Pith. sign in

REVIEW 4 major objections 5 minor 36 references

Dynamically Feasible Path Planning in Cluttered Environments via Reachable Bezier Polytopes

T0 review · 4 major / 5 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read Reachable Bézier polytopes make graph paths provably safe to track

desk verdict Real-time kinodynamic planner with a strong GPU cut heuristic and hardware demo, but the headline safety theorem has a fixable gap: the graph is cut against raw obstacles, not the tracking-error-inflated ones. read the letter →

arxiv 2411.13507 v1 pith:AWURMWL4 submitted 2024-11-20 cs.RO cs.SYeess.SY

Noel Csomay-Shanklin , William D. Compton , Aaron D. Ames This is my paper · ORCID
classification cs.ROcs.SYeess.SY
keywords reachableBézierpolytopeskinodynamicpathplanninglayeredcontrolarchitecturemodelpredictivecurvesreal-timeGPUaccelerationleggedrobots
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 argues that a conventional two-step pipeline — plan a coarse path, then track it with a low-level controller — can be made to carry formal guarantees if the path is built from Bézier curves selected by a reachability oracle. The authors propose building a graph whose edges are Bézier curves that, when tracked, provably keep the full closed-loop system inside its state and input constraints, cutting edges that collide with obstacles and refining the surviving path with model predictive control. Their central result states that if such a graph path exists, the closed-loop MPC execution reaches a neighborhood of the goal while respecting constraints for all time. They demonstrate the architecture in real time on the 3D hopping robot ARCHER, using GPU acceleration to replan through cluttered environments at 10–50 Hz. This matters because it replaces the usual leap of faith that a low-level tracker will keep the robot safe with a check performed at planning time.

What carries the argument

The central object is the reachable Bézier polytope — a polytopic reachable set in the space of polynomial Bézier reference trajectories, encoded by matrices F and G (Theorem 1). For any two boundary states x1, x2 satisfying F[x1;x2] ≤ G, there exists a Bézier curve that, when tracked by the given feedback controller, keeps the closed-loop system within the specified state and input constraint sets. This oracle lets the planner treat dynamics as a graph edge check rather than a differential constraint. Around it, the paper assembles standard Bézier facts — the convex-hull property (which makes collision checking a linear program over control points), a path-length bound used as the edge cost, and subdivision used to discretize the curve into MPC references — and adds a three-stage heuristic that eliminates most collision checks and maps cleanly onto a GPU.

What would settle it

Take ARCHER's actual Raibert-style controller, compute (or measure) its worst-case tracking error bound E, and search one of the Bézier graphs built in the paper for a pair of vertices satisfying F[x1;x2] ≤ G whose connecting Bézier curve, when tracked in closed loop, leaves the widened corridor Xd\O⊕E or violates an input limit; if such a pair exists, the oracle is too optimistic and Theorem 2's conclusion collapses.

Watch

Extended reading notes

Core claim

The paper's central claim is Theorem 2: if the graph-search problem Problem 2 is feasible, then applying the MPC refinement (MPC) in closed loop solves Problem 1 — the full-order closed-loop system keeps its projected state inside the free space, respects input bounds for all time, and converges to a neighborhood of the goal. The proof works by showing that the graph path gives a feasible warm start for MPC, and that each MPC segment, connected to the next by a Bézier curve guaranteed by Theorem 1 from the companion paper, keeps the closed-loop trajectory inside the same constraint sets; standard robust tube MPC theory then supplies recursive feasibility and stability. The paper thereby claims that checking the pair of endpoint states against the reachable-polytope matrices F and G is sufficient to certify dynamic feasibility of an entire path through nonconvex space.

Load-bearing premise

The entire guarantee rests on the companion-paper oracle (Theorem 1): that any pair of states passing the F,G check really can be connected by a Bézier curve whose closed-loop tracking stays inside the state and input constraints for the specific Raibert-style controller used on ARCHER — and this paper neither proves that theorem nor verifies its conditions or provides the worst-case tracking error bound E for that controller.

