Pith. sign in

REVIEW 4 major objections 4 minor 33 references

BT-TL-DMPs: A Novel Robot TAMP Framework Combining Behavior Tree, Temporal Logic and Dynamical Movement Primitives

T0 review · 4 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 Formal task logic lets robots reuse taught motions in new scenes.

desk verdict Useful modular idea, but Eq. (22) as printed minimizes STL robustness instead of maximizing it, so the central low-level claim is unsupported as written. read the letter →

arxiv 2507.14582 v1 pith:7WSBZ566 submitted 2025-07-19 cs.RO

Zezhi Liu , Shizhen Wu , Hanqian Luo , Deyun Qin , Yongchun Fang This is my paper
classification cs.RO
keywords taskandmotionplanninglearningfromdemonstrationdynamicmovementprimitivessignaltemporallogicbehaviortreeslong-horizonmanipulationformalmethodsinrobotics
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

BT-TL-DMPs is a hierarchical framework for making learned manipulation skills reusable in new scenes. It takes a Signal Temporal Logic (STL) specification of a long-horizon task and splits it into two uses: the temporal structure is converted into a Behavior Tree that decides which learned sub-skill runs when, and the remaining constraints are fed into an optimization that adjusts the forcing term of each Dynamic Movement Primitive. The paper's central claim is that this forcing-term optimization, rather than trajectory matching, lets the robot satisfy spatiotemporal requirements such as via-points, obstacle avoidance, and velocity limits while preserving the dynamics seen in demonstrations. The paper supports the claim with simulations of four constraint types and with two real-world long-horizon tasks, breakfast preparation and afternoon-tea preparation.

What carries the argument

The load-bearing object is the optimized DMP forcing term $F^\epsilon_{\mathrm{opt}}$, obtained from Eq. (22) by combining STL robustness with a weighted forcing-term fidelity objective. The weighting matrix $W$, built from the variance of multiple demonstrations, emphasizes low-variability parts of the motion. The second mechanism is the STL-to-LTL-to-BT pipeline: temporal operators in the task formula become ordering constraints, and each atomic manipulation is wrapped as a BT node with pre- and post-conditions, so the tree emits the switching signal $\epsilon(s)$ that selects which DMP runs. Together they propagate task-level spatiotemporal requirements down to the continuous controller.

What would settle it

Take a long-horizon task whose STL formula contains a nested temporal operator, such as $\square_{[0,T]}(\lozenge_{[a,b]}\varphi)$ or an until over an interval, and execute the generated Behavior Tree while the environment changes mid-task. If the tree ever orders sub-tasks in a way that violates the original formula, or if the executed trajectory has negative STL robustness despite the optimizer reporting a positive margin, the central claim is refuted.

Watch

Extended reading notes

Core claim

The central discovery claimed is that the forcing term of a DMP is the right object to optimize when a learned skill must respect a temporal-logic constraint. With $F^\epsilon_{\mathrm{opt}} = \arg\min_{F^\epsilon} \lambda_1 J_{\mathrm{STL}}(F^\epsilon) + \lambda_2 J_{\mathrm{DMP}}(F^\epsilon)$, where $J_{\mathrm{STL}}$ is the STL robustness $\rho(\varphi,p,0)$ of the induced trajectory and $J_{\mathrm{DMP}} = \| W F^\epsilon - W F^\epsilon_{\mathrm{lrn}}\|^2$ is a variance-weighted distance from the demonstrated forcing term, the resulting motions satisfy the formula while keeping the qualitative character of the demonstration. At the task level the paper claims that the same STL description, once abstracted to LTL, can be turned into a Behavior Tree that supplies the sub-task switching signal, so the formal specification drives both levels of the hierarchy.

Load-bearing premise

The framework's task-level guarantees depend on the assumption that the generated Behavior Tree preserves the meaning of the original temporal-logic task description; the paper shows a manual example but no proof that the conversion is correct in general.

