Humanoid Loco-manipulation Planning based on Graph Search and Reachability Maps

Abderrahmane Kheddar; Fumio Kanehiro; Iori Kumagai; Masaki Murooka; Mitsuharu Morisawa

REVIEW 3 major objections 5 minor 33 references

Humanoid Loco-manipulation Planning based on Graph Search and Reachability Maps

T0 review · 3 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read A graph search plans a humanoid's steps, grasps, and rolling as one problem.

desk verdict Genuinely new integration of footstep and regrasp planning via switchable reachability maps, with a rolling-bobbin first, but validation is thinner than the claims. read the letter →

arxiv 2505.23505 v1 pith:FIGDLTGF submitted 2025-05-29 cs.RO

Masaki Murooka , Iori Kumagai , Mitsuharu Morisawa , Fumio Kanehiro , Abderrahmane Kheddar This is my paper

classification cs.RO

keywords loco-manipulationplanninghumanoidrobotsgraphsearchreachabilitymapregraspingobjectrollingfootstepanytimedynamicA*

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 proposes a planning framework that lets a humanoid robot decide, in one shot, where to step and which hand to grasp with while moving a large object. The central claim is that loco-manipulation can be written as a graph search over states that include both feet, the object pose, and the grasping hand, with transition feasibility checked by precomputed reachability maps. The maps are relocated and switched as the object rolls, so the moving grasp point is accounted for without running inverse kinematics during search. The authors demonstrate the planner on a bobbin-rolling task, a door-opening task, and a cart-pushing task, and report that it automatically generates a rolling-with-regrasping motion for the bobbin.

What carries the argument

The load-bearing mechanism is the transition model with switchable reachability maps. Its movable condition checks the object pose at the middle timing of a step against a map anchored at the stance foot, and the end pose against a map anchored at the midpoint of the two feet; for rolling objects, maps are switched according to the distance rolled since the last regrasp, so the changing grasp point is handled by map selection instead of inverse-kinematics calls during the search.

What would settle it

Take a planned bobbin-rolling sequence and execute it on a real or high-fidelity simulated humanoid while recording the actual CoM and object pose through each step; if the hand loses the grasp or the reachability condition fails at some intermediate time even though the middle and end poses passed, the transition model's most-severe-case assumption is false.

Watch

Extended reading notes

Core claim

The paper claims that a humanoid's loco-manipulation planning can be posed as a graph path search over states describing the stance foot, swing foot, object pose, and grasping hand, where each transition is accepted only if a reachability map certifies that the hand can hold the object at the middle and end of the footstep. For rolling objects, several reachability maps indexed by the distance rolled since the last regrasp are switched during search, capturing the motion of the grasp point without expensive kinematic recomputation. The authors report that this yields the first automatically planned motion of a large cylindrical object moved by rolling with regrasping.

Load-bearing premise

The feasibility check assumes the robot's center of mass passes through the stance foot at the middle of each step and the object moves to the middle pose exactly in sync, so checking only the middle and end poses is enough to guarantee reachability throughout the transition.

Editorial extensions

If this is right

A single planner covers pure locomotion, pure manipulation, and combined stepping-while-moving-object behaviors, because the transition set includes no-stepping and zero-object-progress actions.
Regrasping emerges automatically when obstacles make it cheaper, since hand-label transitions are part of the search state, as shown in the door-opening case.
Rolling objects are handled in the same framework as sliding or fixed-grasp objects by switching reachability maps according to the distance rolled since the last regrasp.
An initial solution is found in roughly 100 milliseconds for the tested tasks, allowing the robot to keep moving while the anytime search improves the plan.

Reading between the lines

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

The same switchable-map idea should extend to any object motion whose grasp point moves predictably, such as sliding, pivoting, or objects carried on a tray, without changing the search formulation.
Because the reachability maps are precomputed in posture space, the planner's runtime depends mostly on the graph search itself rather than on the complexity of the robot model.
The cost model's constant step and regrasp penalties could be replaced with task-specific costs such as energy, time, or stability margin without altering the transition model.
The assumption that checking only the middle and end poses guarantees reachability could be tested offline by sampling dense CoM and object trajectories from the generated whole-body motion, giving a quantitative safety margin for execution.

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.

