Pith. sign in

REVIEW 4 major objections 6 minor 68 references

Knowledge Integration Strategies in Autonomous Vehicle Prediction and Planning: A Comprehensive Survey

T0 review · 4 major / 6 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read This survey organizes all methods for integrating traffic rules and domain knowledge into autonomous-vehicle trajectory prediction and planning into a single taxonomy of six families, and argues that hybrid neuro-symbolic approaches hold…

desk verdict A genuinely useful map of knowledge-integration work in AV prediction/planning, but the taxonomy's inclusion rule is looser than advertised and needs an audit before the 'comprehensive' claim stands. read the letter →

arxiv 2502.10477 v2 pith:FUB64L6T submitted 2025-02-13 cs.AI cs.LG

Kumar Manas , Adrian Paschke This is my paper
classification cs.AIcs.LG
keywords autonomousdrivingtrajectorypredictionplanningknowledgeintegrationneuro-symbolicAIlargelanguagemodelsformallogicdiffusion
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

The paper sets out to organize the scattered literature on injecting explicit knowledge — traffic rules, ontologies, domain priors, commonsense — into autonomous-vehicle trajectory prediction and planning. It builds a unified taxonomy of six families: knowledge-graph/ontology methods, reinforcement learning, large language models with retrieval, formal logic, hybrid symbolic-neural systems, and diffusion-based generators. It compares how each family encodes rules explicitly or implicitly and which trade-offs (interpretability, adaptability, provable safety, scalability) follow. Readers in autonomous driving or AI safety get a structured map of existing methods and a claim that hybrid approaches are the most promising route to rule-compliant, interpretable driving systems.

What carries the argument

The central object is the taxonomy itself: a six-category classification of knowledge integration methods (knowledge graphs/ontologies, reinforcement learning, LLM/RAG, formal logic, hybrid symbolic-neural, diffusion) defined by how domain knowledge is represented and embedded. Table I condenses the taxonomy into a comparative overview, assigning each representative work a primary class with cross-references for overlaps and listing advantages and limitations. The taxonomy works by making the field's structure explicit, so that trade-offs such as interpretability versus adaptability, or provable safety versus scalability, become visible across methods rather than within isolated papers.

What would settle it

A reader could test the taxonomy by taking a random sample of recent prediction and planning papers from the same venues and checking whether each can be placed in exactly one category; if most papers cannot be assigned a single primary class, or if a prominent knowledge-integration method fits none of the six categories, the taxonomy's completeness and stability would be undermined.

Watch

Extended reading notes

Core claim

The survey's central claim is that the field of knowledge integration in autonomous driving can be mapped by a single taxonomy spanning purely symbolic to purely data-driven generative methods, and that this map reveals systematic strengths and weaknesses. No prior review, it argues, has covered both trajectory prediction and planning under the lens of explicit knowledge integration while including recent LLM, RAG, and diffusion techniques. The taxonomy's organizing axis is how knowledge enters the pipeline: as explicit symbolic constraints, as soft penalties in losses or rewards, or implicitly learned from data. The paper also claims emerging trends: interpretability, formal verification for safety, and hybrid neuro-symbolic methods are gaining importance, and foundation models offer new ways to convert regulatory text into usable constraints.

Load-bearing premise

The taxonomy assumes that 'explicit knowledge integration' can be cleanly separated from implicit learning, a boundary the paper itself concedes many ML pipelines blur into soft constraints.

Editorial extensions

If this is right

  • Researchers can place any new prediction or planning method into a shared taxonomy and compare it against known trade-offs.
  • Hybrid symbolic-neural approaches are positioned as the most promising balance between rule compliance and learned adaptability, directing attention toward their integration challenges.
  • LLM- and RAG-based methods are recognized as a viable path for translating regulatory text into machine-usable rules, despite latency and hallucination risks.
  • Formal verification and set-based reachability provide provable safety envelopes that can complement learned components, reinforcing the value of hybrid architectures.
  • The field is moving toward knowledge-driven autonomous driving as a distinct research direction, with foundation and diffusion models as key enablers.

