Pith. sign in

REVIEW 4 major objections 4 minor 1 cited by

Contact-Aware Motion Planning Among Movable Objects

T0 review · 4 major / 4 minor · reviewed 2026-08-09 · deepseek-v4-flash

Pith's one-line read CAMP, a contact-aware motion planner, lets a mobile robot push movable objects during planned trajectories, expanding its reachable space and raising task success rates in simulation and real-world tests.

desk verdict A useful integration of complementarity constraints into MINCO-based planning, but the friction model is too hand-wavy to support the feasibility claim. read the letter →

arxiv 2502.03317 v1 pith:BXHWSRUS submitted 2025-02-05 cs.RO

classification cs.RO
keywords contact-awaremotionplanningcomplementarityconstraintsaugmentedLagrangianmethodnavigationamongmovableobjectsrearrangementoftrajectoryoptimizationmobilerobotnon-prehensilemanipulation
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 mobile robot motion planning should not treat every object as a static obstacle. It introduces CAMP, a trajectory optimization framework in which contact between the robot and movable objects is planned rather than avoided. Contact is encoded as complementarity constraints, and the resulting optimization is solved with an augmented Lagrangian method. In simulations, CAMP raises average success in navigation among movable objects (NAMO) to 95% from 54.3% for a collision-free baseline, and achieves 95-100% success on rearrangement of movable objects (RAMO). Real-world tests show the planned pushing trajectories are executable, which matters because deliberate contact lets robots move through cluttered human environments.

What carries the argument

The load-bearing object is the complementarity constraint on contact, written as $\lVert \dot{O}_i(t)-\dot{O}_j(t)\rVert_2 \cdot \lVert f_C(\ddot{O}_k(t))\rVert_2 = 0$ with both factors constrained non-negative: agents may move apart or transmit contact force, but not both at the same point. This expresses non-penetration and stick/slip behavior without prescribing a mode schedule in advance. The trajectory parameterization is piecewise polynomial with minimum control effort, built from the state sequence q and time allocation T; collision avoidance uses GJK distance computation and an ESDF field; the dynamics for movable objects are $M\ddot{C}_x - J^T\lambda + f_D = 0$; and the augmented Lagrangian method, using L-BFGS with a Lewis-Overton line search, solves the resulting optimization problem with complementarity constraints.

What would settle it

Run the same planned pushing trajectory on floors with different friction, such as smooth tile versus carpet, or with objects of different mass, and measure the pushed object's path; if the deviation from the planned object path scales with ground friction or mass and exceeds the mean 0.13-0.20 meter tracking errors reported in Table II, the no-ground-friction quasi-static premise is falsified.

Watch

Extended reading notes

Core claim

The paper tries to establish that a robot's useful workspace expands when the planner is permitted to make contact with movable objects, and that such contact can be made predictable enough to plan as part of an optimization. It encodes contact as complementarity: for any robot-object pair, the relative speed and the contact force cannot both be nonzero, so the agents either separate or push while the contact sticks. This constraint is embedded in an ALM-based trajectory optimization over polynomial trajectories, whose decision variables are waypoints plus segment times. The reported results are that NAMO success rises from a 54.3% baseline average to 95%, RAMO tasks succeed 95-100% with back-end optimization times on the order of seconds, and experiments with a real omnidirectional robot pushing cylinders and cubes produce feasible trajectories. The authors claim this demonstrates a general contact-aware planning paradigm rather than a task-specific controller.

Load-bearing premise

The contact model assumes the friction between the robot and a pushed object is much stronger than the friction between the object and the ground, so the ground's grip can be ignored during low-speed pushing.

Editorial extensions

If this is right

  • In scenes with movable objects, a robot can navigate through spaces that collision-free planners treat as blocked, because the planner can choose to push an object aside instead of routing around it.
  • NAMO success rate rises as the number of movable objects grows, with success exceeding 90% when more than two movable objects are present, suggesting the method converts clutter into usable degrees of freedom.
  • RAMO tasks with cylinder and cube objects can be planned with back-end times of roughly 3 to 4 seconds and 95-100% simulation success, enabling rearrangement rather than only avoidance.
  • The same framework can be customized by adding objective terms, such as a preference for the pushed object's final pose: setting that weight to one makes the robot leave the object displaced instead of pushing it back.
  • Long-distance, long-duration pushing trajectories are executable in the real world, with movable-object position tracking errors averaging 0.13 to 0.20 meters and maxima up to 0.51 meters.