Editorial extensions

If this is right

  • A skill learned once can be redeployed in a new layout by re-optimizing its forcing term under the new STL constraints, without collecting new demonstrations.
  • A single formal specification can drive both the ordering of sub-tasks and the continuous motion, so high-level correctness and low-level constraint satisfaction no longer need separate hand-coded bridges.
  • Because the optimization acts on forcing terms rather than on raw trajectories, DMP convergence to the goal and the temporal profile of the motion are preserved when constraints change.
  • In the reported experiments, reactive Behavior Trees let the robot monitor task progress, retry failed steps, and still meet spatiotemporal constraints in long-horizon tasks such as breakfast and afternoon-tea preparation.
  • The proposed objective can be combined with any STL fragment whose robustness is differentiable, so additional constraint types beyond the four simulated ones could be added without changing the framework.

Reading between the lines

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

  • An implication the paper leaves implicit is that the forcing-term formulation is not tied to DMPs in principle; the same variance-weighted fidelity objective could be applied to other parameterized movement generators, so the low-level mechanism may transfer beyond dynamic movement primitives.
  • The variance-based weighting suggests a measurable prediction: high-variance segments of a demonstrated skill should be the ones that change most under optimization. A per-segment comparison of optimized versus demonstrated forcing terms under increasing constraint weight would test this directly.
  • If the STL-to-BT conversion were paired with a natural-language-to-STL translator, the framework would become an end-to-end pipeline from instruction to constrained execution; the paper leaves that integration to future work.
  • Task-level soundness would become checkable if the conversion were accompanied by a theorem stating which fragment of STL is preserved by the generated Behavior Tree; without such a theorem, the high-level guarantee is only illustrated by example.
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 / 4 minor

Summary. The paper proposes BT-TL-DMPs, a hierarchical task and motion planning framework for long-horizon robotic manipulation. At the task level, Signal Temporal Logic (STL) specifications are abstracted into Linear Temporal Logic (LTL) formulas and used to construct Behavior Trees (BTs) for high-level action sequencing. At the motion level, the forcing terms of learned Dynamic Movement Primitives (DMPs) are optimized so that the generated trajectories satisfy the STL constraints while remaining close to the demonstrated forcing terms. The framework is evaluated in simulations with several STL-constrained scenarios and in two real-world long-horizon tabletop manipulation tasks.

Significance. If the formal link from STL to BTs were rigorous and the DMP optimization problem were correctly formulated, the framework would offer a useful interface between temporal-logic task planning and learning-from-demonstration skill generalization. The paper addresses a relevant problem, uses the public PyTeLo toolbox, and provides qualitative real-world demonstrations. However, the central optimization is written with the wrong sign: it minimizes STL robustness rather than maximizing it, which drives trajectories away from the satisfying set. In addition, the STL-to-LTL-to-BT transformation is only illustrated by a manual example and is not shown to preserve task semantics, and the experiments report no quantitative robustness values. The contributions as stated are therefore not yet supported.