Reading between the lines

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

  • The taxonomy's categories could be operationalized into a benchmark that scores systems by how explicitly they encode traffic rules, turning the explicit/implicit spectrum into a measurable quantity.
  • A chronological meta-analysis of the cited works (for example, counting publications per category by year) could test whether the claimed shift toward hybrid approaches is real or an artifact of selection.
  • Because the survey treats prediction and planning separately, it may underweight end-to-end LLM systems that fuse both tasks; a unified prediction-planning view could change the taxonomy's boundaries.
  • The admission that explicit priors often become implicit soft constraints suggests that 'knowledge integration' is a graded property; a continuous formalization of knowledge explicitness would make the taxonomy more robust.
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

4 major / 6 minor

Summary. This survey proposes a taxonomy of knowledge integration methods for autonomous vehicle trajectory prediction and planning, categorizing approaches into knowledge-graph/ontology-based, reinforcement learning, LLM/RAG, formal logic, hybrid neuro-symbolic, and diffusion/other learning-based methods. It provides a comparative table (Table I) of representative works and discusses challenges and future directions. The central claim is that it delivers a unified taxonomy with comparative analysis of how traffic rules and domain priors are explicitly and implicitly encoded.

Significance. If the taxonomy were rigorously grounded, the survey would fill a useful niche by spanning symbolic, hybrid, and generative approaches in one review. It covers recent LLM and diffusion methods that earlier surveys omit, and Table I gives a quick reference. However, the value of the survey hinges on the consistency of its inclusion criteria and category assignments, which are currently not systematic. The paper is honest about some limitations (e.g., acknowledging that explicit knowledge often becomes implicit in ML pipelines), but this honesty undercuts the taxonomy's stability. The lack of a search protocol and the presence of several contradictory classifications mean the 'comprehensive' framing is not yet substantiated.

major comments (4)
  1. [Section I (focus statement) and Section II-E] Section I states: "We focus exclusively on works that explicitly integrate rules or knowledge, excluding approaches that utilize environmental information without incorporating traffic regulations." Section II-E, however, includes [63] (inverse reinforcement learning) as a hybrid approach, describing it as "learning implicit behavioral rules or preferences directly from data" and "effectively learning the 'knowledge' embedded in expert behavior." This is an implicit, data-only method, directly contradicting the stated inclusion boundary. The boundary must be revised or the entry removed or justified, otherwise the taxonomy's scope is ill-defined.
  2. [Section II-F and Table I] Section II-F classifies RuleFuser [65] under "Diffusion & Other Learning-Based Approaches" and describes it as "combining symbolic reasoning with data-driven predictors," but RuleFuser (as per the cited title) injects rules into evidential networks and is not a diffusion model. The same work also appears in Section II-D as a formal logic-based method and in Table I under Hybrid Symbolic & Neural Methods. Since the paper acknowledges multiple categorizations but never defines the primary-class assignment rule, the category counts and the claimed trend of increasing diffusion-based methods are not reliable.
  3. [Title, Abstract, Section I-A] The survey claims to be "comprehensive" but provides no systematic search protocol, inclusion/exclusion criteria, or coverage statistics (e.g., number of papers screened, databases, years). Without such methodology, the representativeness of the 68 referenced works and Table I cannot be assessed, and the "unified taxonomy" may reflect a convenience sample. This is a load-bearing gap for a survey whose central contribution is a structured map of the field.
  4. [Section II (first paragraph) and Section III] The taxonomy's stability is undermined by the survey's own concession that "in many cases, particularly in ML-based methods, this explicit prior knowledge does not become explicit; instead, in the overall pipeline, it becomes implicit or soft constraints." If explicit knowledge can become implicit, then the distinction between knowledge-integration methods and ordinary ML is not crisp, and the categorization of methods such as RL reward shaping (Section II-B) and IRL (Section II-E) requires a more principled criterion than the one given. The paper says "some works fall under multiple categories; we assign each a primary class but list cross-references" but never specifies the assignment rule, leaving the comparative analysis in Table I and the trends in Section III not fully supported.
minor comments (6)
  1. [Copyright notice on first page] The phrase "Perm ission" contains an erroneous space and should read "Permission."
  2. [Section I, first paragraph] The sentence "Machine learning (ML)- based approaches typically require large datasets but struggles in out-of-distribution scenarios" has a subject-verb agreement error: "struggles" should be "struggle."
  3. [Section II-F] The text refers to "Lupanov function-based reachability encoding," which appears to be a typo for "Lyapunov function-based," consistent with reference [67]'s title.
  4. [Reference [40]] The venue for DiLu is listed as "International Conference on Learning Research (ICLR)"; the correct name is "International Conference on Learning Representations."
  5. [Table I] The row "Diffusion & Other Learning-Based Approaches" lists only three works, one of which ([65]) is not a diffusion method; consider renaming the row to avoid overstating the diffusion category, or add more representative diffusion works.
  6. [Section III, first paragraph] The survey states it spans "classical methods (e.g., geometric models, optimization-based strategies)," but such classical methods are not discussed in the body of the survey; either add a brief discussion or remove the claim.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the survey is a descriptive taxonomy with no derivation chain that reduces to its own inputs.