Desk Editor's Note

read the letter

The paper's real contribution is the transition model: a graph-search state that couples stance/swing feet, hand labels, and object pose, with reachability maps relocated and switched as the object moves. That integration is new, and the rolling-with-regrasping result looks like a genuine first. The planning times are impressive — initial solutions in ~100 ms, final solutions in 1–2 s for fairly complex scenes — and the three demos (bobbin rolling, door opening, cart pushing) show the framework is versatile. The writing is clear and the equations are stated carefully. Credit where due: this is a solid, useful piece for the humanoid loco-manipulation subfield.

The weak spot is the movable-condition check in Eq. (7). It samples two instants — the middle and end of a footstep transition — under an assumed CoM trajectory and synchronous object motion. No argument shows this is conservative or sufficient; reachability maps are not convex, so checking the 'most severe' CoM pose does not bound intermediate poses without a monotonicity proof. For rolling objects, the maps are switched in 5-degree increments and the grasping point migrates, so the midpoint check is especially fragile. The paper explicitly acknowledges the assumption, which is honest, but it never validates it against a full WBM trajectory or a real robot. The one ZMP trace in Fig. 15 is suggestive, not systematic.

Also, the comparison with Jorgensen et al. is qualitative — they report their own times but don't run the same benchmarks on the prior method. And there's no sensitivity analysis for the grid size, step/regrasp costs, or the nominal foot pose heuristic. These are minor-to-moderate issues, not fatal ones. The framework is a planning heuristic; the assumption in Eq. (7) is a clear limitation that could be patched with post-hoc verification or a conservative construction.

Who should read this: anyone working on humanoid loco-manipulation or footstep planning with manipulation constraints. It deserves a serious referee. A referee should ask for empirical validation of the movable condition on several random scenarios (comparing planned vs. full-WBM execution) and a direct timing comparison on shared benchmarks. The core idea holds up and the paper is worth engaging with.

Referee Report

3 major / 5 minor

Summary. The paper proposes a three-stage planning framework for humanoid loco-manipulation: object path planning using RRT*, footstep and regrasping planning using AD* graph search with precomputed reachability maps, and whole-body motion planning using QP-based inverse kinematics. The main contribution is a transition model that represents combined locomotion and manipulation primitives, with feasibility evaluated by relocating and switching reachability maps according to the motion of the robot and the object. The framework is demonstrated in simulation on bobbin rolling with regrasping, door opening, and cart pushing tasks, with reported computation times of the order of seconds.

Significance. If the movability filter is reliable, the framework is a practically useful contribution to humanoid loco-manipulation planning: it is clearly formulated, modular, and computationally efficient, with initial solutions obtained in roughly 100 ms and the ability to handle rolling objects with automatic regrasping. The strengths include an explicit graph-search formulation, precomputed reachability maps that avoid repeated inverse kinematics, and a comparison of planning times against a prior method. However, the central claim rests on an assumption about CoM and object trajectories that is not validated against whole-body motion planning or hardware, which limits the strength of the feasibility claims. The paper also does not provide a formal correctness argument for the transition filter, so the current evidence is not sufficient to fully establish the purported versatility.

major comments (3)

[Section IV-C3, Eq. (7)] The movable condition F_movable is evaluated at only two discrete instants, under the assumption that the CoM is on the destination stance foot at the middle timing and at the mid-foot pose at the end, while the object moves synchronously to the path midpoint. No argument is provided that these two samples bound the hardest reachability condition over the continuous CoM and object trajectories. Reachability maps are not convex in CoM position or object pose, so checking the "most severe" CoM pose does not in general imply reachability for all intermediate CoM positions. The actual CoM from preview control (Section V) is not constrained to pass through the stance foot at the middle timing, and the object is not forced to move synchronously to the midpoint. For rolling objects, the reachability map is switched only at finite rolling-angle increments (5 degrees), leaving intermediate angles unchecked. Consequently, the planner may accept transitions that are not executable by the WBM layer. The paper explicitly acknowledges the assumption in Fig. 6(B), but neither proves its conservatism nor validates it against full whole-body trajectories or hardware experiments. This issue is load-bearing because it directly affects the feasibility of the planned footstep and regrasp sequences, including the claimed rolling-with-regrasping result.
[Section IV-B5, Eq. (5)] The heuristic h_nominal is inadmissible, as the paper itself states ("Without h_nominal, the heuristic is admissible"). Since AD* relies on an admissible heuristic for its suboptimality guarantees, the reported "heuristics inflation factor" in Table II does not provide a rigorous bound on solution quality when h_nominal is used. The authors should either provide an analysis of the bound with h_nominal or explicitly state that the anytime optimality guarantee is heuristic only, so that readers are not misled about the formal properties of the planner.
[Section VI-B4 and Section V] The only dynamics validation of the planned motions is a ZMP check on a forward-rolling motion. ZMP staying within the support region does not guarantee that the hand can reach the object at all intermediate times, which is precisely what the movable condition (7) approximates. The paper should report whether the QP-based WBM planner actually finds feasible whole-body trajectories for the planned footstep and object sequences, and ideally execute the motions on a real robot, to substantiate the executability of the planned transitions. Without this validation, the experimental section is insufficient to support the claim that the rolling-with-regrasping motion is automatically planned and feasible.