major comments (4)
  1. [IV-C, Eqs. (20)-(22)] The optimization in Eq. (22) is F_opt = argmin_F λ1 J_STL(F) + λ2 J_DMP(F) with λ1, λ2 in R_+, and J_STL(F) = ρ(φ, p, 0). Under the standard STL semantics used in Section II-C and Table I, a positive robustness value means satisfaction and larger values mean larger satisfaction margins. Minimizing ρ therefore drives the trajectory away from the satisfying set, toward ρ ≤ 0, i.e., toward violation. No hard constraint ρ ≥ 0 or penalty for violation appears in Eq. (22). The formulation should minimize −ρ, equivalently maximize ρ, or include a constraint that forces the robustness to be positive. As written, the central low-level claim that the optimized DMPs satisfy the STL constraints is not supported.
  2. [IV-B, Eqs. (12)-(13)] The transformation from the STL task description to LTL and then to a Behavior Tree is presented only on a single manual example. No theorem or algorithm is stated showing that the generated BT preserves the satisfaction semantics of the original STL specification, despite Contribution 1 claiming soundness of high-level decisions. The LTL formulas in Eqs. (12) and (13) are written in a non-standard way (e.g., U(F(ψ_Pick, U(Fψ_move, Fψ_place)))) and the abstraction from continuous STL predicates to Boolean atomic propositions is not formally defined. This gap must be addressed, either by stating and proving a preservation result or by clearly circumscribing the claim to the given examples.
  3. [IV-C, Eq. (17)] The paper claims that optimizing the forcing term while minimizing its deviation from the learned forcing term preserves the convergence and stability properties of the original DMP. No stability proof is provided. The optimized forcing term F_opt is not constrained to vanish as the phase variable tends to zero, and the objective in Eq. (22) does not enforce the contraction properties of the underlying spring-damper system. A proof, or an added constraint guaranteeing convergence to the goal, is needed before the claim of preserving demonstrated dynamics can be sustained.
  4. [V and VI] The simulation and real-world experiments report only qualitative trajectories and state that all specified STL constraints are satisfied, but no robustness values or quantitative satisfaction margins are reported, and no statistics over repeated runs are given. Since the optimized trajectory is obtained from an objective that contains the STL robustness term, demonstrating that the resulting trajectory looks compliant is to a large degree a restatement of the optimization. Reporting ρ values for the optimized and unoptimized motions, along with success rates and parameter settings, is necessary to validate the generalization and constraint-satisfaction claims.
minor comments (4)
  1. [Table I] In the quantitative semantics for conjunction and disjunction, the second argument is repeated as ρ(φ1, ξ, t); it should be ρ(φ2, ξ, t) in both rows.
  2. [Eq. (12)] The LTL notation U(F(...), U(...)) is non-standard and the derivation from Eq. (9) is not explained. Please define the exact syntax and semantics used, or rewrite the formulas in a conventional LTL form.
  3. [Eq. (11)] The Action_STL predicate requires the cup position to follow p_DMP(t), which is the trajectory being generated by the low-level optimization. This makes the specification self-referential; please clarify whether this is intended as a task constraint or an assumption about the grasp.
  4. [Section V] The experimental descriptions do not report the values of λ1 and λ2, the DMP basis-function parameters, the number of demonstrations, or the PyTeLo configuration, which makes the results difficult to reproduce.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: the STL-DMP optimization is an input-driven objective, not a self-referential prediction.

full rationale

The paper's central chain is: given an STL task specification, generate a BT for high-level sequencing, and optimize DMP forcing terms so the motion satisfies the low-level STL fragment C_STL while staying close to the learned forcing term. I find no step where an output is defined in terms of the quantity it is said to predict. Eq. (22) minimizes lambda1 J_STL(F) + lambda2 J_DMP(F), where J_STL is the STL robustness of the trajectory generated by F and J_DMP is the forcing-term distance to the learned term. Claiming that the resulting trajectories 'satisfy' the constraints is an evaluation of the optimizer, not a prediction of an independently measured quantity; the constraints are inputs, not fitted parameters. The Action_STL predicate in Eq. (11) is a modeling convention (the cup follows the DMP trajectory), and Section IV-C explicitly routes only C_STL to the motion optimizer, so the low-level optimization is not vacuous. The high-level conversion from STL to LTL to BT is asserted by appeal to external works [32,33] and lacks a semantics-preservation proof; this is a soundness gap, not circularity, because the paper does not define soundness in terms of its own output. Self-citations [25,31] appear only as related work and are not load-bearing. The known sign issue in Eq. (22) (minimizing positive robustness rather than maximizing it) is a correctness defect, not a circular reduction.

