Pith. sign in

REVIEW 4 major objections 5 minor 37 references

DIT* speeds up sampling-based optimal planning by pruning neighbor candidates whose direction does not move toward the goal, while keeping completeness and asymptotic optimality.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

DIT* speeds up sampling-based path planning by removing edges that point away from the goal and scoring remaining edges with a weighted-direction cost.

T0 review reviewed 2026-08-05 challenge →

load-bearing objection DIT* has a genuinely useful heuristic idea and real-robot results, but the paper's proof of asymptotic optimality doesn't account for its own direction filter, and the filter pseudocode appears inverted. the 4 major comments →

arxiv 2508.19168 v1 pith:55OCVYF7 submitted 2025-08-26 cs.RO

Direction Informed Trees (DIT*): Optimal Path Planning via Direction Filter and Direction Cost Heuristic

Liding Zhang , Kejia Chen , Kuanqi Cai , Yu Zhang , Yixuan Dang , Yansong Wu , Zhenshan Bing , Fan Wu
show 2 more authors
Sami Haddadin Alois Knoll
This is my paper
classification cs.RO
keywords sampling-based motion planningasymptotic optimalityprobabilistic completenessdirection heuristicinformed treesgoal biashigh-dimensional planningcosine similarity
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

The reading

Direction Informed Trees (DIT*) is a sampling-based path planner built on the idea that the direction of a candidate edge, not just its length or estimated remaining cost, tells you whether it is worth exploring. DIT* represents each edge as a generalized vector, measures how well a candidate edge aligns with the direction to the goal using a weighted cosine similarity, and discards neighbors whose alignment is no better than the direction the search is already moving. That same directional score supplies a cost heuristic used to break ties between edges of similar computational effort, biasing the tree toward the goal. The paper claims this speeds up initial solution finding and lowers path cost relative to EIT* and other single-query planners in random-rectangle and narrow-passage benchmarks from R4 to R16 and in real 7-DoF and 14-DoF manipulation tasks, while retaining probabilistic completeness and asymptotic optimality. If true, DIT* identifies direction information as a practical lever for making optimal sampling-based planning faster in high-dimensional spaces.

Core claim

The paper's central claim is that DIT* converges to an initial solution and then to a low-cost path faster than existing single-query sampling-based planners because its direction filter removes unpromising nearest-neighbor candidates before expensive edge checks. Formally, DIT* treats edges as vectors in the configuration space, normalizes them, and computes a weighted cosine similarity between the candidate edge vector, the current search-direction vector, and the goal vector. A candidate neighbor survives the filter only when its direction is at least as close to the goal direction as the current edge direction (Φcheck ≥ Φtarget). Surviving edges are scored by an inadmissible direction co

What carries the argument

The mechanism is the direction filter plus direction-cost heuristic built on generalized edge vectors. Every edge (xs, xt) is encoded as a vector in R^n; three unit vectors—target (xlast → xsource), check (xsource → xneighbor), and goal (xlast → xgoal)—are compared with a weighted cosine similarity index Φ ∈ [−1,1]. Algorithm 1 keeps a neighbor only if Φcheck ≥ Φtarget, so the planner refuses to expand edges that would turn the search away from the goal relative to its current heading. Algorithm 2 converts the two similarity indices and vector norms into a scalar inadmissible direction cost, which Algorithm 3 minimizes among edges whose effort estimates are within a threshold. The filter doe

Load-bearing premise

The optimality guarantee is inherited from the standard un-filtered random geometric graph rewiring argument, but DIT* prunes candidate neighbors before rewiring and the paper gives no bound on the probability that the filter keeps the neighbor required for the optimal rewiring step.

What would settle it

Run DIT* on a benchmark where the only feasible path begins with a segment that points away from the goal (e.g., a narrow passage whose entrance lies behind the start). If DIT*'s success rate or final path cost degrades as the required initial deviation angle grows, the filter is discarding necessary directions. A sharper test: construct a single RGG sample where the unique optimal rewiring edge has Φcheck < Φtarget; if DIT*'s Algorithm 1 removes it, that sample cannot rewire optimally, so the planner's asymptotic-optimality claim would fail unless such samples have vanishing probability—which

Watch this falsifier. Get emailed when new claim-graph text bears on it.