Editorial extensions

If this is right

  • Any graph edge that passes the F,G check is a provably trackable reference for the full-order system, so dynamic feasibility is certified at the planning layer rather than assumed of the tracker.
  • Because the graph path is a feasible warm start for MPC, the refined trajectory inherits the guarantee: if the graph problem is feasible at any replan instant, the closed-loop system stays inside Xd\O and U for all time.
  • The architecture is agnostic to the specific robot: it needs only a tracking controller with a bounded error set E and the corresponding F,G matrices, so the same pipeline can be retargeted to other nonlinear systems.
  • Real-time replanning at 10–50 Hz with 50,000 edges is achievable by GPU offloading of the cut heuristic and the separating-hyperplane computations, enabling online response to moving obstacles.
  • The path-length upper bound from Property 2 gives an admissible cost for Dijkstra, so the coarse path is optimal with respect to that bound before MPC refines it.

Reading between the lines

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

  • Editorial extension: the same reachable-polytope oracle could be dropped into sampling-based kinodynamic planners (e.g., RRT*) as a steering primitive, replacing two-point boundary value solves with a single matrix inequality check.
  • Editorial extension: the paper's guarantee is only as tight as the worst-case tracking error bound E; if an implementation cannot compute a trustworthy E for its controller, the practical safety margin is unknown, and an optimistic E could invalidate the collision-avoidance claim.
  • Editorial extension: because the cut heuristic returns 'indeterminate' for a small fraction of edges and the QP resolves those, the GPU speedup is preserved; a similar three-stage pruning pattern could accelerate collision checking for other convex-hull parameterized curves (B-splines, NURBS).
  • Editorial extension: the F,G oracle implicitly assumes the obstacle map is static between replans; with moving obstacles the graph must be rebuilt each cycle, and the 50 Hz replan rate shown here suggests the practical limit on obstacle dynamics the current architecture can tolerate.
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 / 5 minor

Summary. The paper proposes a layered kinodynamic planning framework for nonlinear systems in cluttered environments. A graph is built by sampling reduced-order states and connecting them with Bézier curves whose dynamic feasibility is certified by reachable polytopes from a companion preprint [30]. The graph is pruned by checking whether the convex hull of each Bézier curve's control points intersects obstacles, a shortest path is found with Dijkstra, and the path is refined by an MPC that tracks the graph solution while enforcing a collision-free corridor. The main theoretical claim (Theorem 2) is that a feasible graph path guarantees that the closed-loop system under MPC solves the original constrained control problem. The approach is demonstrated on the ARCHER 3D hopping robot in simulation and on hardware, with GPU acceleration for real-time performance.

Significance. If the stated guarantees held, the architecture would be a valuable example of combining discrete graph search with continuous optimization while preserving feasibility for nonlinear systems. The hardware experiments and GPU implementation are significant engineering contributions, and the paper provides code and video links, which is commendable. However, the central theoretical guarantee has a substantial gap: the graph is cut against raw obstacles rather than the E-inflated obstacles required by the tracking-error bound, and the proof of Theorem 2 does not verify the conditions needed for the cited tube MPC theory. These issues undermine the closed-loop guarantee as stated and require a major revision.

