scenario-based test reduction and prioritization for multi-module autonomous driving systems. In Proceedings of the 30th ACM Joint European Software Engi- neering Conference and Symposium on the Foundation...

    2022

  3. [11]

    Alexey Dosovitskiy, German Ros, Felipe Codevilla, Antonio Lopez, and Vladlen Koltun. 2017. CARLA: An Open Urban Driving Simulator. InProceedings of the 1st Annual Conference on Robot Learning (Proceedings of Machine Learning Research, Vol. 78), Sergey Levine, Vincent Vanhoucke...

    2017

  4. [12]

    EDM.com Staff. 2024. Google’s Self-Driving Waymo Cars Appear to Stall, Block Traffic Out of Portola Festival. EDM.com (2024). https://edm.com/news/google- STCLocker: Deadlock Avoidance Testing for Autonomous Driving Systems Conference acronym ’XX, June 03–05, 2018, Woodstock, ...

    2024

  5. [13]

    Azadeh Emami, Majid Sarvi, and Saeed Asadi Bagloee. 2019. Using Kalman filter algorithm for short-term traffic flow prediction in a connected vehicle environment. Journal of Modern Transportation 27 (2019), 222–232

    2019

  6. [14]

    Shuo Feng, Haowei Sun, Xintao Yan, Haojie Zhu, Zhengxia Zou, Shengyin Shen, and Henry X Liu. 2023. Dense reinforcement learning for safety validation of autonomous vehicles. Nature 615, 7953 (2023), 620–627

    2023

  7. [15]

    Alessio Gambi, Tri Huynh, and Gordon Fraser. 2019. Generating effective test cases for self-driving cars from police reports. In Proceedings of the 2019 27th ACM Joint Meeting on European Software Engineering Conference and Symposium on the Foundations of Software Engineering ...

    2019

  8. [16]

    Alessio Gambi, Marc Mueller, and Gordon Fraser. 2019. Automatically testing self- driving cars with search-based procedural content generation. In Proceedings of the 28th ACM SIGSOFT International Symposium on Software Testing and Analysis . ACM, Beijing, China, 318–328

    2019

  9. [17]

    Xin Gao, Xinyu Zhang, Yiguo Lu, Yuning Huang, Lei Yang, Yijin Xiong, and Peng Liu. 2024. A Survey of Collaborative Perception in Intelligent Vehicles at Intersections. IEEE Transactions on Intelligent Vehicles (2024), 1–20. doi:10.1109/ TIV.2024.3395783

    2024

  10. [18]

    Joshua Garcia, Yang Feng, Junjie Shen, Sumaya Almanee, Yuan Xia, and Qi Alfred Chen. 2020. A comprehensive study of autonomous vehicle bugs. In Proceedings of the ACM/IEEE 42nd International Conference on Software Engineering . IEEE, Seoul, South Korea, 385–396

    2020

  11. [19]

    Ashkan Gholamhosseinian and Jochen Seitz. 2022. A Comprehensive Survey on Cooperative Intersection Management for Heterogeneous Connected Vehicles. IEEE Access 10 (2022), 7937–7972. doi:10.1109/ACCESS.2022.3142450

    2022

  12. [20]

    Ionel Gog, Sukrit Kalra, Peter Schafhalter, Matthew A Wright, Joseph E Gonzalez, and Ion Stoica. 2021. Pylot: A modular platform for exploring latency-accuracy tradeoffs in autonomous vehicles. In2021 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 8806–8813

    2021

  13. [21]

    An Guo, Yuan Zhou, Haoxiang Tian, Chunrong Fang, Yunjian Sun, Weisong Sun, Xinyu Gao, Anh Tuan Luu, Yang Liu, and Zhenyu Chen. 2024. Sovar: Build generalizable scenarios from accident reports for autonomous driving testing. In Proceedings of the 39th IEEE/ACM International Con...

    2024

  14. [22]

    Jia Cheng Han and Zhi Quan Zhou. 2020. Metamorphic fuzz testing of au- tonomous vehicles. In Proceedings of the IEEE/ACM 42nd International Conference on Software Engineering Workshops. 380–385

    2020

  15. [23]

    Seunghee Han, Jaeuk Kim, Geon Kim, Jaemin Cho, Jiin Kim, and Shin Yoo. 2021. Preliminary evaluation of path-aware crossover operators for search-based test data generation for autonomous driving. In 2021 IEEE/ACM 14th International Workshop on Search-Based Software Testing (SB...