If this is right

  • In the paper's benchmarks, time to initial solution improves by roughly 11–45% over EIT* depending on dimension and scenario, so the filter translates directly into faster first paths in 4-to-16-dimensional configuration spaces.
  • Because the direction heuristic is kept in the lowest tier of the queue key, the admissible cost layer still controls the search, which is how DIT* aims to preserve asymptotic optimality while gaining goal bias.
  • Filtering neighbor candidates reduces the number of edges sent to collision checking, the dominant computational cost in sampling-based planning.
  • Real-world results on a 7-DoF kitchen robot and a 14-DoF dual-arm cable-routing task indicate the speedup is not limited to synthetic benchmarks.

Where Pith is reading between the lines

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

  • The asymptotic-optimality argument inherits the unpruned random-geometric-graph lemmas without quantifying how often the direction filter deletes the optimal rewiring parent; a survival-probability bound for near-optimal edges would close this gap.
  • Because the filter compares only directions, it will reject any first move that must head away from the goal to navigate around an obstacle; an environment with a U-shaped detour would directly stress this limitation.
  • The tunable weights and thresholds (ω1, ω2, ξ1, ξ2, µ) are fixed across experiments; adapting them to local obstacle density or homotopy class could be a natural follow-up.
  • DIT* samples uniformly despite using informed-style heuristics; combining its direction filter with ellipsoidal informed sampling could further improve convergence in narrow-passage problems.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. The manuscript proposes Direction Informed Trees (DIT*), a sampling-based motion planner that augments the EIT*/BIT*-style forward/reverse search with (i) a direction filter that prunes nearest-neighbor candidates according to their alignment with the goal direction and (ii) a direction-cost heuristic based on a weighted cosine similarity. The authors claim probabilistic completeness and asymptotic optimality by uniform sampling and by citing Karaman-Frazzoli lemmas, and they report faster initial-solution convergence than EIT*, AIT*, BIT*, etc. in simulated R4–R16 problems and in two real-world manipulation tasks.

Significance. The intended contribution—using directional information to bias edge selection while maintaining optimality guarantees—is relevant to high-dimensional planning, and the real-world experimental domains are appropriate. However, the paper as written does not deliver a rigorous theoretical foundation: the asymptotic optimality argument bypasses the effect of the direction filter, the central filter definition appears internally inconsistent, and the mathematical definition of the similarity index is incorrect. In addition, the experimental reporting is incomplete and contains an internal contradiction. If the algorithm and proofs were corrected, the idea could be valuable; in its current form, the central claims are not supported.

major comments (4)
  1. [Section III-C, Eqs. (12)–(14)] The proof of probabilistic completeness and asymptotic optimality is transferred directly from Karaman-Frazzoli via Lemmas 56/71/72, but the DIT* graph is not the RGG or EIT* graph: Algorithm 1 removes a neighbor whenever Φcheck ≥ Φtarget, before choose-parent/rewire. The probability that the optimal rewiring parent (or any edge needed for connectivity) survives the filter is never analyzed. Without such a bound, the lemmas do not apply, and Eq. (14) is unsupported. The proof must either show that the filter preserves the required graph properties with high probability or explicitly analyze the pruned graph.
  2. [Algorithm 1 and Section III-A] The filter acceptance condition appears inverted relative to the prose. The text states that the direction is valid when Vcheck lies between Vtarget and Vgoal; in that geometric configuration, Φcheck should be greater than or equal to Φtarget (the check vector is closer to the goal direction). However, line 11 of Algorithm 1 deletes xneighbor whenever Φcheck ≥ Φtarget, i.e., it removes exactly the goal-aligned neighbors and retains those deviating away from the goal. Either the pseudocode or the description is wrong. Since no code is provided, the reported empirical results cannot be attributed to a well-defined method.
  3. [Eq. (3)] The weighted cosine similarity is incorrectly formulated. The weights ω1 and ω2 appear only in the denominator, so for unit vectors the index equals cos θ / (|ω1 ω2|), which can exceed 1 in absolute value with the stated weights (ω1 = 0.6, ω2 = 0.4). The claim Φi ∈ [−1, 1] is therefore false, and all threshold comparisons (e.g., |Φ| ≈ 1, Φcheck ≥ Φtarget) used in the algorithm are not well-defined. The formula needs correction (e.g., weights in the numerator as well), and the range proof must be supplied.
  4. [Section IV-A and appended PDT report] Table I reports only medians, with no confidence intervals or statistical tests, despite the paper claiming 'robust' improvements. More seriously, the appended report block in the manuscript (after the references) is a Planner Developer Tools report for 'defaultRandomRectangles2D' that states 'executed 1 runs' and reports identical min/median/max values, while Section IV-A claims 100 runs per scenario. This internal inconsistency undermines the reproducibility of the benchmark and the central empirical claim of faster convergence.
minor comments (5)
  1. [Abstract] 'DIT* convergence faster' should be 'DIT* converges faster'.
  2. [Algorithm 1, line 7] Returning 'False' from a function whose return type is a neighbor set is confusing; this likely should be 'continue' or 'skip'.
  3. [Eq. (2)] The definition Vec := VF(xs) − VF(xt) sets the direction convention; the text should state clearly whether the vector points from source to target or the reverse, because the filter comparisons depend on this orientation.
  4. [Section III-C] The statement that DIT* uses the same choose-parent and rewire strategies as EIT* is vague; a precise description or reference to the EIT* algorithm is needed to support the proof transfer.
  5. [Appendix/PDT block] The automatically generated PDT report appended to the manuscript should be removed or properly integrated with a consistent run count; as written it contradicts the 100-run setup and introduces an apparent artifact.