full rationale

This is a survey paper and contains no formal derivation, fitted parameters, or predictive equations that could reduce to their own inputs. Its central claim is to deliver a unified taxonomy of knowledge integration methods; that taxonomy is an organizational summary of the cited literature rather than a consequence of a self-imposed definition. The only self-referential elements are three works authored by the survey authors ([31] TR2MTL, [60] SHIFT, and [68] legal compliance checking), but each is presented as one representative example inside a category, and neither the existence of any category nor the survey's overall structure depends on accepting those works' results. The paper itself acknowledges the softness of its classification boundary, stating that 'in many cases, particularly in ML-based methods, this explicit prior knowledge does not become explicit; instead, in the overall pipeline, it becomes implicit or soft constraints,' and that 'some works fall under multiple categories; we assign each a primary class but list cross-references.' These are scope and classification limitations, not circular reductions. No step in the survey's argument is shown to be equivalent, by construction or by self-citation, to the conclusion it supports.

Assumptions & free parameters 0 free parameters · 3 assumptions · 0 invented entities

The survey rests on assumptions about category exhaustiveness, the explicit/implicit knowledge boundary, and representativeness of selected works. No free parameters or invented entities apply because the paper is a literature review.

assumptions (3)
  • domain assumption The field can be exhaustively partitioned into the five categories in Table I (ontology, RL, LLM/RAG, formal logic, hybrid, diffusion).
    The survey's taxonomy depends on categories being comprehensive and mostly disjoint, but the text notes some works fall under multiple categories and assigns primary classes.
  • domain assumption Explicit rule knowledge can be reliably distinguished from implicit/soft constraints in ML pipelines.
    Section II introduction admits that in many ML-based methods explicit prior knowledge becomes implicit or soft constraints, undermining the clean boundary the taxonomy needs.
  • domain assumption The selected references are representative of the broader literature.
    No systematic search strategy or inclusion criteria are provided, so representativeness is assumed rather than demonstrated.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Knowledge Integration Strategies in Autonomous Vehicle Prediction and Planning: A Comprehensive Survey." pith.science (2026). https://pith.science/paper/FUB64L6T

@misc{pith2026250210477,
  author       = {Pith},
  title        = {Pith review of: Knowledge Integration Strategies in Autonomous Vehicle Prediction and Planning: A Comprehensive Survey},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/FUB64L6T}},
  note         = {Machine review of arXiv:2502.10477}
}
read the original abstract

This comprehensive survey examines the integration of knowledge-based approaches in autonomous driving systems, specifically focusing on trajectory prediction and planning. We extensively analyze various methodologies for incorporating domain knowledge, traffic rules, and commonsense reasoning into autonomous driving systems. The survey categorizes and analyzes approaches based on their knowledge representation and integration methods, ranging from purely symbolic to hybrid neuro-symbolic architectures. We examine recent developments in logic programming, foundation models for knowledge representation, reinforcement learning frameworks, and other emerging technologies incorporating domain knowledge. This work systematically reviews recent approaches, identifying key challenges, opportunities, and future research directions in knowledge-enhanced autonomous driving systems. Our analysis reveals emerging trends in the field, including the increasing importance of interpretable AI, the role of formal verification in safety-critical systems, and the potential of hybrid approaches that combine traditional knowledge representation with modern machine learning techniques.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