Reading between the lines

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

  • Inference: the paper's contribution is mostly at the trajectory level; task-level contact decisions still come from a front-end search on a masked map, so coupling CAMP with learned or semantic object-choice policies is a natural next step.
  • Inference: because the contact model ignores ground friction under a quasi-static assumption, the method should degrade on high-friction floors or with heavy objects, and adding ground-friction estimation or closed-loop correction is a testable extension.
  • Inference: the complementarity-plus-ALM formulation is not restricted to mobile robots; it could transfer to manipulation arms or legged robots that also plan through intentional contact, provided the same convex-shape collision geometry is available.
  • Inference: the reported object tracking errors, with maxima near half a meter, suggest that the planned contact model is only approximately correct; fusing force or tactile sensing during execution would be a direct way to close that gap.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

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 manuscript proposes CAMP, an optimization-based contact-aware motion planner for mobile robots operating among movable objects. Contact between the robot and movable objects is encoded as complementarity constraints (Section V-B5), and the resulting nonconvex trajectory optimization problem is solved with an augmented Lagrangian method (Section V-C), initialized by front-end path searches for NAMO and RAMO tasks (Section IV). The method is evaluated in randomized NAMO and RAMO simulations against the GCOPTER baseline (Table I) and in real-world experiments with cylinder and cube objects, with tracking errors reported in Table II. The paper claims that CAMP expands the robot's reachable space, substantially improves task success rates, and produces physically feasible contact trajectories.

Significance. If the modeling gaps identified below are addressed, CAMP would be a valuable step toward treating deliberate contact with movable objects as a first-class citizen in mobile robot trajectory optimization. The ALM formulation is a sensible choice for complementarity-constrained programs, the NAMO comparison is informative, and the real-world experiments provide a useful sanity check. The paper also states an intention to open-source the code, which would help reproducibility. However, the contact dynamics and complementarity model are under-specified, the ground-friction approximation in Appendix A undermines the physical feasibility claim as stated, and the RAMO baseline and success-rate definitions are missing. These issues are load-bearing for the central claims, so the paper needs a revision before it can be accepted.

major comments (4)
  1. [Sec. V-A / V-B4] The decision variables in Eq. (6) are listed as x = [p, o_1, ..., o_N, t_1, ..., t_M], but the dynamics constraint h_dyn in Section V-B4 contains lambda, the contact impulse force magnitudes, and Section V-B5 uses f_C in the contact constraints without giving f_C as an explicit function of C_x and t. The Jacobian J is only described as being determined during collision detection. As written, h_dyn and h_comp are not evaluable functions of the decision variables, so the optimizer cannot enforce them. Please add lambda (or an explicit contact-force law such as a penalty model or friction-cone inequality) to the problem formulation, or state precisely how lambda and f_C are computed from C_x. This is load-bearing because the paper's feasibility claim rests on these constraints.
  2. [Appendix A / Sec. V-B4] Appendix A states that, under the quasi-static assumption, "we can ignore the frictional force between the objects and the ground." This is not the pusher-slider quasistatic approximation used in the cited literature: in [31] and [32], the limit surface is precisely the mapping from the contact force required to overcome ground friction to the resulting object velocity. If ground friction is omitted from h_dyn, then a constant-velocity plan with zero contact force satisfies h_dyn (with air drag neglected) and h_comp (zero force), so the planned object motion is not anchored to a physically realizable push. The movable-object tracking errors in Table II (means 0.13-0.20 m, maxima up to 0.51 m) are consistent with this model mismatch. Please either include a ground-friction term (e.g., a Coulomb friction cone or limit surface) in h_dyn, or restrict and validate the approximation by reporting object mass, floor material, and friction coefficients and by showing that the tracking errors remain within a task-specific tolerance.
  3. [Table I / Abstract] Table I reports no GCOPTER baseline for the RAMO tasks (the RAMO rows show "----"), but the abstract claims that CAMP yields "a significant improvement in the success rate of two types of fundamental tasks." The success-rate improvement is therefore supported only for NAMO. In addition, "success rate" is never defined: the reader cannot tell whether it means planner convergence, constraint satisfaction at sampled points, or physical task completion. Please define the metric and provide the missing RAMO baseline comparison, or restrict the claim to NAMO. Please also report what constitutes a failed trial (e.g., timeout, constraint violation, or execution failure) and where failures occur.
  4. [Sec. V-B5] The inequality constraints g_i,j = -||dot-O_i - dot-O_j|| <= 0 and g_k = -||f_C(\ddot O_k)|| <= 0 are automatically satisfied by every trajectory because norms are nonnegative; the actual contact condition is carried entirely by the equality h_comp. However, the equality uses only magnitudes, so it cannot distinguish normal velocity from tangential sliding velocity or normal force from friction force. This loses the directional information needed for sliding contact, which is central to pushing an object along the floor. Please replace the norm-based complementarity by a component-wise formulation (normal gap/velocity versus normal force, and tangential velocity versus friction force) or state and justify the modeling assumptions under which the magnitude version remains sufficient for the NAMO and RAMO scenarios.
minor comments (4)
  1. [Sec. V-A] The notation s = n*(N+1) does not match the displayed decision vector x = [p, o_1, ..., o_N, t_1, ..., t_M]; please clarify the intended dimensions.
  2. [Table I] The RAMO rows leave the GCOPTER entries as "----" without explanation; please add a caption note or state clearly in the text that no RAMO baseline was run.
  3. [Abstract / Contributions] The paper says the code will be open-sourced, but no repository URL or availability statement is provided; please include one if the code is available at the time of the revised submission.
  4. [Sec. VI-A] The reported success rate increases as the number of movable objects grows (Table I, NAMO rows, 75% to 100%), which is counterintuitive and not explained; a short discussion of why more movable objects make the task easier for CAMP would help the reader interpret the comparison.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the paper's planning claims are empirically validated and its cited prior components do not reduce the result to its inputs.