major comments (4)
  1. [III-B and III-D, Theorem 2 proof] The graph cut in Section III-B only removes edges whose Bézier control-point convex hull intersects the raw obstacle set O (conv(P) ∩ O = ∅), whereas the MPC constraint (7c) requires the nominal state sequence to lie in Xd \(O ⊕ E) and Definition 1 implies the full-order state satisfies Π(x(t)) ∈ x_d(t) ⊕ E. Therefore, an edge surviving the graph cut can pass within distance E of an obstacle, and the closed-loop trajectory can enter O even though the nominal curve avoids O. The proof of Theorem 2 states 'As v_k is in the collision-free graph C, it satisfies (7d) and (7c),' but this implication is false: membership in C only guarantees that the convex hull avoids O, not O ⊕ E. Consequently, the graph path is not generally a feasible warm start for the MPC, and the recursive-feasibility argument collapses. The repair is to run the graph cut against O ⊕ E (or erode the free space accordingly) and to state this explicitly; this is a load-bearing fix, not a cosmetic one.
  2. [Theorem 2 proof] The proof appeals to 'standard Robust tube MPC theory [31]' to claim recursive feasibility and robust stability, but the MPC formulation (7) is not shown to satisfy the assumptions of that theory. In particular, the terminal equality constraint (7e) x_N = r_N is not accompanied by a terminal invariant set or a terminal controller, and no tube tightening is defined for the state and input constraints under the tracking error E. Without these ingredients, the cited reference does not by itself establish recursive feasibility or robust stability. The authors should either provide a self-contained proof of these properties for (7) or explicitly verify the hypotheses of the tube MPC result they invoke, including the role of the reachability oracle in providing a terminal controller.
  3. [II-B, Theorem 1] The central feasibility oracle, Theorem 1 from the companion preprint [30], is imported without proof or numerical verification in this paper. The matrices F and G, the worst-case tracking error bound E, and the class of tracking controllers for which the theorem holds are not specified for the ARCHER system. Since every graph edge and the MPC constraint (7d) depend on this theorem, a failure of its assumptions would invalidate Theorem 2. The authors should make the companion result accessible (e.g., include the theorem statement with its assumptions, or provide the code/data that verifies the oracle for the specific controller), or state clearly that the main claim is conditional on an unverified external result.
  4. [III-D, Eq. (7c)] The set membership notation in D[x_k; x_{k+1}] ∈ Xd \(O ⊕ E) is ambiguous because D[x_k; x_{k+1}] is a matrix of Bézier control points, not an element of R^n. If the intended meaning is that each control point lies in the eroded free space, this is still a different condition from the convex hull avoiding O ⊕ E, and the paper should specify which condition is used. This ambiguity matters because the proof of Theorem 2 relies on (7c) being satisfied by the graph path, and the current text does not make the precise constraint clear.
minor comments (5)
  1. [Abstract] The word 'polyonmials' in the abstract is a typo and should be 'polynomials'.
  2. [Section II-B, Lemma 1] The statement 'there exists a unique matrix D' could benefit from a brief explanation of how uniqueness follows from the chosen Bézier degree p = 2γ−1, since this is not immediately obvious from the text.
  3. [Section III-C] The text uses 'Dijkstra‘s algorithm' with a curly apostrophe; please use a standard apostrophe for consistency.
  4. [Table I] The table columns are not labeled; the reader cannot immediately tell which column corresponds to which function (cutHeuristic vs. adjacentHyperplane). Adding column headers would improve clarity.
  5. [Section IV] The description of the projection map Π as 'the restriction of the full order state to the center of mass x and y positions and velocities' is clear, but the sentence immediately following it ('This desired quaternion is then tracked by a low-level controller...') switches to 'u(x,t)' without defining all variables; please define q, q_d, and ω_d in the text for completeness.

Circularity Check

1 steps flagged · score 3.0 of 10

Theorem 2's guarantee is structurally underwritten by the self-cited reachable-polytope oracle [30]; a separate O vs O+E gap affects correctness, not circularity.

  1. self citation load bearing [Section II-B Theorem 1 and Section III-D Theorem 2 proof]
    "Theorem 1 ([30]). Given a convex state constraint set Xd ⊂ Rn, input constraint set U ⊂RM and error tracking bound ¯E, there exist matrices F and G such that any two points x1, x2 ∈ Rn satisfying: F [x⊤1 x⊤2]⊤ ≤ G implies the existence of a Bézier curve xd with xd(0) = x1, xd(T ) = x2, such that, when tracked, the closed loop system satisfies Π(xcl(t)) ∈ Xd and k(xcl(t), xd(t)) ∈ U."

    The paper's central guarantee, Theorem 2, is proved by invoking this companion-preprint oracle twice: graph edges are admitted exactly when they satisfy F[x1; x2] ≤ G, and the closed-loop segment property in the MPC recursion is asserted 'By Theorem 1'. The theorem is from [30], authored by the same research group, and is not proved, machine-checked, or validated in this paper; no worst-case error bound E is computed for the ARCHER Raibert-style controller, which is only written down as a formula. Thus the claimed first-principles solution of Problem 1 inherits its core feasibility content from an unverified self-citation rather than from a derivation contained in this paper.

full rationale

The derivation chain is not circular in the fitting or renaming sense: no parameter is fitted to data and then reported as a prediction, and the Bézier graph/MPC pipeline exploits standard Bézier properties plus a reachable-set abstraction. The main circularity-adjacent concern is load-bearing self-citation: Theorem 1 from [30] supplies both the graph feasibility oracle and the closed-loop invariant needed by the proof of Theorem 2. Since [30] is a same-author preprint not verified here, the strongest guarantee reduces to that citation. Separately, there is a correctness gap (not a circularity) in the proof of Theorem 2: the graph cut (Cut-QP) checks conv(P) ∩ O = ∅, while MPC constraint (7c) requires avoidance of O ⊕ E; the sentence 'As v∗k is in the collision-free graph C, it satisfies (7d) and (7c)' is therefore unjustified, and the graph path may not be a feasible warm start even if Theorem 1 is accepted. This affects soundness, not self-reference. The hardware experiments and code provide independent empirical support for the practical pipeline, so the score is moderate rather than high.