Circularity Check

0 steps flagged

No significant circularity: the direction filter/heuristic are explicit design definitions, the speedup is measured empirically, and the asymptotic-optimality claim rests on an external lemma rather than a self-referential reduction.

full rationale

DIT*'s central claims are (i) a new algorithmic mechanism (direction filter + direction cost heuristic) and (ii) measured faster convergence on benchmarks. Neither reduces to its own inputs by construction. The filter and heuristic are explicitly defined in Alg. 1 and Eqs. 3-4 in terms of weighted cosine similarity to the goal vector; describing this as "goal bias" restates the design rather than deriving a prediction from it. The empirical convergence improvements in Section IV are measurements, not mathematical consequences of the fitted parameters (ξ1, ξ2, ω1, ω2). The probabilistic-completeness/asymptotic-optimality argument in Section III-C cites Karaman-Frazzoli Lemmas 56/71/72 (Eq. 14) and asserts DIT* uses the same choose-parent/rewire as EIT*, but Alg. 1 prunes neighbors before rewiring, so the cited lemmas' hypotheses (Eq. 13's connection radius for the full RGG) are not re-derived for the filtered graph. That is a missing-support / proof-transfer gap, not circularity: the cited lemmas are external positive results, and the paper does not define the filter in terms of the AO conclusion. The self-citations present (e.g., refs. [7], [23]) appear in related-work context and are not load-bearing for the core claims. The apparent inversion between the prose "direction is valid when V_check lies between V_target and V_goal" and Alg. 1 line 11 (removing Φ_check >= Φ_target) is an internal-consistency issue that should be corrected, but it is not a circular-reasoning step.

Axiom & Free-Parameter Ledger

5 free parameters · 4 axioms · 0 invented entities

The central contribution rests on five hand-set parameters, an unproven extension of standard optimality lemmas to a modified algorithm, and a similarity measure that is not mathematically consistent with its claimed range.

free parameters (5)
  • omega1, omega2 (cosine similarity weights) = omega1=0.6, omega2=0.4
    Chosen by hand in Eq. 3; no sensitivity analysis. They also break the [-1,1] bound claim.
  • xi1, xi2 (direction cost weights) = xi1=0.6, xi2=0.8
    Tuning parameters in Eq. 4 that directly determine the direction cost heuristic.
  • mu (effort tie-break threshold) = 5000
    Hand-set to prioritize direction cost when effort differences are under threshold.
  • eta (RGG constant) = 1.001
    Set in experiments; affects rewiring radius inequality.
  • rewire factor = 1.1
    Set uniformly for all planners; influences graph connectivity.
axioms (4)
  • ad hoc to paper Karaman-Frazzoli lemmas (56, 71, 72) remain valid after applying the direction filter.
    The filter can exclude neighbors, which is not covered by the standard proof assumptions. Used in Section III-C.
  • domain assumption Eq. 3 defines a similarity index in [-1,1].
    The formula is not normalized correctly when omega1 != omega2, yet the paper relies on this range for filtering and cost.
  • standard math Uniform sampling of the informed set yields an implicit RGG as in Penrose/BIT*.
    Standard model, but the direction filter biases edge selection, altering the effective RGG.
  • domain assumption The direction cost heuristic can be used as a third key in the lexicographic queue without losing search completeness.
    No proof that prioritization by direction cost preserves the completeness of the underlying search.