68 extracted references · 62 canonical work pages

  1. [63]

    Driving in Real Life with Inverse Reinforcement Learning

    T. Phan-Minh et al. , “Driving in real life with inverse reinforcement learning,” arXiv preprint arXiv:2206.03004 , 2022

    work page Pith review arXiv 2022

  2. [65]

    RuleFuser: An Evidential Bayes Approach for Rule Injection in Imitation Learned Planners and Predictors for Robustness under Distribution Shifts

    J. Patrikar et al. , “Rulefuser: Injecting rules in evidential networks for robust out-of-distribution trajectory prediction,” arXiv preprint arXiv:2405.11139, 2024

    work page Pith review arXiv 2024

  3. [1]

    A survey on trajectory-prediction methods for autonomous dr iving,

    Y . Huang, J. Du, Z. Y ang, Z. Zhou, L. Zhang, and H. Chen, “A survey on trajectory-prediction methods for autonomous dr iving,” IEEE Transactions on Intelligent V ehicles , 2022

    work page 2022

  4. [2]

    The integration of prediction and planning in deep learning automated driving systems: A review,

    S. Hagedorn et al. , “The integration of prediction and planning in deep learning automated driving systems: A review,” arXiv preprint arXiv:2308.05731, 2024

    arXiv 2024

  5. [3]

    End- to-end autonomous driving: Challenges and frontiers,

    L. Chen, P . Wu, K. Chitta, B. Jaeger, A. Geiger, and H. Li, “ End- to-end autonomous driving: Challenges and frontiers,” arXiv preprint arXiv:2306.16927, 2024

    arXiv 2024

  6. [4]

    Formal methods to comply with rules of the road in autonomous driving: State of the art and grand challe nges,

    N. Mehdipour et al. , “Formal methods to comply with rules of the road in autonomous driving: State of the art and grand challe nges,” Automatica, 2023

    work page 2023

  7. [5]

    Integrating knowledge graphs int o au- tonomous vehicle technologies: A survey of current state an d future directions,

    S. Htun and K. Fukuda, “Integrating knowledge graphs int o au- tonomous vehicle technologies: A survey of current state an d future directions,” Information, 2024

    work page 2024

  8. [6]

    Autonomous driving system: A comprehensive survey,

    J. Zhao et al., “Autonomous driving system: A comprehensive survey,” Expert Syst. Appl. , 2024

    work page 2024

Show all 68 references
  1. [7]

    A sce nario- adaptive driving behavior prediction approach to urban aut onomous driving,

    X. Geng, H. Liang, B. Y u, P . Zhao, L. He, and R. Huang, “A sce nario- adaptive driving behavior prediction approach to urban aut onomous driving,” Applied Sciences , vol. 7, no. 4, p. 426, Apr. 2017

    2017

  2. [8]

    Ontology-based decision making on uncontrolled intersec tions and narrow roads,

    L. Zhao, R. Ichise, T. Y oshikawa, T. Naito, T. Kakinami, a nd Y . Sasaki, “Ontology-based decision making on uncontrolled intersec tions and narrow roads,” in IEEE Intelligent V ehicles Symposium (IV) , 2015

    2015

  3. [9]

    Knowledge representa tion and planning for on-road driving,

    S. B. Balakirsky and C. Scrapper, “Knowledge representa tion and planning for on-road driving,” Robotics Auton. Syst. , vol. 49, 2004

    2004

  4. [10]

    SemanticFo rmer: Holistic and semantic traffic scene representation for trajectory pr ediction using knowledge graphs,

    Z. Sun, Z. Wang, L. Halilaj, and J. Luettin, “SemanticFo rmer: Holistic and semantic traffic scene representation for trajectory pr ediction using knowledge graphs,” IEEE Robotics and Automation Letters , 2024

    2024

  5. [11]

    Using ontology-based traffic models for mor e efficient decision making of autonomous vehicles,