    2021

  16. [24]

    Fitash Ul Haq, Donghwan Shin, and Lionel Briand. 2022. Efficient online testing for DNN-enabled systems using surrogate-assisted and many-objective optimiza- tion. In Proceedings of the 44th International Conference on Software Engineering . IEEE, Pittsburgh Pennsylvania, 811–822

    2022

  17. [25]

    Fitash Ul Haq, Donghwan Shin, and Lionel C Briand. 2023. Many-objective rein- forcement learning for online testing of dnn-enabled systems. In 2023 IEEE/ACM 45th International Conference on Software Engineering (ICSE) . IEEE, 1814–1826

    2023

  18. [26]

    Carl Hildebrandt, Meriel von Stein, and Sebastian Elbaum. 2023. PhysCov: Phys- ical Test Coverage for Autonomous Vehicles. In Proceedings of the 32nd ACM SIGSOFT International Symposium on Software Testing and Analysis . 449–461

    2023

  19. [27]

    Yihan Hu, Jiazhi Yang, Li Chen, Keyu Li, Chonghao Sima, Xizhou Zhu, Siqi Chai, Senyao Du, Tianwei Lin, Wenhai Wang, Lewei Lu, Xiaosong Jia, Qiang Liu, Jifeng Dai, Yu Qiao, and Hongyang Li. 2023. Planning-oriented Autonomous Driving. In Proceedings of the IEEE/CVF Conference on...

    2023

  20. [28]

    Yuqi Huai, Sumaya Almanee, Yuntianyi Chen, Xiafa Wu, Qi Alfred Chen, and Joshua Garcia. 2023. sceno RITA: Generating Diverse, Fully-Mutable, Test Scenar- ios for Autonomous Vehicle Planning. IEEE Transactions on Software Engineering (2023)

    2023

  21. [29]

    Yuqi Huai, Yuntianyi Chen, Sumaya Almanee, Tuan Ngo, Xiang Liao, Ziwen Wan, Qi Alfred Chen, and Joshua Garcia. 2023. Doppelgänger test generation for revealing bugs in autonomous driving software. In 2023 IEEE/ACM 45th International Conference on Software Engineering (ICSE) . ...

    2023

  22. [30]

    Yangjie Ji, Zewei Zhou, Ziru Yang, Yanjun Huang, Yuanjian Zhang, Wanting Zhang, Lu Xiong, and Zhuoping Yu. 2024. Toward autonomous vehicles: A survey on cooperative vehicle-infrastructure system. iScience 27, 5 (2024), 109751. doi:10.1016/j.isci.2024.109751

    2024

  23. [31]

    Shinpei Kato, Shota Tokunaga, Yuya Maruyama, Seiya Maeda, Manato Hirabayashi, Yuki Kitsukawa, Abraham Monrroy, Tomohito Ando, Yusuke Fujii, and Takuya Azumi. 2018. Autoware on board: Enabling autonomous vehicles with embedded systems. In 2018 ACM/IEEE 9th International Confere...

    2018

  24. [32]

    Seulbae Kim, Major Liu, Junghwan" John" Rhee, Yuseok Jeon, Yonghwi Kwon, and Chung Hwan Kim. 2022. DriveFuzz: Discovering Autonomous Driving Bugs through Driving Quality-Guided Fuzzing. In Proceedings of the 2022 ACM SIGSAC Conference on Computer and Communications Security . ...

    2022

  25. [33]

    Smith, Yu- Chiang Frank Wang, and Min-Hung Chen

    Hsu kuang Chiu, Ryo Hachiuma, Chien-Yi Wang, Stephen F. Smith, Yu- Chiang Frank Wang, and Min-Hung Chen. 2025. V2V-LLM: Vehicle-to-Vehicle Cooperative Autonomous Driving with Multi-Modal Large Language Models. arXiv:2502.09980 [cs.CV] https://arxiv.org/abs/2502.09980

    2025

  26. [34]

    Selvaraj Vasantha Kumar. 2017. Traffic flow prediction using Kalman filtering technique. Procedia Engineering 187 (2017), 582–587

    2017

  27. [35]

    Guanpeng Li, Yiran Li, Saurabh Jha, Timothy Tsai, Michael Sullivan, Siva Ku- mar Sastry Hari, Zbigniew Kalbarczyk, and Ravishankar Iyer. 2020. AV-FUZZER: Finding safety violations in autonomous driving systems. In 2020 IEEE 31st Inter- national Symposium on Software Reliabilit...