full rationale

The paper's central claims are empirical rather than derivational: success rates in Table I and tracking errors in Table II compare CAMP with a baseline and with real-world execution. The optimization pipeline uses MINCO trajectories from the authors' prior work [1] as a parameterization and GCOPTER [1] as a baseline, but neither is used to define the success metric or to assert the feasibility conclusion; MINCO is a published, code-reproduced component and GCOPTER is only a comparison method. The complementarity constraints and ALM solver are supported by external citations [7, 11, 39], not by a self-citation chain. Appendix A's quasi-static assumption that ground friction can be ignored is a physical modeling assumption with stated justification from prior pusher-slider work [32]; it may be inaccurate, and Table II's movable-object tracking errors may reflect that inaccuracy, but an inaccurate assumption is not a circular one unless the conclusion is defined in terms of the assumption. No equation in the paper predicts a quantity that was used to fit a parameter, and no target result appears as an input to the optimization. The derivation chain is therefore self-contained with respect to circularity concerns.

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

The central claim rests on standard numerical optimization tools (MINCO, ALM, L-BFGS) plus domain assumptions about quasi-static pushing and convex geometry. No new physical entity is introduced. The main free parameters are objective weights, penalty update factors, convergence thresholds, and state bounds, none of which are reported with concrete fitted values.

free parameters (6)
  • Objective weight matrix W
    User-chosen diagonal weights balancing smoothness and time in the cost, Section V-A, Eq. (6).
  • Time penalty coefficient alpha
    Coefficient on total time in the objective; set by hand with no stated tuning rule.
  • Contact cost coefficient beta = 0 or 1
    In Section VI-C, beta is set to 0 and 1 to show trajectory customization; a user-chosen objective weight.
  • ALM penalty update factor gamma
    Penalty parameter rho is updated as rho = gamma * rho with gamma > 1, Section V-C; the actual value is not reported.
  • Convergence thresholds epsilon_opt, epsilon_feas
    User-defined feasibility and optimality thresholds in Section III-B; values are not reported.
  • State bounds vmax, amax
    Velocity and acceleration bounds in Section V-B3; chosen physical limits, not fitted, but needed for feasibility.
assumptions (6)
  • standard math MINCO trajectory generation theorem (Theorem 2 in [1])
    The back-end represents each segment as a minimum-energy polynomial using MINCO [1]; the paper relies on this result without deriving it.
  • standard math ALM convergence for problems with complementarity constraints
    Section V-C asserts that ALM does not require LICQ and converges, citing [11, 36, 38].
  • domain assumption Quasi-static assumption that friction between robot and movable object dominates, so ground friction can be ignored
    Appendix A: this assumption underlies the contact force and complementarity model; if false, planned object trajectories may not be executable.
  • domain assumption Coulomb friction and limit surface model from pusher-slider literature [31, 32]
    Section II-B and Appendix A use these to characterize contact direction and stick or slip modes.
  • domain assumption Robot and objects represented as circles or spheres and convex hulls, with GJK collision handling
    Section V-B2 restricts the geometry to convex shapes, which limits applicability to non-convex objects.
  • domain assumption Point contact with force direction determined by geometry during collision detection
    Section V-B4 states contact point and normal are fixed during detection, simplifying but also limiting possible contact modes.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Contact-Aware Motion Planning Among Movable Objects." pith.science (2026). https://pith.science/paper/BXHWSRUS