Assumptions & free parameters 5 free parameters · 5 assumptions · 0 invented entities

The central feasibility oracle, the tracking error bound, and the recursive feasibility of the MPC are all assumed rather than derived in this paper. The graph size and timing parameters are hand-chosen. No new physical entities are introduced.

free parameters (5)
  • Graph node count N = 5000
    Number of sampled states in the graph, chosen by hand. It controls coverage and runtime, and no completeness guarantee is given.
  • Bezier curve time interval T = 0.5 s
    Horizon of each graph edge, chosen by hand. It determines the reachable Bezier polytopes and the size of the graph.
  • MPC horizon and time step = N=50, h=0.1 s
    Chosen for real-time operation at 100-200 Hz; affects optimality and feasibility of the refinement stage.
  • SQP iterations = 1
    One SQP iteration per MPC solve in practice; no proof that a single linearization preserves the collision-avoidance guarantee.
  • Path length vs tracking cost weights = scaled significantly higher
    Hand-tuned to incentivize shorter paths; not reported numerically.
assumptions (5)
  • domain assumption Theorem 1 holds: there exist matrices F and G defining reachable Bezier polytopes such that boundary points satisfying F[x1;x2] <= G imply existence of a dynamically feasible Bezier curve that satisfies constraints when tracked.
    Central oracle, cited from same-author preprint [30]; not proved or independently verified here.
  • domain assumption A tracking controller k with a bounded worst-case tracking error E exists for the ARCHER robot (Definition 1).
    Section IV assumes a Raibert-style controller and low-level quaternion tracking, but does not compute E or verify the invariant set.
  • domain assumption Robust tube MPC theory ([31]) applies to the MPC formulation (7) including terminal equality (7e) and linearized obstacle constraints.
    Theorem 2 proof appeals to this theory without verifying terminal ingredients, tube tightening, or disturbance model.
  • domain assumption Free space is given by Xd\O with each obstacle Oi a convex polytope (Assumption 1).
    Used for the collision cut QP and separating hyperplane checks; real obstacles from SAM2 segmentation are not guaranteed convex.
  • standard math Bezier subdivision (Property 3) allows sampling the graph curve to produce reference points with the same dynamic feasibility property.
    Used to generate r_k for MPC; the subdivided curve between sampled points is a Bezier curve, but whether it inherits Theorem 1's reachability condition is not shown.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Dynamically Feasible Path Planning in Cluttered Environments via Reachable Bezier Polytopes." pith.science (2026). https://pith.science/paper/AWURMWL4

@misc{pith2026241113507,
  author       = {Pith},
  title        = {Pith review of: Dynamically Feasible Path Planning in Cluttered Environments via Reachable Bezier Polytopes},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/AWURMWL4}},
  note         = {Machine review of arXiv:2411.13507}
}
read the original abstract

The deployment of robotic systems in real world environments requires the ability to quickly produce paths through cluttered, non-convex spaces. These planned trajectories must be both kinematically feasible (i.e., collision free) and dynamically feasible (i.e., satisfy the underlying system dynamics), necessitating a consideration of both the free space and the dynamics of the robot in the path planning phase. In this work, we explore the application of reachable Bezier polytopes as an efficient tool for generating trajectories satisfying both kinematic and dynamic requirements. Furthermore, we demonstrate that by offloading specific computation tasks to the GPU, such an algorithm can meet tight real time requirements. We propose a layered control architecture that efficiently produces collision free and dynamically feasible paths for nonlinear control systems, and demonstrate the framework on the tasks of 3D hopping in a cluttered environment.

Figures

Figures reproduced from arXiv: 2411.13507 by the authors.