reviewed 2026-08-05 · how reviews work

0 comments
Cite this review

Pith. "Pith review of Direction Informed Trees (DIT*): Optimal Path Planning via Direction Filter and Direction Cost Heuristic." pith.science (2026). https://pith.science/paper/55OCVYF7

@misc{pith2026250819168,
  author       = {Pith},
  title        = {Pith review of: Direction Informed Trees (DIT*): Optimal Path Planning via Direction Filter and Direction Cost Heuristic},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/55OCVYF7}},
  note         = {Machine review of arXiv:2508.19168}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

Optimal path planning requires finding a series of feasible states from the starting point to the goal to optimize objectives. Popular path planning algorithms, such as Effort Informed Trees (EIT*), employ effort heuristics to guide the search. Effective heuristics are accurate and computationally efficient, but achieving both can be challenging due to their conflicting nature. This paper proposes Direction Informed Trees (DIT*), a sampling-based planner that focuses on optimizing the search direction for each edge, resulting in goal bias during exploration. We define edges as generalized vectors and integrate similarity indexes to establish a directional filter that selects the nearest neighbors and estimates direction costs. The estimated direction cost heuristics are utilized in edge evaluation. This strategy allows the exploration to share directional information efficiently. DIT* convergence faster than existing single-query, sampling-based planners on tested problems in R^4 to R^16 and has been demonstrated in real-world environments with various planning tasks. A video showcasing our experimental results is available at: https://youtu.be/2SX6QT2NOek

Figures

Figures reproduced from arXiv: 2508.19168 by Alois Knoll, Fan Wu, Kejia Chen, Kuanqi Cai, Liding Zhang, Sami Haddadin, Yansong Wu, Yixuan Dang, Yu Zhang, Zhenshan Bing.

Figure 1
Figure 1. Figure 1: Planner comparison on dual-arm manipulator robot during [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Last state xlast to source state xsource forms the target vector V⃗target, xsource to next state xneighbor forms vector to check V⃗check, xlast to goal state forms goal vector V⃗goal. Additional V⃗goal,shifted starts at xsource is shifted from xlast and similar to V⃗goal. V⃗goal,shifted and V⃗check created area to explore represented in light purple; this will filter out redundant samples to optimize edge … view at source ↗
Figure 3
Figure 3. Figure 3: Illustrates the real-world scenarios of a manipulator (7-DoF) for the kitchen cooking task, (a) shows the start configuration of the [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Two snapshots of 2D depiction to demonstrate the simulated [PITH_FULL_IMAGE:figures/full_fig_p005_4.png] view at source ↗
Figure 1
Figure 1. Figure 1: Top: Percentage of runs that found a solution at any given time with a Clopper Fig. 5: Detailed experimental results from Section IV-A are presented above. Fig. (a), (c), and (e) depict ten random rectangle outcomes [PITH_FULL_IMAGE:figures/full_fig_p006_1.png] view at source ↗

discussion (0)

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

Reference graph

Works this paper leans on

37 extracted references · 23 canonical work pages

  1. [1]

    Asymptotically optimal sampling- based motion planning methods,

    J. D. Gammell and M. P. Strub, “Asymptotically optimal sampling- based motion planning methods,” Annual Review of Control, Robotics, and Autonomous Systems , vol. 4, pp. 295–318, 2021

    work page 2021

  2. [2]

    A formal basis for the heuristic determination of minimum cost paths,

    P. E. Hart, N. J. Nilsson, and B. Raphael, “A formal basis for the heuristic determination of minimum cost paths,” IEEE transactions on Systems Science and Cybernetics , vol. 4, no. 2, pp. 100–107, 1968

    1968

  3. [3]

    Dynamic programming, princeton univ,

    R. Bellman, “Dynamic programming, princeton univ,” Press Prince- ton, New Jersey, 1957

    work page 1957

  4. [4]