Assumptions & free parameters 6 free parameters · 4 assumptions · 0 invented entities

The framework relies on an assumed STL specification, a sound abstraction from STL to LTL and BT, and stability of the optimized DMP; none of these are proved in the paper. The free parameters, including regularization weights, DMP basis settings, STL thresholds, and variance weighting, are hand-chosen and unreported.

free parameters (6)
  • lambda_1 and lambda_2
    Regularization weights in Eq. (22) balance STL satisfaction against preservation of the demonstrated forcing term; no values are given in the paper.
  • DMP basis function centers and widths
    The centers c_i and widths h_i in Eq. (3) are set in advance but no values or selection procedure are reported.
  • GMM-GMR hyperparameters
    Number of Gaussian components and covariance regularization used to compute the mean and variance of demonstrations are not specified.
  • Variance weighting formulas
    The exponential weights in Appendix A are introduced ad hoc to avoid zero variance at start and goal; the functional form is not derived or justified.
  • STL time intervals and thresholds
    Intervals such as [t0,t1], [t1,t2] and thresholds such as 0.01, 0.1 m are chosen per scenario without a tuning rule.
  • DMP time constant and gains
    tau, alpha, beta in Eq. (1) are standard but values are not reported; results depend on them.
assumptions (4)
  • domain assumption The given STL specification Psi_STL correctly describes the long-horizon task.
    Section IV-A states it is assumed the STL formula can describe the task correctly; the paper does not verify the specification against the actual environment.
  • ad hoc to paper The STL-to-LTL abstraction and the subsequent BT construction from [32] are sound and preserve task satisfaction.
    Section IV-B invokes [32] and [33] without proving that the abstracted LTL formula and generated BT correctly realize the original continuous STL task.
  • ad hoc to paper The optimized DMP retains the convergence and stability properties of the original DMP.
    No stability analysis is given for the DMP with the optimized forcing term F_opt; the paper assumes forcing-term closeness preserves convergence.
  • domain assumption Demonstrations are adequately captured and GMM-GMR yields a reliable mean and variance for each skill.
    The paper relies on Qualisys motion capture and GMM-GMR without reporting validation of the learned models.

how reviews work

0 comments
Cite this review

Pith. "Pith review of BT-TL-DMPs: A Novel Robot TAMP Framework Combining Behavior Tree, Temporal Logic and Dynamical Movement Primitives." pith.science (2026). https://pith.science/paper/7WSBZ566

@misc{pith2026250714582,
  author       = {Pith},
  title        = {Pith review of: BT-TL-DMPs: A Novel Robot TAMP Framework Combining Behavior Tree, Temporal Logic and Dynamical Movement Primitives},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/7WSBZ566}},
  note         = {Machine review of arXiv:2507.14582}
}
read the original abstract

In the field of Learning from Demonstration (LfD), enabling robots to generalize learned manipulation skills to novel scenarios for long-horizon tasks remains challenging. Specifically, it is still difficult for robots to adapt the learned skills to new environments with different task and motion requirements, especially in long-horizon, multi-stage scenarios with intricate constraints. This paper proposes a novel hierarchical framework, called BT-TL-DMPs, that integrates Behavior Tree (BT), Temporal Logic (TL), and Dynamical Movement Primitives (DMPs) to address this problem. Within this framework, Signal Temporal Logic (STL) is employed to formally specify complex, long-horizon task requirements and constraints. These STL specifications are systematically transformed to generate reactive and modular BTs for high-level decision-making task structure. An STL-constrained DMP optimization method is proposed to optimize the DMP forcing term, allowing the learned motion primitives to adapt flexibly while satisfying intricate spatiotemporal requirements and, crucially, preserving the essential dynamics learned from demonstrations. The framework is validated through simulations demonstrating generalization capabilities under various STL constraints and real-world experiments on several long-horizon robotic manipulation tasks. The results demonstrate that the proposed framework effectively bridges the symbolic-motion gap, enabling more reliable and generalizable autonomous manipulation for complex robotic tasks.

Figures

Figures reproduced from arXiv: 2507.14582 by the authors.