@misc{pith2026250203317,
  author       = {Pith},
  title        = {Pith review of: Contact-Aware Motion Planning Among Movable Objects},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/BXHWSRUS}},
  note         = {Machine review of arXiv:2502.03317}
}
read the original abstract

Most existing methods for motion planning of mobile robots involve generating collision-free trajectories. However, these methods focusing solely on contact avoidance may limit the robots' locomotion and can not be applied to tasks where contact is inevitable or intentional. To address these issues, we propose a novel contact-aware motion planning (CAMP) paradigm for robotic systems. Our approach incorporates contact between robots and movable objects as complementarity constraints in optimization-based trajectory planning. By leveraging augmented Lagrangian methods (ALMs), we efficiently solve the optimization problem with complementarity constraints, producing spatial-temporal optimal trajectories of the robots. Simulations demonstrate that, compared to the state-of-the-art method, our proposed CAMP method expands the reachable space of mobile robots, resulting in a significant improvement in the success rate of two types of fundamental tasks: navigation among movable objects (NAMO) and rearrangement of movable objects (RAMO). Real-world experiments show that the trajectories generated by our proposed method are feasible and quickly deployed in different tasks.

Figures

Figures reproduced from arXiv: 2502.03317 by the authors.

Figure 1
Figure 1. Compare to contact-avoidance method, contact-aware [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. The overview of the CAMP framework. The figure illustrates the workflow of CAMP. The perception module of the [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. The figure illustrates front-end computations in CAMP [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: The figure of the feasibility constraints. The two [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: The omnidirectional mobile robot used in our real [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 6
Figure 6. Figure 6: Experiments demonstrations. The first and second rows show a series of snapshots, illustrating the robot’s execution of [PITH_FULL_IMAGE:figures/full_fig_p009_6.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

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

  1. LLM-Land: Large Language Models for Context-Aware Drone Landing

    cs.RO 2025-05 conditional novelty 5.0 of 10

    A drone landing system that uses BLIP captions, a RAG-grounded lightweight LLM, and MPC raises simulated landing success against dynamic obstacles from 34% to 96% in open-field trials.

Reference graph

Works this paper leans on

45 extracted references · 36 canonical work pages · cited by 1 Pith paper

  1. [31]

    Planar slid- ing with dry friction part 1. limit surface and moment function,

    S. Goyal, A. Ruina, and J. Papadopoulos, “Planar slid- ing with dry friction part 1. limit surface and moment function,” Wear, vol. 143, no. 2, pp. 307–330, 1991

  2. [32]

    Feedback control of the pusher-slider system: A story of hybrid and under- actuated contact dynamics,

    F. R. Hogan and A. Rodriguez, “Feedback control of the pusher-slider system: A story of hybrid and under- actuated contact dynamics,” in Algorithmic F oundations of Robotics XII: Proceedings of the Twelfth Workshop on the Algorithmic F oundations of Robotics . Springer, 2020, pp. 800–815

  3. [1]

    Geometrically constrained trajectory optimization for multicopters,

    Z. Wang, X. Zhou, C. Xu, and F. Gao, “Geometrically constrained trajectory optimization for multicopters,” IEEE Transactions on Robotics , vol. 38, no. 5, pp. 3259– 3278, 2022

  4. [2]

    Motion planning around obstacles with convex optimization,

    T. Marcucci, M. Petersen, D. von Wrangel, and R. Tedrake, “Motion planning around obstacles with convex optimization,” Science Robotics , vol. 8, no. 84, p. eadf7843, 2023. [Online]. Available: https://www. science.org/doi/abs/10.1126/scirobotics.adf7843