Figure 1
Figure 1. The proposed framework performing path planning around [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. The path planning framework presented in Algorithm 1. From left to right: a) A Bézier graph is constructed, b) it is cut based on [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. The heuristic employed to check if the Bézier curve is [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: 500 randomly generated obstacles with graph replanning at [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]
Figure 6
Figure 6. Figure 6: Experiments run on the ARCHER hardware Platform. (Top) 3 snapshots of the graph solve for various obstacle configurations. [PITH_FULL_IMAGE:figures/full_fig_p006_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

36 extracted references · 25 canonical work pages

  1. [31]

    Rawlings, D

    J. Rawlings, D. Mayne, and M. Diehl, Model Predictive Control: The- ory, Computation, and Design . Nob Hill Publishing, 2017. [Online]. Available: https://books.google.com/books?id=MrJctAEACAAJ

    work page 2017

  2. [30]

    Reachable bézier polytopes: A primitive for layered motion planning,

    N. Csomay-Shanklin and A. D. Ames, “Reachable bézier polytopes: A primitive for layered motion planning,” preprint, 2024. [Online]. Available: https://noelc-s.github.io/website/research/2024/09/30/ACC_ Bezier/

    work page 2024

  3. [1]

    A Mobile Automaton: An Application of Artificial Intelligence Techniques:,

    N. J. Nilsson, “A Mobile Automaton: An Application of Artificial Intelligence Techniques:,” Defense Technical Information Center, Fort Belvoir, V A, Tech. Rep., Jan. 1969. [Online]. Available: http://www.dtic.mil/docs/citations/ADA459660

    work page 1969

  4. [2]

    A survey of learning-based robot motion planning,

    J. Wang, T. Zhang, N. Ma, Z. Li, H. Ma, F. Meng, and M. Q.-H. Meng, “A survey of learning-based robot motion planning,” IET Cyber- Systems and Robotics , vol. 3, no. 4, pp. 302–314, 2021, _eprint: https://onlinelibrary.wiley.com/doi/pdf/10.1049/csy2.12020. [Online]. Available: https://onlinelibrary.wiley.com/doi/abs/10.1049/csy2.12020

    work page doi:10.1049/csy2.12020 2021

  5. [3]

    Motion Planning Networks: Bridging the Gap Between Learning-based and Classical Motion Planners

    A. H. Qureshi, Y . Miao, A. Simeonov, and M. C. Yip, “Motion Planning Networks: Bridging the Gap Between Learning-based and Classical Motion Planners,” Jun. 2020, arXiv:1907.06013 [cs]. [Online]. Available: http://arxiv.org/abs/1907.06013

    work page Pith review arXiv 2020

  6. [4]

    From Perception to Decision: A Data-driven Approach to End-to-end Motion Planning for Autonomous Ground Robots

    M. Pfeiffer, M. Schaeuble, J. Nieto, R. Siegwart, and C. Cadena, “From Perception to Decision: A Data-driven Approach to End-to-end Motion Planning for Autonomous Ground Robots,” in 2017 IEEE International Conference on Robotics and Automation (ICRA) , May 2017, pp. 1527–1533, arXiv:1609.07910 [cs]. [Online]. Available: http://arxiv.org/abs/1609.07910

    work page Pith review arXiv 2017

  7. [5]

    Neural Path Planning: Fixed Time, Near-Optimal Path Generation via Oracle Imitation,

    M. J. Bency, A. H. Qureshi, and M. C. Yip, “Neural Path Planning: Fixed Time, Near-Optimal Path Generation via Oracle Imitation,” Apr. 2019, arXiv:1904.11102 [cs]. [Online]. Available: http://arxiv.org/abs/1904.11102

    arXiv 2019

  8. [6]

    Kinodynamic Planning,

    E. Schmerling and M. Pavone, “Kinodynamic Planning,” in Encyclope- dia of Robotics, M. H. Ang, O. Khatib, and B. Siciliano, Eds. Berlin, Heidelberg: Springer Berlin Heidelberg, 2021, pp. 1–9

    work page 2021

Show all 36 references
  1. [7]

    Kinodynamic motion planning,

    B. Donald, P. Xavier, J. Canny, and J. Reif, “Kinodynamic motion planning,” Journal of the ACM , vol. 40, no. 5, pp. 1048–1066, Nov. 1993

    1993

  2. [8]

    Randomized Kinodynamic Planning,

    S. M. LaValle and J. J. Kuffner, “Randomized Kinodynamic Planning,” The International Journal of Robotics Research , vol. 20, no. 5, pp. 378–400, May 2001

    2001

  3. [9]

    Kinodynamic RRT*: Asymptotically optimal motion planning for robots with linear dynamics,

    D. J. Webb and J. van den Berg, “Kinodynamic RRT*: Asymptotically optimal motion planning for robots with linear dynamics,” in 2013 IEEE International Conference on Robotics and Automation . Karlsruhe, Germany: IEEE, May 2013, pp. 5054–5061. [Online]. Available: http://ieeexpl...