    2020

  28. [36]

    Zhongrui Li, Jiarun Dai, Zongan Huang, Nianhao You, Yuan Zhang, and Min Yang. 2024. Viohawk: Detecting traffic violations of autonomous driving systems through criticality-guided simulation testing. In Proceedings of the 33rd ACM SIGSOFT International Symposium on Software Tes...

    2024

  29. [37]

    Guannan Lou, Yao Deng, Xi Zheng, Mengshi Zhang, and Tianyi Zhang. 2022. Testing of autonomous driving systems: where are we and where should we go?. In Proceedings of the 30th ACM Joint European Software Engineering Conference and Symposium on the Foundations of Software Engin...

    2022

  30. [38]

    Chengjie Lu, Yize Shi, Huihui Zhang, Man Zhang, Tiexin Wang, Tao Yue, and Shaukat Ali. 2022. Learning configurations of operating environment of au- tonomous vehicles to maximize their collisions. IEEE Transactions on Software Engineering 49, 1 (2022), 384–402

    2022

  31. [39]

    You Lu, Yifan Tian, Yuyang Bi, Bihuan Chen, and Xin Peng. 2024. DiaVio: LLM-Empowered Diagnosis of Safety Violations in ADS Simulation Testing. In Proceedings of the 33rd ACM SIGSOFT International Symposium on Software Testing and Analysis (Vienna, Austria) (ISSTA 2024). Assoc...

    2024

  32. [40]

    Wassim G Najm, John D Smith, Mikio Yanagisawa, A John, et al. 2007. Pre-crash scenario typology for crash avoidance research . Technical Report. United States. Department of Transportation. National Highway Traffic Safety

    2007

  33. [41]

    Wassim G Najm, Samuel Toma, John Brewer, et al. 2013. Depiction of priority light-vehicle pre-crash scenarios for safety applications based on vehicle-to-vehicle communications. Technical Report DOT HS 811 732. National Highway Traffic Safety Administration, U.S. Department of...

    2013

  34. [42]

    Hieu Ngo, Hua Fang, and Honggang Wang. 2023. Cooperative Perception With V2V Communication for Autonomous Vehicles. IEEE Transactions on Vehicular Technology 72, 9 (2023), 11122–11131. doi:10.1109/TVT.2023.3264020

    2023

  35. [43]

    Philippe Nitsche, Pete Thomas, Rainer Stuetz, and Ruth Welsh. 2017. Pre-crash scenarios at road junctions: A clustering method for car crash data. Accident Analysis & Prevention 107 (2017), 137–151

    2017

  36. [44]

    Jan-Pieter Paardekooper, S Montfort, Jeroen Manders, Jorrit Goos, E de Gelder, O Camp, O Bracquemond, and Gildas Thiolon. 2019. Automatic identification of critical scenarios in a public dataset of 6000 km of public-road driving. In 26th International Technical Conference on t...

    2019

  37. [45]

    Qi Pang, Yuanyuan Yuan, and Shuai Wang. 2022. MDPFuzz: Testing Models Solving Markov Decision Processes. In Proceedings of the 31st ACM SIGSOFT International Symposium on Software Testing and Analysis (Virtual, South Korea) (ISSTA 2022). Association for Computing Machinery, Ne...

    2022

  38. [46]

    Federico Pratissoli, Riccardo Brugioni, Nicola Battilani, and Lorenzo Sabattini

  39. [47]

    HongSheng Qi, Yi Yu, Qing Tang, and XianBiao Hu. 2022. Intersection traffic deadlock formation and its probability: A petri net-based modeling approach. IET Intelligent Transport Systems 16, 10 (2022), 1342–1363

    2022

  40. [48]

    Sophia Koepke, Zeynep Akata, and Andreas Geiger

    Katrin Renz, Kashyap Chitta, Otniel-Bogdan Mercea, A. Sophia Koepke, Zeynep Akata, and Andreas Geiger. 2022. PlanT: Explainable Planning Transformers via Object-Level Representations. In Conference on Robotic Learning (CoRL)

    2022

  41. [49]

    Christian Roesener, Felix Fahrenkrog, Axel Uhlig, and Lutz Eckstein. 2016. A scenario-based assessment approach for automated driving by using time series classification of human-driving behaviour. In 2016 IEEE 19th international con- ference on intelligent transportation syst...