minor comments (5)

[Section IV-C3, Eq. (7)] The definition of c_mid_obj as c_P[(idx(c_obj[k]) + idx(c_obj[k+1]))/2] is ambiguous when the sum of the two indices is odd; please clarify whether the midpoint pose is obtained by interpolation, rounding, or some other rule.
[Section IV-C1] The reachability map grid resolution is reported as 100 mm and 10 degrees in Section VI-B1, but the paper does not specify how an arbitrary object pose that does not coincide with a grid node is evaluated (e.g., nearest neighbor, bilinear interpolation). Please describe the evaluation rule.
[Figure 14(B)] The horizontal axis of Fig. 14(B) is labeled "time [s]", but the figure shows the length of the footstep sequence over time; please clarify whether this is computation time and what the unit refers to.
[Section VI-B2] The use of the Halton sequence to generate the footstep action set is interesting; please state the dimensionality of the sampled space and the number of samples used for the nominal action set, since these affect the planner's behavior.
[Introduction, Contributions] The claim of being the first study to automatically plan rolling operation with regrasping could be tempered, since the experiments are simulations and prior work on non-prehensile manipulation may cover related tasks; a more careful comparison to existing rolling-manipulation literature would strengthen the novelty statement.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: reachability maps are precomputed and task-independent; the Eq. (7) assumption is an acknowledged approximation, not a fitted input.

full rationale

The paper's derivation chain is not circular. OP-planning produces an object path using RRT*, FR-planning searches a graph whose transition feasibility is filtered by precomputed reachability maps, and WBM-planning solves a QP-based inverse kinematics problem. The reachability maps are computed offline from the specific robot models (HRP-5P, HRP-2Kai, HRP-4) with fixed grid sizes of 100 mm and 10 degrees, and for rolling objects in 5-degree rolling-angle increments; they are not fitted to the search results and do not encode the target task. The movable-condition check in Eq. (7) is the only point where the paper replaces a continuous CoM/object trajectory with an assumed worst-case mid-transition pose, but the paper explicitly states this assumption: 'we assume the CoM is on the stance foot at the middle timing of the transition... the object is considered to be in the middle pose.' This is a stated heuristic approximation, not a derived-from-inputs prediction. Using the same IK solver for map generation and WBM-planning is a consistency choice, not a fitted-parameter loop. The ZMP check is a consistency validation of the generated motion, not a result that is forced by the input planning parameters. Self-citations ([1], [2], [26], [29], [32]) support specific tools, prior robot models, or earlier applications, but the framework's key novelty—the switching reachability-map transition model—does not rest on any self-cited uniqueness theorem or on an ansatz smuggled in via citation. The acknowledged potential failure of Eq. (7) to guarantee continuous reachability is a correctness and robustness concern, not circularity.

Assumptions & free parameters 5 free parameters · 3 assumptions · 0 invented entities

The framework uses standard planning and optimization components (RRT*, AD*, QP-based IK) from the literature. Its specific assumptions are the approximate CoM/object trajectory model behind the movable condition and the reliance on precomputed reachability maps as a feasibility oracle.

free parameters (5)

Reachability map grid size = 100 mm, 10 degrees
Resolution chosen by authors; affects the granularity of reachability tests and planning time.
Nominal foot pose distance (h_nominal) = 1.2 m
Hand-tuned heuristic distance of nominal foot pose behind the object to guide the search; speeds up initial solution.
Cost constants c_step, c_regrasp = not specified
Constant costs for stepping and regrasping determine the trade-off; values not given but affect planned motion.
Rolling angle range and increment = 0 to 45 degrees, 5 degree increments
Discretization of rolling angles for reachability map generation; chosen by authors.
N_obj incremental limit = not specified
Maximum object index increment per transition in graph search (Eq. 2f); controls branching factor.