    2013

  4. [10]

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

    S. M. LaValle, “Rapidly-exploring random trees : A new tool for path planning,” The annual research report , 1998

    1998

  5. [11]

    Reachability-guided sam- pling for planning under differential constraints,

    A. Shkolnik, M. Walter, and R. Tedrake, “Reachability-guided sam- pling for planning under differential constraints,” in 2009 IEEE In- ternational Conference on Robotics and Automation , 2009, pp. 2859– 2865

    2009

  6. [12]

    Optimal paths for a car that goes both forwards and backwards

    J. A. Reeds and L. A. Shepp, “Optimal paths for a car that goes both forwards and backwards.” Pacific Journal of Mathematics, Oct. 1990. [Online]. Available: https://www.scinapse.io/papers/1971998222

    1990

  7. [13]

    Planning, Fast and Slow: A Framework for Adaptive Real-Time Safe Trajectory Planning,

    D. Fridovich-Keil, S. L. Herbert, J. F. Fisac, S. Deglurkar, and C. J. Tomlin, “Planning, Fast and Slow: A Framework for Adaptive Real-Time Safe Trajectory Planning,” Mar. 2018, arXiv:1710.04731 [cs]. [Online]. Available: http://arxiv.org/abs/1710.04731

    2018 arXiv

  8. [14]

    Reachability-guided sam- pling for planning under differential constraints,

    A. Shkolnik, M. Walter, and R. Tedrake, “Reachability-guided sam- pling for planning under differential constraints,” in 2009 IEEE International Conference on Robotics and Automation , ser. 1, May 2009, pp. 2859–2865, iSSN: 1050-4729

    2009

  9. [15]

    R3T: Rapidly-exploring Random Reachable Set Tree for Optimal Kinodynamic Planning of Nonlinear Hybrid Systems,

    A. Wu, S. Sadraddini, and R. Tedrake, “R3T: Rapidly-exploring Random Reachable Set Tree for Optimal Kinodynamic Planning of Nonlinear Hybrid Systems,” in 2020 IEEE International Conference on Robotics and Automation (ICRA), ser. 1, May 2020, pp. 4245–4251, iSSN: 2577-087X

    2020

  10. [16]

    Mixed- integer programming in motion planning,

    D. Ioan, I. Prodan, S. Olaru, F. Stoican, and S.-I. Niculescu, “Mixed- integer programming in motion planning,” Annual Reviews in Control, vol. 51, pp. 65–87, 2021

    2021

  11. [17]

    Efficient mixed-integer planning for UA Vs in cluttered environments,

    R. Deits and R. Tedrake, “Efficient mixed-integer planning for UA Vs in cluttered environments,” in 2015 IEEE International Conference on Robotics and Automation (ICRA) , ser. 1. Seattle, W A, USA: IEEE, May 2015, pp. 42–49. [Online]. Available: http://ieeexplore.ieee.org/docu...

    2015

  12. [18]

    Shortest paths in graphs of convex sets,

    T. Marcucci, J. Umenberger, P. A. Parrilo, and R. Tedrake, “Shortest paths in graphs of convex sets,” arXiv:2101.11565, 2021

    2021 arXiv

  13. [19]

    Motion Planning around Obstacles with Convex Optimization,

    T. Marcucci, M. Petersen, D. von Wrangel, and R. Tedrake, “Motion Planning around Obstacles with Convex Optimization,” May 2022, arXiv:2205.04422 [cs]. [Online]. Available: http://arxiv.org/abs/2205. 04422

    2022 arXiv

  14. [20]

    An efficient and direct method for trajectory optimization of robots constrained by contact kinematics and forces,

    J. Lee, E. Bakolas, and L. Sentis, “An efficient and direct method for trajectory optimization of robots constrained by contact kinematics and forces,” Autonomous Robots, vol. 45, pp. 135–153, 2021

    2021

  15. [21]