    2016

  42. [50]

    Hao Shao, Letian Wang, Ruobing Chen, Hongsheng Li, and Yu Liu. 2023. Safety- enhanced autonomous driving using interpretable sensor fusion transformer. In Conference on Robot Learning . PMLR, 726–737

    2023

  43. [51]

    Smith, and Daniela L

    Daniel Eduardo Soltero, Stephen L. Smith, and Daniela L. Rus. 2011. Collision avoidance for persistent monitoring in multi-robot systems with intersecting trajectories. In 2011 IEEE/RSJ International Conference on Intelligent Robots and Systems

    2011

  44. [52]

    Yang Sun, Christopher M Poskitt, Jun Sun, Yuqi Chen, and Zijiang Yang. 2022. LawBreaker: An approach for specifying traffic laws and fuzzing autonomous ve- hicles. In Proceedings of the 37th IEEE/ACM International Conference on Automated Conference acronym ’XX, June 03–05, 201...

    2022

  45. [53]

    Shuncheng Tang, Zhenya Zhang, Jixiang Zhou, Lei Lei, Yuan Zhou, and Yinxing Xue. 2024. Legend: A top-down approach to scenario generation of autonomous driving systems assisted by large language models. In Proceedings of the 39th IEEE/ACM International Conference on Automated ...

    2024

  46. [54]

    Shuncheng Tang, Zhenya Zhang, Jixiang Zhou, Yuan Zhou, Yan-Fu Li, and Yinxing Xue. 2023. EvoScenario: Integrating Road Structures into Critical Sce- nario Generation for Autonomous Driving System Testing. In 2023 IEEE 34th International Symposium on Software Reliability Engine...

    2023

  47. [55]

    Wenbing Tang, Mingfei Cheng, Yuan Zhou, and Yang Liu. 2025. Moral Testing of Autonomous Driving Systems. arXiv preprint arXiv:2505.03683 (2025)

    2025 arXiv

  48. [56]

    Yun Tang, Yuan Zhou, Yang Liu, Jun Sun, and Gang Wang. 2021. Collision avoidance testing for autonomous driving systems on complete maps. In 2021 IEEE Intelligent Vehicles Symposium (IV) . IEEE, Nagoya, Japan, 179–185

    2021

  49. [57]

    Yun Tang, Yuan Zhou, Fenghua Wu, Yang Liu, Jun Sun, Wuling Huang, and Gang Wang. 2021. Route coverage testing for autonomous vehicles via map modeling. In 2021 IEEE International Conference on Robotics and Automation (ICRA) . IEEE, Xi’an, China, 11450–11456

    2021

  50. [58]

    Yun Tang, Yuan Zhou, Tianwei Zhang, Fenghua Wu, Yang Liu, and Gang Wang

  51. [59]

    Gokulnath Thandavarayan, Miguel Sepulcre, and Javier Gozalvez. 2020. Coopera- tive Perception for Connected and Automated Vehicles: Evaluation and Impact of Congestion Control. IEEE Access 8 (2020), 197665–197683. doi:10.1109/ACCESS. 2020.3035119

    2020

  52. [60]

    Eric Thorn, Shawn C Kimmel, Michelle Chaka, Booz Allen Hamilton, et al. 2018. A framework for automated driving system testable cases and scenarios . Technical Report. United States. Department of Transportation. National Highway Traffic Safety

    2018

  53. [61]

    Sebastian Thrun. 2002. Probabilistic robotics. Commun. ACM 45, 3 (2002), 52–57

    2002

  54. [62]

    Haoxiang Tian, Yan Jiang, Guoquan Wu, Jiren Yan, Jun Wei, Wei Chen, Shuo Li, and Dan Ye. 2022. MOSAT: finding safety violations of autonomous driving sys- tems using multi-objective genetic algorithm. InProceedings of the 30th ACM Joint European Software Engineering Conference...

    2022

  55. [63]

    Robin van der Made, Martijn Tideman, Ulrich Lages, Roman Katz, and Martin Spencer. 2015. Automated generation of virtual driving scenarios from test drive data. In 24th International Technical Conference on the Enhanced Safety of Vehicles (ESV) National Highway Traffic Safety ...