    R. Regele, “Using ontology-based traffic models for mor e efficient decision making of autonomous vehicles,” in F ourth International Conference on Autonomic and Autonomous Systems (ICAS’08) , 2008

    2008

  6. [12]

    Bayesian-knowled ge driven ontologies: A framework for fusion of semantic knowledge un der uncertainty and incompleteness,

    E. S. Jr, J. Jurmain, and A. Ragazzi, “Bayesian-knowled ge driven ontologies: A framework for fusion of semantic knowledge un der uncertainty and incompleteness,” PLOS ONE , 2024

    2024

  7. [13]

    Ontology-based re asoning approach for long-term behavior prediction of road users,

    F. Fang, S. Y amaguchi, and A. Khiat, “Ontology-based re asoning approach for long-term behavior prediction of road users,” in 2019 IEEE Intelligent Transportation Systems Conference (ITSC ), 2019

    2019

  8. [14]

    Utilizing Background Knowledge for Robust Reason- ing over Traffic Situations,

    J. Zhang et al., “Utilizing Background Knowledge for Robust Reason- ing over Traffic Situations,” in W orkshop on Knowledge-Augmented Methods for Natural Language Processing , 2022

    2022

  9. [15]

    Ontology-ba sed driving scene modeling, situation assessment and decision making for autonomous vehicles,

    L. Huang, H. Liang, B. Y u, B. Li, and H. Zhu, “Ontology-ba sed driving scene modeling, situation assessment and decision making for autonomous vehicles,” in 2019 4th Asia-Pacific Conference on Intelligent Robot Systems (ACIRS) , 2019

    2019

  10. [16]

    Using ontologies for the formalization and recognition of criticality for automated driving,

    L. Westhofen, C. Neurohr, M. Butz, M. Scholtes, and M. Sc huldes, “Using ontologies for the formalization and recognition of criticality for automated driving,” IEEE Open Journal of Intelligent Transporta- tion Systems , vol. 3, 2022

    2022

  11. [17]

    Ontology based knowledge representati on tech- nique, domain modeling languages and planners for robotic p ath planning: A survey,

    G. R and V . Uma, “Ontology based knowledge representati on tech- nique, domain modeling languages and planners for robotic p ath planning: A survey,” ICT Express , vol. 4, p. 69–74, 2018

    2018

  12. [18]

    An ontology for sustainable intelligent transportation systems,

    A. Giret et al. , “An ontology for sustainable intelligent transportation systems,” 2018

    2018

  13. [19]

    Ontology development for sustainable intelligent transportation systems,

    B. P . Bhuyan et al., “Ontology development for sustainable intelligent transportation systems,” in 2024 IEEE 29th Asia Pacific Conference on Communications (APCC) , 2024

    2024

  14. [20]

    Toward s the automation of knowledge graph construction using large lan guage models

    V . K. Kommineni, B. K¨ onig-Ries, and S. Samuel, “Toward s the automation of knowledge graph construction using large lan guage models.”

  15. [21]

    Graph-based autonomous driving with traffic- rule-enhanced curriculum learning,

    L. F. Peiss et al. , “Graph-based autonomous driving with traffic- rule-enhanced curriculum learning,” in 2023 IEEE 26th International Conference on Intelligent Transportation Systems (ITSC) , 2023

    2023

  16. [22]

    Road traffic law adaptive decision-making for sel f-driving vehicles,

    J. Lin, W. Zhou, H. Wang, Z. Cao, W. Y u, C. Zhao, D. Zhao, D. Y ang, and J. Li, “Road traffic law adaptive decision-making for sel f-driving vehicles,” in 2022 IEEE 25th International Conference on Intelligent Transportation Systems (ITSC) , 2022

    2022

  17. [23]

    Evolutionary decision-making and planning for autonomous driving: A hybrid augmented intelligence frame work,

    K. Y uan et al. , “Evolutionary decision-making and planning for autonomous driving: A hybrid augmented intelligence frame work,” IEEE Transactions on Intelligent Transportation Systems , 2024

    2024

  18. [24]

    Let hybrid a* path plan ner obey traf- fic rules: A deep reinforcement learning-based planning fra mework,