Figure 1
Figure 1. The proposed robot TAMP framework which is combined with BT, TL and DMPs. [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. The utilization of LTL and STL to generate BT (in task [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. The overall task and motion level planning framework. [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Motion planning framework: the proposed method gen [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: Comparison of DMP simulations: (a) generation with new start/goal positions, (b) obstacle avoidance via trajectory [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 6
Figure 6. Figure 6: Different scenarios for STL-DMP generalization performance: first row shows via-point, obstacle avoidance, space-limit, [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]
Figure 7
Figure 7. Figure 7: Experiment of breakfast preparation task. [PITH_FULL_IMAGE:figures/full_fig_p008_7.png]
Figure 8
Figure 8. Figure 8: Experiment of afternoon tea preparation task. [PITH_FULL_IMAGE:figures/full_fig_p009_8.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

33 extracted references · 23 canonical work pages

  1. [8]

    Reactive temporal logic-based planning and control for interactive robotic tasks,

    F. Nawaz, S. Peng, L. Lindemann, N. Figueroa, and N. Matni, “Reactive temporal logic-based planning and control for interactive robotic tasks,” inIEEE/RSJ Inter- national Conference on Intelligent Robots and Systems. IEEE, 2024, pp. 12 108–12 115

    work page 2024

  2. [32]

    Designing Behavior Trees from Goal-Oriented LTLf Formulas

    A. Neupane, E. G. Mercer, and M. A. Goodrich, “De- signing behavior trees from goal-oriented ltlf formulas,” arXiv preprint arXiv:2307.06399, 2023

    work page Pith review arXiv 2023

  3. [33]

    Spatial: monitoring and planning of robotic tasks using spatio-temporal logic specifications,

    C. Pek, G. F. Schuppe, F. Esposito, J. Tumova, and D. Kragic, “Spatial: monitoring and planning of robotic tasks using spatio-temporal logic specifications,”Au- tonomous Robots, vol. 47, no. 8, pp. 1439–1462, 2023. 11 APPENDIX A. Construction for the Weight of Optimization Function in Generalizing Multi-demon-learned DMPs As described by equation (18) in ...

    work page 2023

  4. [1]

    Integrated task and motion planning,

    C. R. Garrett, R. Chitnis, R. Holladay, B. Kim, T. Sil- ver, L. P. Kaelbling, and T. Lozano-P ´erez, “Integrated task and motion planning,”Annual Review of Control, Robotics, and Autonomous Systems, vol. 4, no. 1, pp. 265–293, 2021

    work page 2021

  5. [2]

    Recent trends in task and motion planning for robotics: A survey,

    H. Guo, F. Wu, Y . Qin, R. Li, K. Li, and K. Li, “Recent trends in task and motion planning for robotics: A survey,”ACM Computing Surveys, vol. 55, no. 13s, pp. 1–36, 2023

    2023

  6. [3]

    A survey of optimization-based task and motion planning: From classical to learning approaches,

    Z. Zhao, S. Cheng, Y . Ding, Z. Zhou, S. Zhang, D. Xu, and Y . Zhao, “A survey of optimization-based task and motion planning: From classical to learning approaches,” IEEE/ASME Transactions on Mechatronics, pp. 1–27, 2024

    work page 2024

  7. [4]

    Recent advances in robot learning from demonstration,

    H. Ravichandar, A. S. Polydoros, S. Chernova, and A. Billard, “Recent advances in robot learning from demonstration,”Annual Review of Control, Robotics, and Autonomous Systems, vol. 3, pp. 297–330, 2020

    2020

  8. [5]

    A constrained dmps framework for robot skills learning and generalization from human demonstrations,

    Z. Lu, N. Wang, and C. Yang, “A constrained dmps framework for robot skills learning and generalization from human demonstrations,”IEEE/ASME Transactions on Mechatronics, vol. 26, no. 6, pp. 3265–3275, 2021. 10

    work page 2021

Show all 33 references
  1. [6]

    Incremental learning robot task representation and identification,

    X. Zhou, H. Wu, J. Rojas, Z. Xu, S. Li, X. Zhou, H. Wu, J. Rojas, Z. Xu, and S. Li, “Incremental learning robot task representation and identification,”Nonparametric Bayesian Learning for Collaborative Robot Multimodal Introspection, pp. 29–49, 2020

    2020

  2. [7]

    Rule-based safe probabilistic movement primitive control via control barrier functions,