    2015

  56. [64]

    Renzhi Wang, Mingfei Cheng, Xiaofei Xie, Yuan Zhou, and Lei Ma. 2025. MoDi- tector: Module-Directed Testing for Autonomous Driving Systems.arXiv preprint arXiv:2502.08504 (2025)

    2025 arXiv

  57. [65]

    Tsun-Hsuan Wang, Sivabalan Manivasagam, Ming Liang, Bin Yang, Wenyuan Zeng, and Raquel Urtasun. 2020. V2VNet: Vehicle-to-Vehicle Communication for Joint Perception and Prediction. In Computer Vision – ECCV 2020 , Andrea Vedaldi, Horst Bischof, Thomas Brox, and Jan-Michael Frah...

    2020

  58. [66]

    Waymo. 2024. Waymo One Autonomous Ride-Hailing in San Francisco. https: //waymo.com/intl/zh-cn/waymo-one-san-francisco/ Accessed: 2025-05-16

    2024

  59. [67]

    Greg Welch, Gary Bishop, et al. 1995. An introduction to the Kalman filter. (1995)

    1995

  60. [68]

    Runsheng Xu, Yi Guo, Xu Han, Xin Xia, Hao Xiang, and Jiaqi Ma. 2021. OpenCDA: an open cooperative driving automation framework integrated with co-simulation. In 2021 IEEE International Intelligent Transportation Systems Con- ference (ITSC). IEEE, 1155–1162

    2021

  61. [69]

    Bayen, and Cathy Wu

    Zhongxia Yan, Abdul Rahman Kreidieh, Eugene Vinitsky, Alexandre M. Bayen, and Cathy Wu. 2023. Unified Automatic Control of Vehicular Systems With Re- inforcement Learning. IEEE Transactions on Automation Science and Engineering 20, 2 (2023), 789–804. doi:10.1109/TASE.2022.3168621

    2023

  62. [70]

    Zhongxia Yan, Han Zheng, and Cathy Wu. 2024. Multi-agent path finding for cooperative autonomous driving. In2024 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 12361–12367

    2024

  63. [71]

    Xudong Zhang and Yan Cai. 2023. Building Critical Testing Scenarios for Au- tonomous Driving from Real Accidents. In Proceedings of the 32nd ACM SIGSOFT International Symposium on Software Testing and Analysis . 462–474

    2023

  64. [72]

    Xiaodong Zhang, Wei Zhao, Yang Sun, Jun Sun, Yulong Shen, Xuewen Dong, and Zijiang Yang. 2023. Testing automated driving systems by breaking many laws efficiently. In Proceedings of the 32nd ACM SIGSOFT International Symposium on Software Testing and Analysis. 942–953

    2023

  65. [73]

    Zhejun Zhang, Alexander Liniger, Dengxin Dai, Fisher Yu, and Luc Van Gool

  66. [74]

    Ziyuan Zhong, Gail Kaiser, and Baishakhi Ray. 2023. Neural network guided evolutionary fuzzing for finding traffic violations of autonomous vehicles. IEEE Transactions on Software Engineering 49, 4 (2023), 1860–1875

    2023

  67. [75]

    Yuan Zhou, Yang Sun, Yun Tang, Yuqi Chen, Jun Sun, Christopher M Poskitt, Yang Liu, and Zijiang Yang. 2023. Specification-based Autonomous Driving System Testing. IEEE Transactions on Software Engineering (2023), 1–19

    2023

  68. [76]

    In Proceedings of the IEEE/CVF international conference on computer vision

    End-to-end urban driving by imitating a reinforcement learning coach. In Proceedings of the IEEE/CVF international conference on computer vision . 15222– 15232

  69. [2021]

    In Proceedings of the 36th IEEE/ACM International Conference on Automated Software Engineering (ASE)

    Systematic testing of autonomous driving systems using map topology- based scenario classification. In Proceedings of the 36th IEEE/ACM International Conference on Automated Software Engineering (ASE). IEEE, Melbourne, Australia, 1342–1346

  70. [2023]

    IEEE Transactions on Automation Science and Engineering (2023)

    Hierarchical traffic management of multi-AGV systems with deadlock prevention applied to industrial environments. IEEE Transactions on Automation Science and Engineering (2023)

    2023

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.