    X. Li, S. Patel, and C. B¨ uskens, “Let hybrid a* path plan ner obey traf- fic rules: A deep reinforcement learning-based planning fra mework,” ArXiv, vol. abs/2407.01216, 2024

    2024 arXiv

  19. [25]

    Personalized decision-making and control for auto- mated vehicles based on generative adversarial imitation l earning,

    X. Tang et al. , “Personalized decision-making and control for auto- mated vehicles based on generative adversarial imitation l earning,” in IEEE International Conference on Intelligent Transportat ion Systems , 2023

    2023

  20. [26]

    Virtual to real reinfo rcement learning for autonomous driving,

    X. Pan, Y . Y ou, Z. Wang, and C. Lu, “Virtual to real reinfo rcement learning for autonomous driving,” in Procedings of the British Machine Vision Conference 2017 . British Machine Vision Association, 2017

    2017

  21. [27]

    Deep reinforcement learning for autonomous driving: A survey,

    B. R. Kiran et al. , “Deep reinforcement learning for autonomous driving: A survey,” IEEE Transactions on Intelligent Transportation Systems, 2022

    2022

  22. [28]

    Robust autonomy emerges from self- play,

    M. Cusumano-Towner et al. , “Robust autonomy emerges from self- play,” arXiv preprint arXiv:2502.03349 , 2025

    2025 arXiv

  23. [29]

    Driving everywhere with large language model policy adaptation,

    B. Li et al. , “Driving everywhere with large language model policy adaptation,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , 2024

    2024

  24. [30]

    Driving with regulation: Interpretable decision-making for auton omous vehi- cles with retrieval-augmented reasoning via LLM,

    T. Cai, Y . Liu, Z. Zhou, H. Ma, S. Z. Zhao, Z. Wu, and J. Ma, “ Driving with regulation: Interpretable decision-making for auton omous vehi- cles with retrieval-augmented reasoning via LLM,” Oct. 202 4

  25. [31]

    TR2MTL: llm b ased framework for metric temporal logic formalization of traffi c rules,

    K. Manas, S. Zwicklbauer, and A. Paschke, “TR2MTL: llm b ased framework for metric temporal logic formalization of traffi c rules,” in 2024 IEEE Intelligent V ehicles Symposium (IV) , 2024

    2024

  26. [32]

    Drivegpt4: Interpretable end-to-end autonomous drivin g via large language model,

    Z. Xu et al., “Drivegpt4: Interpretable end-to-end autonomous drivin g via large language model,” IEEE Robotics and Automation Letters , 2024

    2024

  27. [33]

    A language agent for autonomous driving,

    J. Mao, J. Y e, Y . Qian, M. Pavone, and Y . Wang, “A language agent for autonomous driving,” Conference on Language Modeling , vol. abs/2311.10813, 2024

    2024 arXiv

  28. [34]

    EMMA: End-to-end multimodal model for au- tonomous driving,

    J.-J. Hwang et al. , “EMMA: End-to-end multimodal model for au- tonomous driving,” 2024

    2024

  29. [35]

    DriveLLM: Charting the path toward full autonomous driving with large language models,

    Y . Cui et al. , “DriveLLM: Charting the path toward full autonomous driving with large language models,” IEEE Transactions on Intelligent V ehicles, 2024

    2024

  30. [36]

    LanguageMPC: Large language models as decision makers for autonomous driving,

    H. Sha et al. , “LanguageMPC: Large language models as decision makers for autonomous driving,” 2023

    2023

  31. [37]

    Drive like a human: Rethinking autonomous driving with large language models,

    D. Fu et al. , “Drive like a human: Rethinking autonomous driving with large language models,” in 2024 IEEE/CVF Winter Conference on Applications of Computer Vision W orkshops (WACVW) , 2024

    2024

  32. [38]

    RAG-Driver: Generalisable driving explanations with retrieval-augmented in-context learning in multi-modal l arge language model,

    J. Y uan et al. , “RAG-Driver: Generalisable driving explanations with retrieval-augmented in-context learning in multi-modal l arge language model,” Robotics: Science and Systems (RSS) , 2024

    2024

  33. [39]