assumptions (3)

domain assumption Reachability map membership implies whole-body kinematic feasibility for the relevant CoM frame and hand.
The whole approach relies on precomputed IK-based reachability maps being a sufficient and accurate model of what the robot can reach; the maps are generated with the same kinematics as WBM-planning.
domain assumption During a footstep transition, the CoM lies on the stance foot at the middle timing and the object moves synchronously, so checking the middle and end poses is sufficient.
Stated in Section IV-C3 and used to derive Eq. (7); not proven to be conservative for all walking motions.
domain assumption The object path from RRT* is discretizable into a sequence of poses that the robot can follow with the planned steps and grasps.
FR-planning uses only the discretized path points and moves the object monotonically along increasing indices.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Humanoid Loco-manipulation Planning based on Graph Search and Reachability Maps." pith.science (2026). https://pith.science/paper/FIGDLTGF

@misc{pith2026250523505,
  author       = {Pith},
  title        = {Pith review of: Humanoid Loco-manipulation Planning based on Graph Search and Reachability Maps},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/FIGDLTGF}},
  note         = {Machine review of arXiv:2505.23505}
}

read the original abstract

In this letter, we propose an efficient and highly versatile loco-manipulation planning for humanoid robots. Loco-manipulation planning is a key technological brick enabling humanoid robots to autonomously perform object transportation by manipulating them. We formulate planning of the alternation and sequencing of footsteps and grasps as a graph search problem with a new transition model that allows for a flexible representation of loco-manipulation. Our transition model is quickly evaluated by relocating and switching the reachability maps depending on the motion of both the robot and object. We evaluate our approach by applying it to loco-manipulation use-cases, such as a bobbin rolling operation with regrasping, where the motion is automatically planned by our framework.

Figures

Figures reproduced from arXiv: 2505.23505 by the authors.

**Figure 1.** Humanoid loco-manipulation motions. by applying this transition model to a sophisticated graph search algorithm. It is noteworthy that, to the best of our knowledge, this is the first study in which a humanoid robot automatically plans the motion of moving a large cylindrical object by the rolling operation with regrasping. A. Related Works 1) Large Object Manipulation: Various types of large object manipulation are… view at source ↗

**Figure 2.** Overview of the planning framework. The loco-manipulation planning framework consists of three processes: object path planning (OP-planning), footstep and regrasping planning (FR-planning), and whole-body motion planning (WBM-planning). and multi-contact motion has been proposed recently with the advantage of being able to handle motion without discretization [19], [20]. However, in this letter, graph search is use… view at source ↗

**Figure 4.** Problem settings of FR-planning. A robot is moving an object from the start position to the goal position. Green rectangle markers represent the footstep action set for the right foot; a blue line represents the object path; and markers with yellow to red gradation represent the reachability map. set A and the reachability map M are described in Sections VI-B and IV-C, respectively. B. Formulation as Graph Search Pr… view at source ↗

Figures from the paper (8 more)

**Figure 6.** Figure 6: Frame and trajectory of CoM (A) mid(cL-foot, cR-foot ) is the middle pose between the left and right foot poses. The position of CoM frame ccom coincides with the robot’s CoM, and its orientation coincides with the middle pose. (B) The red curve illustrates an example …

**Figure 7.** Figure 7: An example of reachability map. A reachability map for opening the door by grasping the doorknob with the left hand is shown. A reachability map is represented as a set of cells on a 3D grid consisting of the X and Y positions and the Z-angle (i.e., rotation angle arou…

**Figure 8.** Figure 8: Transition evaluation based on reachability map. [PITH_FULL_IMAGE:figures/full_fig_p005_8.png]

**Figure 9.** Figure 9: An example of reachability maps for a rolling object. [PITH_FULL_IMAGE:figures/full_fig_p006_9.png]

**Figure 10.** Figure 10: Tasks and constraints in WBM-planning. Whole-body IK consists of three types of equality tasks (boxed in red) and two types of inequality constraints (boxed in blue). Links checked for selfcollision are displayed in their green strictly convex hulls. generated by int…

**Figure 12.** Figure 12: Door opening task. In (A), HRP-2Kai humanoid opens the door with the left hand until the end, whereas in (B), the robot switches from left to right hand to avoid the obstacle visualized in blue [PITH_FULL_IMAGE:figures/full_fig_p007_12.png]