    Embedding nonlinear optimization in rrt for optimal kinodynamic planning,

    S. Stoneman and R. Lampariello, “Embedding nonlinear optimization in rrt for optimal kinodynamic planning,” in 53rd IEEE Conference on Decision and Control . IEEE, 2014, pp. 3737–3744

    2014

  16. [22]

    A kinody- namic steering-method for legged multi-contact locomotion,

    P. Fernbach, S. Tonneau, A. Del Prete, and M. Taïx, “A kinody- namic steering-method for legged multi-contact locomotion,” in 2017 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). IEEE, 2017, pp. 3701–3707

    2017

  17. [23]

    A quantitative framework for layered multirate control: Toward a theory of control architecture,

    N. Matni, A. D. Ames, and J. C. Doyle, “A quantitative framework for layered multirate control: Toward a theory of control architecture,” IEEE Control Systems Magazine , vol. 44, no. 3, pp. 52–94, 2024

    2024

  18. [24]

    Fast Path Planning Through Large Collections of Safe Boxes,

    T. Marcucci, P. Nobel, R. Tedrake, and S. Boyd, “Fast Path Planning Through Large Collections of Safe Boxes,” May 2023, arXiv:2305.01072 [cs, eess]. [Online]. Available: http://arxiv.org/abs/ 2305.01072

    2023 arXiv

  19. [25]

    Multi-Rate Planning and Control of Uncertain Nonlinear Systems: Model Predictive Control and Control Lyapunov Functions,

    N. Csomay-Shanklin, A. J. Taylor, U. Rosolia, and A. D. Ames, “Multi-Rate Planning and Control of Uncertain Nonlinear Systems: Model Predictive Control and Control Lyapunov Functions,” in 2022 IEEE 61st Conference on Decision and Control (CDC) , Dec. 2022, pp. 3732–3739, iSSN:...

    2022

  20. [26]

    Creating ARCHER: A 3D Hopping Robot with Flywheels for Attitude Control,

    E. R. Ambrose, “Creating ARCHER: A 3D Hopping Robot with Flywheels for Attitude Control,” Ph.D. dissertation, California Institute of Technology, 2022

    2022

  21. [27]

    Fast online trajectory optimization for the bipedal robot cassie

    T. Apgar, P. Clary, K. Green, A. Fern, and J. W. Hurst, “Fast online trajectory optimization for the bipedal robot cassie.” in Robotics: Science and Systems , vol. 101, 2018, p. 14

    2018

  22. [28]

    On characterizations of the input- to-state stability property,

    E. D. Sontag and Y . Wang, “On characterizations of the input- to-state stability property,” Systems & Control Letters , vol. 24, no. 5, pp. 351–359, Apr. 1995. [Online]. Available: https: //www.sciencedirect.com/science/article/pii/0167691194000506

    1995

  23. [29]

    A primer on bézier curves,

    M. Kamermans, “A primer on bézier curves,” (online book) , 2020. [Online]. Available: https://pomax.github.io/bezierinfo/

    2020

  24. [32]

    Supplemental video

    “Supplemental video.” [Online]. Available: https://vimeo.com/ 1009702220

  25. [33]

    [Online]

    “Code,” 2024. [Online]. Available: https://github.com/noelc-s/ PathPlanner

    2024

  26. [34]

    Nonlinear model predictive control of a 3d hopping robot: Leveraging lie group integrators for dynamically stable behaviors,

    N. Csomay-Shanklin, V . D. Dorobantu, and A. D. Ames, “Nonlinear model predictive control of a 3d hopping robot: Leveraging lie group integrators for dynamically stable behaviors,” in 2023 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2023, pp. 12 106–12 112

    2023

  27. [35]

    OSQP: an operator splitting solver for quadratic programs,

    B. Stellato, G. Banjac, P. Goulart, A. Bemporad, and S. Boyd, “OSQP: an operator splitting solver for quadratic programs,” Mathematical Programming Computation , vol. 12, no. 4, pp. 637–672, 2020. [Online]. Available: https://doi.org/10.1007/s12532-020-00179-2

    2020 doi

  28. [36]

    Segment anything,

    A. Kirillov, E. Mintun, N. Ravi, H. Mao, C. Rolland, L. Gustafson, T. Xiao, S. Whitehead, A. C. Berg, W.-Y . Lo, P. Dollár, and R. Gir- shick, “Segment anything,” arXiv:2304.02643, 2023

    2023 arXiv

Pith tools

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