    Analysis of probabilistic roadmaps for path planning,

    L. E. Kavraki, M. N. Kolountzakis, and J.-C. Latombe, “Analysis of probabilistic roadmaps for path planning,” IEEE Transactions on Robotics and automation , vol. 14, no. 1, pp. 166–171, 1998

    work page 1998

  5. [5]

    Rapidly-exploring random trees: A new tool for path planning,

    S. M. LaValle, “Rapidly-exploring random trees: A new tool for path planning,” in Technical Report. Computer Science Department, Iowa State University, 1998

    work page 1998

  6. [6]

    RRT-connect: An efficient approach to single-query path planning,

    J. J. Kuffner and S. M. LaValle, “RRT-connect: An efficient approach to single-query path planning,” in Proceedings 2000 ICRA. Millennium Conference. IEEE International Conference on Robotics and Automa- tion. Symposia Proceedings (Cat. No. 00CH37065) , vol. 2. IEEE, 2000, pp. 995–1001

    work page 2000

  7. [7]

    Tree- based grafting approach for bidirectional motion planning with local subsets optimization,

    L. Zhang, Y . Ling, Z. Bing, F. Wu, S. Haddadin, and A. Knoll, “Tree- based grafting approach for bidirectional motion planning with local subsets optimization,” IEEE Robotics and Automation Letters, vol. 10, no. 6, pp. 5815–5822, 2025

    2025

  8. [8]

    Approaches for heuristically biasing RRT growth,

    C. Urmson and R. Simmons, “Approaches for heuristically biasing RRT growth,” in Proceedings 2003 IEEE/RSJ International Con- ference on Intelligent Robots and Systems (IROS 2003)(Cat. No. 03CH37453), vol. 2. IEEE, 2003, pp. 1178–1183

    work page 2003

  9. [9]

    Bidirectional sampling-based motion plan- ning without two-point boundary value solution,

    S. Nayak and M. W. Otte, “Bidirectional sampling-based motion plan- ning without two-point boundary value solution,” IEEE Transactions on Robotics, vol. 38, no. 6, pp. 3636–3654, 2022

    2022

  10. [10]

    Adaptively informed trees (AIT*) and effort informed trees (EIT*): Asymmetric bidirectional sampling- based path planning,

    M. P. Strub and J. D. Gammell, “Adaptively informed trees (AIT*) and effort informed trees (EIT*): Asymmetric bidirectional sampling- based path planning,” The International Journal of Robotics Research, vol. 41, no. 4, pp. 390–417, 2022

    2022

  11. [11]

    Sampling-based algorithms for optimal motion planning,

    S. Karaman and E. Frazzoli, “Sampling-based algorithms for optimal motion planning,” The International Journal of Robotics Research , vol. 30, no. 7, pp. 846–894, 2011

    2011

  12. [12]

    Informed RRT*: Optimal sampling-based path planning focused via direct sampling of an admissible ellipsoidal heuristic,

    J. D. Gammell, S. S. Srinivasa, and T. D. Barfoot, “Informed RRT*: Optimal sampling-based path planning focused via direct sampling of an admissible ellipsoidal heuristic,” in 2014 IEEE/RSJ International Conference on Intelligent Robots and Systems. IEEE, 2014, pp. 2997– 3004

    work page 2014

  13. [13]

    Informed sampling for asymptotically optimal path planning,

    J. D. Gammell, T. D. Barfoot, and S. S. Srinivasa, “Informed sampling for asymptotically optimal path planning,” IEEE Transactions on Robotics, vol. 34, no. 4, pp. 966–984, 2018

    work page 2018

  14. [14]

    Hierarchical rejection sampling for informed kinodynamic planning in high-dimensional spaces,

    T. Kunz, A. Thomaz, and H. Christensen, “Hierarchical rejection sampling for informed kinodynamic planning in high-dimensional spaces,” in 2016 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2016, pp. 89–96

    work page 2016

  15. [15]

    Gener- alizing informed sampling for asymptotically-optimal sampling-based kinodynamic planning via markov chain monte carlo,

    D. Yi, R. Thakker, C. Gulino, O. Salzman, and S. Srinivasa, “Gener- alizing informed sampling for asymptotically-optimal sampling-based kinodynamic planning via markov chain monte carlo,” in 2018 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2018, pp. 7063–7070

    work page 2018

  16. [16]