  5. [3]

    Faster: Fast and safe trajectory planner for navigation in unknown environments,

    J. Tordesillas, B. T. Lopez, M. Everett, and J. P. How, “Faster: Fast and safe trajectory planner for navigation in unknown environments,” IEEE Transactions on Robotics, vol. 38, no. 2, pp. 922–938, 2022

  6. [4]

    Robust and efficient quadrotor trajectory generation for fast au- tonomous flight,

    B. Zhou, F. Gao, L. Wang, C. Liu, and S. Shen, “Robust and efficient quadrotor trajectory generation for fast au- tonomous flight,” IEEE Robotics and Automation Letters , vol. 4, no. 4, pp. 3529–3536, 2019

  7. [5]

    Navigation among mov- able obstacles: Real-time reasoning in complex envi- ronments,

    M. Stilman and J. J. Kuffner, “Navigation among mov- able obstacles: Real-time reasoning in complex envi- ronments,” International Journal of Humanoid Robotics , vol. 2, no. 04, pp. 479–503, 2005

  8. [6]

    Non-smooth trajectory optimization for wheeled balancing robots with contact switches and im- pacts,

    V . Klemm, Y . de Viragh, D. Rohr, R. Siegwart, and M. Tognon, “Non-smooth trajectory optimization for wheeled balancing robots with contact switches and im- pacts,” IEEE Transactions on Robotics , pp. 1–20, 2023

Show all 45 references
  1. [7]

    A direct method for trajectory optimization of rigid bodies through contact,

    M. Posa, C. Cantu, and R. Tedrake, “A direct method for trajectory optimization of rigid bodies through contact,” The International Journal of Robotics Research, vol. 33, no. 1, pp. 69–81, 2014. [Online]. Available: https://doi.org/10.1177/0278364913506757

  2. [8]

    Multiplier and gradient methods,

    M. R. Hestenes, “Multiplier and gradient methods,” Journal of Optimization Theory and Applications , vol. 4, 10 no. 5, pp. 303–320, 1969

  3. [9]

    A method for nonlinear constraints in min- imization problems,

    M. J. Powell, “A method for nonlinear constraints in min- imization problems,” Optimization, pp. 283–298, 1969

  4. [10]

    Augmented lagrange multiplier func- tions and duality in nonconvex programming,

    R. T. Rockafellar, “Augmented lagrange multiplier func- tions and duality in nonconvex programming,” SIAM Journal on Control , vol. 12, no. 2, pp. 268–285, 1974

  5. [11]

    E. G. Birgin and J. M. Mart ´ınez, Practical Augmented Lagrangian Methods for Constrained Optimization . So- ciety for Industrial and Applied Mathematics, 2014

  6. [12]

    Planning for manipula- tion among movable objects: Deciding which objects go where, in what order, and how,

    D. Saxena and M. Likhachev, “Planning for manipula- tion among movable objects: Deciding which objects go where, in what order, and how,” 2023

  7. [13]

    Planning for complex non-prehensile manipulation among movable objects by interleaving multi-agent pathfinding and physics-based simulation,

    D. M. Saxena and M. Likhachev, “Planning for complex non-prehensile manipulation among movable objects by interleaving multi-agent pathfinding and physics-based simulation,” in 2023 IEEE International Conference on Robotics and Automation (ICRA) . IEEE,

  8. [14]

    Hierarchical de- cision theoretic planning for navigation among movable obstacles,

    M. Levihn, J. Scholz, and M. Stilman, “Hierarchical de- cision theoretic planning for navigation among movable obstacles,” in Algorithmic F oundations of Robotics X: Proceedings of the Tenth Workshop on the Algorithmic F oundations of Robotics. Springer, 2013, pp. 19–35

  9. [15]

    Tedrake, Underactuated Robotics , 2023

    R. Tedrake, Underactuated Robotics , 2023. [Online]. Available: https://underactuated.csail.mit.edu

  10. [16]

    Simultaneous learning of contact and continuous dynamics,

    B. Bianchini, M. Halm, and M. Posa, “Simultaneous learning of contact and continuous dynamics,” in Proceedings of The 7th Conference on Robot Learning , ser. Proceedings of Machine Learning Research, J. Tan, M. Toussaint, and K. Darvish, Eds., vol. 229. PMLR, 06–09 Nov 2023, p...