    Llm-assist: Enhancing closed-loop planning with language-based reasoning,

    S. Sharan, F. Pittaluga, M. Chandraker et al. , “Llm-assist: Enhancing closed-loop planning with language-based reasoning,” arXiv preprint arXiv:2401.00125, 2023

    2023 arXiv

  34. [40]

    Dilu: A knowledge-driven approach to autonomous driving with large language models,

    L. Wen et al. , “Dilu: A knowledge-driven approach to autonomous driving with large language models,” in International Conference on Learning Research (ICLR) , 2024

    2024

  35. [41]

    Formalizing traffic rule s for uncon- trolled intersections,

    A. Karimi and P . S. Duggirala, “Formalizing traffic rule s for uncon- trolled intersections,” in 2020 ACM/IEEE 11th International Confer- ence on Cyber-Physical Systems (ICCPS) , 2020

    2020

  36. [42]

    AUTO-DISCERN: Autonomous Driving Using Common Sense Reasoning,

    S. Kothawade et al. , “AUTO-DISCERN: Autonomous Driving Using Common Sense Reasoning,” in W orkshop on Goal-directed Execution of Answer Set Programs , 2021

    2021

  37. [43]

    Assessing drivers situa tion awareness in semi-autonomous vehicles: Asp based characterisations of driving dynamics for modeling scene interpretation and projection ,

    J. Suchan and J.-P . Osterloh, “Assessing drivers situa tion awareness in semi-autonomous vehicles: Asp based characterisations of driving dynamics for modeling scene interpretation and projection ,” in 39th International Conference on Logic Programming (ICLP 2023) , 2023

    2023

  38. [44]

    Using reachable s ets for trajectory planning of automated vehicles,

    S. Manzinger, C. Pek, and M. Althoff, “Using reachable s ets for trajectory planning of automated vehicles,” IEEE Transactions on Intelligent V ehicles, vol. 6, pp. 232–248, 2021

    2021

  39. [45]

    Set-based prediction of traf fic partici- pants considering occlusions and traffic rules,

    M. Koschi and M. Althoff, “Set-based prediction of traf fic partici- pants considering occlusions and traffic rules,” IEEE Transactions on Intelligent V ehicles, vol. 6, pp. 249–265, 2021

    2021

  40. [46]

    Adaptive cruise c ontrol: Hybrid, distributed, and now formally verified,

    S. M. Loos, A. Platzer, and L. Nistor, “Adaptive cruise c ontrol: Hybrid, distributed, and now formally verified,” in FM 2011: F ormal Methods, M. Butler and W. Schulte, Eds. Springer, 2011

    2011

  41. [47]

    Bhuiyan, G

    H. Bhuiyan, G. Governatori, A. Bond, S. Demmel, M. Badiu l Islam, and A. Rakotonirainy, Traffic Rules Encoding Using Defeasible De- ontic Logic . IOS Press, 2020

    2020

  42. [48]

    Formalise r egulations for autonomous vehicles with right-open temporal deontic d efeasible logic,

    P . Y . Chan, X. Li, Y . Lu, Y . Lin, and A. Bundy, “Formalise r egulations for autonomous vehicles with right-open temporal deontic d efeasible logic,” in Artificial Intelligence XLI , 2025

    2025

  43. [49]

    Formalizing traffic rules for machine interpretability,

    K. Esterle, L. Gressenbuch, and A. Knoll, “Formalizing traffic rules for machine interpretability,” in 2020 IEEE 3rd Connected and Automated V ehicles Symposium (CA VS), 2020

    2020

  44. [50]

    Traffic netw ork control from temporal logic specifications,

    S. Coogan, E. A. Gol, M. Arcak, and C. Belta, “Traffic netw ork control from temporal logic specifications,” IEEE Transactions on Control of Network Systems , 2016

    2016

  45. [51]

    Model predictive control with signal temporal logic specifications,

    V . Raman et al. , “Model predictive control with signal temporal logic specifications,” in 53rd IEEE Conference on Decision and Control , 2014

    2014

  46. [52]

    Model predictive control o f urban traffic networks with temporal logic constraints,

    S. Sadraddini and C. Belta, “Model predictive control o f urban traffic networks with temporal logic constraints,” in 2016 American Control Conference (ACC), 2016