    M. Davoodi, A. Iqbal, J. M. Cloud, W. J. Beksi, and N. R. Gans, “Rule-based safe probabilistic movement primitive control via control barrier functions,”IEEE Transactions on Automation Science and Engineering, vol. 20, no. 3, pp. 1500–1514, 2022

    2022

  3. [9]

    Colledanchise and P

    M. Colledanchise and P. ¨Ogren,Behavior trees in robotics and AI: An introduction. CRC Press, 2018

    2018

  4. [10]

    Behavior trees in robot control systems,

    P. ¨Ogren and C. I. Sprague, “Behavior trees in robot control systems,”Annual Review of Control, Robotics, and Autonomous Systems, vol. 5, no. 1, pp. 81–107, 2022

    2022

  5. [11]

    Comparison between behavior trees and finite state machines,

    M. Iovino, J. F ¨orster, P. Falco, J. J. Chung, R. Siegwart, and C. Smith, “Comparison between behavior trees and finite state machines,”arXiv preprint arXiv:2405.16137, 2024

    2024 arXiv

  6. [12]

    Adaptive manipulation using behavior trees,

    J. Cloete, W. Merkt, and I. Havoutis, “Adaptive manipulation using behavior trees,”arXiv preprint arXiv:2406.14634, 2024

    2024 arXiv

  7. [13]

    Backpropa- gation through signal temporal logic specifications: In- fusing logical structure into gradient-based methods,

    K. Leung, N. Ar ´echiga, and M. Pavone, “Backpropa- gation through signal temporal logic specifications: In- fusing logical structure into gradient-based methods,” The International Journal of Robotics Research, vol. 42, no. 6, pp. 356–370, 2023

    2023

  8. [14]

    Nl2tl: Trans- forming natural languages to temporal logics using large language models,

    Y . Chen, R. Gandhi, Y . Zhang, and C. Fan, “Nl2tl: Trans- forming natural languages to temporal logics using large language models,”arXiv preprint arXiv:2305.07766, 2023

    2023 arXiv

  9. [15]

    Chatstl: A framework of translation from nat- ural language to signal temporal logic specifications for autonomous vehicle navigation out of blocked scenarios,

    Y . Wang, Z. Huang, S. Dong, H. Chu, X. Yin, and B. Gao, “Chatstl: A framework of translation from nat- ural language to signal temporal logic specifications for autonomous vehicle navigation out of blocked scenarios,” inInternational Conference on Computer and Automa- tion En...

    2024

  10. [16]

    Formal synthesis of con- trollers for safety-critical autonomous systems: Devel- opments and challenges,

    X. Yin, B. Gao, and X. Yu, “Formal synthesis of con- trollers for safety-critical autonomous systems: Devel- opments and challenges,”Annual Reviews in Control, vol. 57, p. 100940, 2024

    2024

  11. [17]

    On the programming effort required to generate behavior trees and finite state machines for robotic applications,

    M. Iovino, J. F ¨orster, P. Falco, J. J. Chung, R. Siegwart, and C. Smith, “On the programming effort required to generate behavior trees and finite state machines for robotic applications,” inIEEE International Conference on Robotics and Automation. IEEE, 2023, pp. 5807– 5813

    2023

  12. [18]

    A novel robust imitation learning framework for complex skills with limited demonstrations,

    W. Wang, C. Zeng, H. Zhan, and C. Yang, “A novel robust imitation learning framework for complex skills with limited demonstrations,”IEEE Transactions on Au- tomation Science and Engineering, 2024

    2024

  13. [19]

    Beyond predefined actions: Integrating behavior trees and dynamic movement primitives for robot learning from demonstration,

    D. C. Dom ´ınguez, E. Schaffernicht, and T. Stoyanov, “Beyond predefined actions: Integrating behavior trees and dynamic movement primitives for robot learning from demonstration,”arXiv preprint arXiv:2505.08625, 2025

    2025 arXiv

  14. [20]