  11. [17]

    Active extrinsic contact sensing: Application to general peg-in-hole insertion,

    S. Kim and A. Rodriguez, “Active extrinsic contact sensing: Application to general peg-in-hole insertion,” in 2022 International Conference on Robotics and Automa- tion (ICRA) , 2022, pp. 10 241–10 247

  12. [18]

    Tac2pose: Tactile object pose estimation from the first touch,

    M. Bauza, A. Bronars, and A. Rodriguez, “Tac2pose: Tactile object pose estimation from the first touch,” The International Journal of Robotics Research , vol. 42, no. 13, pp. 1185–1209, 2023. [Online]. Available: https://doi.org/10.1177/02783649231196925

  13. [19]

    Direct trajectory opti- mization using nonlinear programming and collocation,

    C. R. Hargraves and S. W. Paris, “Direct trajectory opti- mization using nonlinear programming and collocation,” Journal of guidance, control, and dynamics , vol. 10, no. 4, pp. 338–342, 1987

  14. [20]

    An introduction to trajectory optimization: How to do your own direct collocation,

    M. Kelly, “An introduction to trajectory optimization: How to do your own direct collocation,” SIAM Review , vol. 59, no. 4, pp. 849–904, 2017. [Online]. Available: https://doi.org/10.1137/16M1062569

  15. [21]

    Survey of numerical methods for trajectory optimization,

    J. T. Betts, “Survey of numerical methods for trajectory optimization,” Journal of guidance, control, and dynam- ics, vol. 21, no. 2, pp. 193–207, 1998

  16. [22]

    A fast procedure for computing the distance between complex objects in three-dimensional space,

    E. G. Gilbert, D. W. Johnson, and S. S. Keerthi, “A fast procedure for computing the distance between complex objects in three-dimensional space,” IEEE Journal on Robotics and Automation , vol. 4, no. 2, pp. 193–203, 1988

  17. [23]

    Optimization- based collision avoidance,

    X. Zhang, A. Liniger, and F. Borrelli, “Optimization- based collision avoidance,” IEEE Transactions on Con- trol Systems Technology , vol. 29, no. 3, pp. 972–983, 2021

  18. [24]

    Con- tinuous implicit sdf based any-shape robot trajectory op- timization,

    T. Zhang, J. Wang, C. Xu, A. Gao, and F. Gao, “Con- tinuous implicit sdf based any-shape robot trajectory op- timization,” in 2023 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) , 2023, pp. 282– 289

  19. [25]

    Planning dynamically feasible trajectories for quadrotors using safe flight cor- ridors in 3-d complex environments,

    S. Liu, M. Watterson, K. Mohta, K. Sun, S. Bhattacharya, C. J. Taylor, and V . Kumar, “Planning dynamically feasible trajectories for quadrotors using safe flight cor- ridors in 3-d complex environments,” IEEE Robotics and Automation Letters, vol. 2, no. 3, pp. 1688–1695, 2017

  20. [26]

    Reaching the limit in autonomous racing: Optimal control versus reinforcement learning,

    Y . Song, A. Romero, M. M ¨uller, V . Koltun, and D. Scaramuzza, “Reaching the limit in autonomous racing: Optimal control versus reinforcement learning,” Science Robotics , vol. 8, no. 82, p. eadg1462, 2023. [Online]. Available: https://www.science.org/doi/abs/10. 1126/scirob...

  21. [27]

    Champion-level drone racing using deep reinforcement learning,

    E. Kaufmann, L. Bauersfeld, A. Loquercio, M. M ¨uller, V . Koltun, and D. Scaramuzza, “Champion-level drone racing using deep reinforcement learning,” Nature, vol. 620, pp. 982–987, 2023. [Online]. Available: https://doi.org/10.1038/s41586-023-06419-4

  22. [28]

    A survey on coverage path planning for robotics,

    E. Galceran and M. Carreras, “A survey on coverage path planning for robotics,” Robotics and Autonomous Systems, vol. 61, no. 12, pp. 1258–1276, 2013. [Online]. Available: https://www.sciencedirect.com/science/article/ pii/S092188901300167X

  23. [29]

    Cooper- ative multi-robot box-pushing,

    M. Mataric, M. Nilsson, and K. Simsarin, “Cooper- ative multi-robot box-pushing,” in Proceedings 1995 IEEE/RSJ International Conference on Intelligent Robots and Systems. Human Robot Interaction and Cooperative Robots, vol. 3, 1995, pp. 556–561 vol.3