    2016

  47. [53]

    Model predictive robu stness of signal temporal logic predicates,

    Y .-C. Lin, H. Li, and M. Althoff, “Model predictive robu stness of signal temporal logic predicates,” IEEE Robotics and Automation Letters, 2022

    2022

  48. [54]

    Injecting knowledg e in data-driven vehicle trajectory predictors,

    M. Bahari, I. Nejjar, and A. Alahi, “Injecting knowledg e in data-driven vehicle trajectory predictors,” p. 103010, Jul. 2021

    2021

  49. [55]

    Differentiable logic layer for rule guided trajectory prediction

    X. Li et al. , “Differentiable logic layer for rule guided trajectory prediction.” PMLR, 2021

    2021

  50. [56]

    Towards knowledge-driven autonomous driving,

    X. Li, Y . Bai, P . Cai, L. Wen, D. Fu, B. Zhang, X. Y ang, X. Ca i, T. Ma, J. Guo et al., “Towards knowledge-driven autonomous driving,” arXiv preprint arXiv:2312.04316, 2023

    2023 arXiv

  51. [57]

    A s urvey on hybrid motion planning methods for automated driving sys tems,

    M. A. Sormoli, K. Koufos, M. Dianati, and R. Woodman, “A s urvey on hybrid motion planning methods for automated driving sys tems,” arXiv preprint arXiv:2406.05575 , 2024

    2024 arXiv

  52. [58]

    Logic tensor networks,

    S. Badreddine, A. d. Garcez, L. Serafini, and M. Spranger , “Logic tensor networks,” Artificial Intelligence , vol. 303, p. 103649, 2022

    2022

  53. [59]

    Deepproblog: Neural probabilistic logic program- ming,

    R. Manhaeve et al., “Deepproblog: Neural probabilistic logic program- ming,” Advances in neural information processing systems , 2018

    2018

  54. [60]

    Uncertainty-aware trajectory prediction via rule- regularized heteroscedastic deep classification,

    K. Manas et al. , “Uncertainty-aware trajectory prediction via rule- regularized heteroscedastic deep classification,” in Proceedings of the Robotics: Science and Systems (RSS) , 2025

    2025

  55. [61]

    Cautious pla nning with incremental symbolic perception: Designing verified react ive driving maneuvers,

    D. Kamale, S. Haesaert, and C.-I. V asile, “Cautious pla nning with incremental symbolic perception: Designing verified react ive driving maneuvers,” in IEEE International Conference on Robotics and Au- tomation (ICRA) , 2023

    2023

  56. [62]

    V ehicle trajectory prediction using generative adversa rial network with temporal logic syntax tree features,

    X. Li et al., “V ehicle trajectory prediction using generative adversa rial network with temporal logic syntax tree features,” IEEE Robotics and Automation Letters, 2021

    2021

  57. [64]

    Safetynet: Safe planning for real-world self-driving vehicles using machine-learned policies,

    M. Vitelli et al. , “Safetynet: Safe planning for real-world self-driving vehicles using machine-learned policies,” in 2022 International Con- ference on Robotics and Automation (ICRA) , 2022

    2022

  58. [66]

    Diverse controllable diffusion pol icy with signal temporal logic,

    Y . Meng and C. Fan, “Diverse controllable diffusion pol icy with signal temporal logic,” IEEE Robotics and Automation Letters , 2024

    2024

  59. [67]

    Cobl-diffusion: Diffusion-ba sed conditional robot planning in dynamic environments using control barri er and lya- punov functions,

    K. Mizuta and K. Leung, “Cobl-diffusion: Diffusion-ba sed conditional robot planning in dynamic environments using control barri er and lya- punov functions,” in IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) , 2024

    2024

  60. [68]

    Legal Compliance Checking of A u- tonomous Driving with Formalized Traffic Rule Exceptions,

    K. Manas and A. Paschke, “Legal Compliance Checking of A u- tonomous Driving with Formalized Traffic Rule Exceptions,” in W ork- shop on Logic Programming and Legal Reasoning (LPLR) , 2023

    2023

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.