    Elliptical k-nearest neighbors: Path optimization via coulomb’s law and invalid vertices in c-space obstacles,

    L. Zhang, Z. Bing, Y . Zhang, K. Cai, L. Chen, F. Wu, S. Had- dadin, and A. Knoll, “Elliptical k-nearest neighbors: Path optimization via coulomb’s law and invalid vertices in c-space obstacles,” 2024 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pp. 12 032–12 039, 2024

    2024

  17. [17]

    Energy-efficient path planning of reconfigurable robots in complex environments,

    P. T. Kyaw, A. V . Le, P. Veerajagadheswar, M. R. Elara, T. T. Thu, N. H. K. Nhan, P. Van Duc, and M. B. Vu, “Energy-efficient path planning of reconfigurable robots in complex environments,” IEEE Transactions on Robotics , vol. 38, no. 4, pp. 2481–2494, 2022

    2022

  18. [18]

    Non-parametric informed exploration for sampling-based motion planning,

    S. S. Joshi and T. Panagiotis, “Non-parametric informed exploration for sampling-based motion planning,” in 2019 International Confer- ence on Robotics and Automation (ICRA) . IEEE, 2019, pp. 5915– 5921

    work page 2019

  19. [19]

    Batch informed trees (BIT*): Sampling-based optimal planning via the heuristically guided search of implicit random geometric graphs,

    J. D. Gammell, S. S. Srinivasa, and T. D. Barfoot, “Batch informed trees (BIT*): Sampling-based optimal planning via the heuristically guided search of implicit random geometric graphs,” in 2015 IEEE international conference on robotics and automation (ICRA) . IEEE, 2015, pp. 3067–3074

    work page 2015

  20. [20]

    Batch informed trees (BIT*): Informed asymptotically optimal anytime search,

    J. D. Gammell, T. D. Barfoot, and S. S. Srinivasa, “Batch informed trees (BIT*): Informed asymptotically optimal anytime search,” The International Journal of Robotics Research , vol. 39, no. 5, pp. 543– 567, 2020

    work page 2020

  21. [21]

    Penrose, Random geometric graphs

    M. Penrose, Random geometric graphs . OUP Oxford, 2003, vol. 5

    2003

  22. [22]

    Advanced BIT* (ABIT*): Sampling- based planning with advanced graph-search techniques,

    M. P. Strub and J. D. Gammell, “Advanced BIT* (ABIT*): Sampling- based planning with advanced graph-search techniques,” in 2020 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2020, pp. 130–136

    work page 2020

  23. [23]

    Flexible informed trees (FIT*): Adaptive batch-size approach in informed sampling- based path planning,

    L. Zhang, Z. Bing, K. Chen, L. Chen, K. Cai, Y . Zhang, F. Wu, P. Krumbholz, Z. Yuan, S. Haddadin, and A. Knoll, “Flexible informed trees (FIT*): Adaptive batch-size approach in informed sampling- based path planning,” 2024 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) , pp. 3146–3152, 2024

    2024

  24. [24]

    Adaptively informed trees (AIT*): Fast asymptotically optimal path planning through adaptive heuristics,

    M. P. Strub and J. D. Gammell, “Adaptively informed trees (AIT*): Fast asymptotically optimal path planning through adaptive heuristics,” in 2020 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2020, pp. 3191–3198

    2020

  25. [25]

    Preserving and combining knowledge in robotic lifelong reinforcement learning,

    Y . Meng, Z. Bing, X. Yao, K. Chen, K. Huang, Y . Gao, F. Sun, and A. Knoll, “Preserving and combining knowledge in robotic lifelong reinforcement learning,” Nature Machine Intelligence, pp. 1–14, 2025

    2025

  26. [26]

    Estimated informed anytime search for sampling-based planning via adaptive sampler,

    L. Zhang, K. Cai, Y . Zhang, Z. Bing, C. Wang, F. Wu, S. Haddadin, and A. Knoll, “Estimated informed anytime search for sampling-based planning via adaptive sampler,” IEEE Transactions on Automation Science and Engineering , vol. 22, pp. 18 580–18 593, 2025

    2025

  27. [27]