  24. [30]

    Realization of heavy object transportation by mobile robots using handcarts and outrigger,

    F. Ohashi, K. Kaminishi, J. D. Figueroa Heredia, H. Kato, T. Ogata, T. Hara, and J. Ota, “Realization of heavy object transportation by mobile robots using handcarts and outrigger,” Robomech journal, vol. 3, pp. 1–12, 2016

  25. [33]

    Trajectory generation and control of a quadrotor with a cable- suspended load-a differentially-flat hybrid system,

    K. Sreenath, N. Michael, and V . Kumar, “Trajectory generation and control of a quadrotor with a cable- suspended load-a differentially-flat hybrid system,” in 2013 IEEE international conference on robotics and automation. IEEE, 2013, pp. 4888–4895

  26. [34]

    A review of pseudospec- tral optimal control: From theory to flight,

    I. M. Ross and M. Karpenko, “A review of pseudospec- tral optimal control: From theory to flight,” Annual Reviews in Control , vol. 36, no. 2, pp. 182–197, 2012

  27. [35]

    Sur- vey of uav motion planning,

    L. Quan, L. Han, B. Zhou, S. Shen, and F. Gao, “Sur- vey of uav motion planning,” IET Cyber-systems and 11 Robotics, vol. 2, no. 1, pp. 14–21, 2020

  28. [36]

    Global convergence of augmented lagrangian methods applied to optimization problems with degenerate constraints, including problems with complementarity constraints,

    A. F. Izmailov, M. V . Solodov, and E. Uskov, “Global convergence of augmented lagrangian methods applied to optimization problems with degenerate constraints, including problems with complementarity constraints,” SIAM Journal on Optimization , vol. 22, no. 4, pp. 1579– 1606, 2012

  29. [37]

    Calipso: A differentiable solver for trajectory op- timization with conic and complementarity constraints,

    T. A. Howell, K. Tracy, S. Le Cleac’h, and Z. Manch- ester, “Calipso: A differentiable solver for trajectory op- timization with conic and complementarity constraints,” in The International Symposium of Robotics Research . Springer, 2022, pp. 504–521

  30. [38]

    Local convergence of exact and inexact augmented lagrangian methods under the second-order sufficient optimality condition,

    D. Fern ´andez and M. V . Solodov, “Local convergence of exact and inexact augmented lagrangian methods under the second-order sufficient optimality condition,” SIAM Journal on Optimization , vol. 22, no. 2, pp. 384–407, 2012

  31. [39]

    Contact and friction simulation for computer graphics,

    S. Andrews, K. Erleben, and Z. Ferguson, “Contact and friction simulation for computer graphics,” in ACM SIGGRAPH 2022 Courses , ser. SIGGRAPH ’22. New York, NY , USA: Association for Computing Machinery, 2022. [Online]. Available: https://doi.org/10. 1145/3532720.3535640

  32. [40]

    On the limited memory bfgs method for large scale optimization,

    D. C. Liu and J. Nocedal, “On the limited memory bfgs method for large scale optimization,” Mathematical programming, vol. 45, no. 1-3, pp. 503–528, 1989

  33. [41]

    Nonsmooth optimiza- tion via quasi-newton methods,

    A. S. Lewis and M. L. Overton, “Nonsmooth optimiza- tion via quasi-newton methods,” Mathematical Program- ming, vol. 141, pp. 135–163, 2013

  34. [42]

    D. P. Bertsekas, Constrained optimization and Lagrange multiplier methods . Academic press, 2014

  35. [43]

    Nocedal and S

    J. Nocedal and S. J. Wright, Numerical optimization . Springer, 1999

  36. [44]

    Proximity queries for absolutely continuous parametric curves,

    A. Lakshmanan, A. Patterson, V . Cichella, and N. Ho- vakimyan, “Proximity queries for absolutely continuous parametric curves,” in Proceedings of Robotics: Science and Systems, Freiburg im Breisgau, Germany, June 2019

  37. [2023]

    Available: http://dx.doi.org/10.1109/ ICRA48891.2023.10161006

    [Online]. Available: http://dx.doi.org/10.1109/ ICRA48891.2023.10161006

Pith tools

Reviewed August 9, 2026 · model on record in the stance chip above.