    Motion planning with temporal-logic specifications: Progress and challenges,

    E. Plaku and S. Karaman, “Motion planning with temporal-logic specifications: Progress and challenges,” AI Communications, vol. 29, no. 1, pp. 151–162, 2015

    2015

  15. [21]

    Pddl— the planning domain definition language,

    C. Aeronautiques, A. Howe, C. Knoblock, I. D. McDer- mott, A. Ram, M. Veloso, D. Weld, D. W. Sri, A. Barrett, D. Christiansonet al., “Pddl— the planning domain definition language,”Technical Report, Tech. Rep., 1998

    1998

  16. [22]

    Vernacopter: Disambiguated natural-language- driven robot via formal specifications,

    T. V . Laar, Z. Zhang, S. Qi, S. Haesaert, and Z. Sun, “Vernacopter: Disambiguated natural-language- driven robot via formal specifications,”arXiv preprint arXiv:2409.09536, 2024

    2024 arXiv

  17. [23]

    Control barrier functions for signal temporal logic tasks,

    L. Lindemann and D. V . Dimarogonas, “Control barrier functions for signal temporal logic tasks,”IEEE Control Systems Letters, vol. 3, no. 1, pp. 96–101, 2018

    2018

  18. [24]

    Dynamic movement primitives in robotics: A tutorial survey,

    M. Saveriano, F. J. Abu-Dakka, A. Kramberger, and L. Peternel, “Dynamic movement primitives in robotics: A tutorial survey,”The International Journal of Robotics Research, vol. 42, no. 13, pp. 1133–1184, 2023

    2023

  19. [25]

    A novel dmps framework for robot skill generalizing with obstacle avoidance: Taking volume and orientation into consideration,

    Z. Liu and Y . Fang, “A novel dmps framework for robot skill generalizing with obstacle avoidance: Taking volume and orientation into consideration,”IEEE/ASME Transactions on Mechatronics, 2025

    2025

  20. [26]

    Logic program- ming for deliberative robotic task planning,

    D. Meli, H. Nakawala, and P. Fiorini, “Logic program- ming for deliberative robotic task planning,”Artificial Intelligence Review, vol. 56, no. 9, pp. 9011–9049, 2023

    2023

  21. [27]

    Continuous- time control synthesis under nested signal temporal logic specifications,

    P. Yu, X. Tan, and D. V . Dimarogonas, “Continuous- time control synthesis under nested signal temporal logic specifications,”IEEE Transactions on Robotics, vol. 40, pp. 2272–2286, 2024

    2024

  22. [28]

    A flexible and efficient temporal logic tool for python: Pytelo,

    G. A. Cardona, K. Leahy, M. Mann, and C.-I. Vasile, “A flexible and efficient temporal logic tool for python: Pytelo,”arXiv preprint arXiv:2310.08714, 2023

    2023 arXiv

  23. [29]

    Nl2stl: Transformation from logic natural language to signal temporal logics using llama2,

    Y . Mao, T. Zhang, X. Cao, Z. Chen, X. Liang, B. Xu, and H. Fang, “Nl2stl: Transformation from logic natural language to signal temporal logics using llama2,” inIEEE International Conference on Cybernetics and Intelligent Systems and IEEE International Conference on Robotics, ...

    2024

  24. [30]

    Seeing, saying, solving: An llm-to-tl framework for cooperative robots,

    D. B. Choe, S. V . Sangeetha, S. Emanuel, C.-Y . Chiu, S. Coogan, and S. Kousik, “Seeing, saying, solving: An llm-to-tl framework for cooperative robots,”arXiv preprint arXiv:2505.13376, 2025

    2025 arXiv

  25. [31]

    Enhancing transformation from natural language to signal temporal logic using llms with diverse external knowledge,

    Y . Fang, Z. Jin, J. An, H. Chen, X. Chen, and N. Zhan, “Enhancing transformation from natural language to signal temporal logic using llms with diverse external knowledge,”arXiv preprint arXiv:2505.20658, 2025

    2025 arXiv

Pith tools

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