    Meta-reinforcement learning in non-stationary and dynamic environments,

    Z. Bing, D. Lerch, K. Huang, and A. Knoll, “Meta-reinforcement learning in non-stationary and dynamic environments,” IEEE Trans- actions on Pattern Analysis and Machine Intelligence , vol. 45, no. 3, pp. 3476–3491, 2023

    work page 2023

  28. [28]

    Solving robotic manipulation with sparse reward reinforcement learn- ing via graph-based diversity and proximity,

    Z. Bing, H. Zhou, R. Li, X. Su, F. O. Morin, K. Huang, and A. Knoll, “Solving robotic manipulation with sparse reward reinforcement learn- ing via graph-based diversity and proximity,” IEEE Transactions on Industrial Electronics, vol. 70, no. 3, pp. 2759–2769, 2023

    work page 2023

  29. [29]

    Lateral flexion of a compliant spine improves motor performance in a bioinspired mouse robot,

    Z. Bing, A. Rohregger, F. Walter, Y . Huang, P. Lucas, F. O. Morin, K. Huang, and A. Knoll, “Lateral flexion of a compliant spine improves motor performance in a bioinspired mouse robot,” Science Robotics, vol. 8, no. 85, 2023

    work page 2023

  30. [30]

    Cloud- native fog robotics: Model-based deployment and evaluation of real- time applications,

    L. Wen, Y . Zhang, M. Rickert, J. Lin, F. Pan, and A. Knoll, “Cloud- native fog robotics: Model-based deployment and evaluation of real- time applications,” IEEE Robotics and Automation Letters , vol. 10, no. 1, pp. 398–405, 2025

    work page 2025

  31. [31]

    Motion planning for robotics: A review for sampling-based planners,

    L. Zhang, K. Cai, Z. Sun, Z. Bing, C. Wang, L. Figueredo, S. Had- dadin, and A. Knoll, “Motion planning for robotics: A review for sampling-based planners,” Biomimetic Intelligence and Robotics , vol. 5, no. 1, p. 100207, 2025

    2025

  32. [32]

    Complex robotic manipulation via graph-based hindsight goal generation,

    Z. Bing, M. Brucker, F. O. Morin, R. Li, X. Su, K. Huang, and A. Knoll, “Complex robotic manipulation via graph-based hindsight goal generation,” IEEE Transactions on Neural Networks and Learn- ing Systems, vol. 33, no. 12, pp. 7863–7876, 2022

    work page 2022

  33. [33]

    Contact-aware shaping and maintenance of deformable lin- ear objects with fixtures,

    K. Chen, Z. Bing, F. Wu, Y . Meng, A. Kraft, S. Haddadin, and A. Knoll, “Contact-aware shaping and maintenance of deformable lin- ear objects with fixtures,” in 2023 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) , 2023, pp. 1–8

    work page 2023

  34. [34]

    Context-based meta- reinforcement learning with bayesian nonparametric models,

    Z. Bing, Y . Yun, K. Huang, and A. Knoll, “Context-based meta- reinforcement learning with bayesian nonparametric models,” IEEE Transactions on Pattern Analysis and Machine Intelligence , pp. 1–18, 2024

    work page 2024

  35. [35]

    The open motion planning library,

    I. A. Sucan, M. Moll, and L. E. Kavraki, “The open motion planning library,” IEEE Robotics & Automation Magazine , vol. 19, no. 4, pp. 72–82, 2012

    2012

  36. [36]

    Benchmarking motion planning algorithms: An extensible infrastructure for analysis and visualization,

    M. Moll, I. A. Sucan, and L. E. Kavraki, “Benchmarking motion planning algorithms: An extensible infrastructure for analysis and visualization,” IEEE Robotics & Automation Magazine , vol. 22, no. 3, pp. 96–102, 2015

    work page 2015

  37. [37]

    Planner developer tools (PDT): Reproducible experiments and statistical analysis for developing and testing motion planners,

    J. D. Gammell, M. P. Strub, and V . N. Hartmann, “Planner developer tools (PDT): Reproducible experiments and statistical analysis for developing and testing motion planners,” in Proceedings of the Work- shop on Evaluating Motion Planning Performance (EMPP), IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) , 2022

    work page 2022

This paper was first reviewed by deepseek-v4-flash on August 5, 2026.