**Figure 13.** Figure 13: Cart pushing task. HRP-4 humanoid carries the cart by pushing it with both hands. B. Evaluation of Planning Results 1) Computation Time: Table II shows the detailed results of FR-planning in the examples described in Section VI-A. Compared with the state-of-the-art l…

**Figure 15.** Figure 15: ZMP trajectory of the planned motion. The planned motion is dynamically feasible as the ZMP is within the support region. VII. CONCLUSION In this letter, we proposed a versatile planning framework for loco-manipulation. In order to quickly and flexibly generate comple…

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

33 extracted references · 24 canonical work pages

[1]

Manipulation strategy decision and execution based on strategy proving operation for carrying large and heavy objects,

M. Murooka, S. Noda, S. Nozawa, Y . Kakiuchi, K. Okada, and M. Inaba, “Manipulation strategy decision and execution based on strategy proving operation for carrying large and heavy objects,” inIEEE International Conference on Robotics and Automation, 2014, pp. 3425–3432

2014
[2]

Perception based locomotion system for a humanoid robot with adaptive footstep compensation under task constraints,

I. Kumagai, M. Morisawa, S. Nakaoka, T. Sakaguchi, H. Kaminaga, K. Kaneko, and F. Kanehiro, “Perception based locomotion system for a humanoid robot with adaptive footstep compensation under task constraints,” inIEEE/RSJ International Conference on Intelligent Robots and Systems, 2018, pp. 713–719

work page 2018
[3]

Humanoid teleoperation for whole body manipulation,

M. Stilman, K. Nishiwaki, and S. Kagami, “Humanoid teleoperation for whole body manipulation,” inIEEE International Conference on Robotics and Automation, 2008, pp. 3175–3180

work page 2008
[4]

Pivoting based manipulation by a humanoid robot,

E. Yoshida, M. Poirier, J.-P. Laumond, O. Kanoun, F. Lamiraux, R. Alami, and K. Yokoi, “Pivoting based manipulation by a humanoid robot,”Autonomous Robots, vol. 28, pp. 77–88, 2009

2009
[5]

Hu- manoid full-body controller adapting constraints in structured objects through updating task-level reference force,

S. Nozawa, I. Kumagai, Y . Kakiuchi, K. Okada, and M. Inaba, “Hu- manoid full-body controller adapting constraints in structured objects through updating task-level reference force,” inIEEE/RSJ International Conference on Intelligent Robots and Systems, 2012, pp. 3417–3424

2012
[6]

Finding locomanipulation plans quickly in the locomotion constrained manifold,

S. J. Jorgensen, M. Vedantam, R. Gupta, H. Cappel, and L. Sentis, “Finding locomanipulation plans quickly in the locomotion constrained manifold,” inIEEE International Conference on Robotics and Automa- tion, 2020, pp. 6611–6617

work page 2020
[7]

Manipulation of documented objects by a walking humanoid robot,

S. Dalibard, A. Nakhaei, F. Lamiraux, and J. Laumond, “Manipulation of documented objects by a walking humanoid robot,” inIEEE-RAS International Conference on Humanoid Robots, 2010, pp. 518–523

work page 2010
[8]

Humanoid whole-body planning for loco-manipulation tasks,

P. Ferrari, M. Cognetti, and G. Oriolo, “Humanoid whole-body planning for loco-manipulation tasks,” inIEEE International Conference on Robotics and Automation, 2017, pp. 4741–4746

work page 2017

Show all 33 references

[9]

Motion primitive based random planning for loco-manipulation tasks,

A. Settimi, D. Caporale, P. Kryczka, M. Ferrati, and L. Pallottino, “Motion primitive based random planning for loco-manipulation tasks,” inIEEE-RAS International Conference on Humanoid Robots, 2016, pp. 1059–1066

2016
[10]

Randomized multi-modal motion planning for a humanoid robot manipulation task,

K. Hauser and V . Ng-Thow-Hing, “Randomized multi-modal motion planning for a humanoid robot manipulation task,”The International Journal of Robotics Research, vol. 30, no. 6, pp. 678–698, 2011

2011
[11]

A skill-based motion planning framework for humanoids,

M. Kallmann, Y . Huang, and R. Backman, “A skill-based motion planning framework for humanoids,” inIEEE International Conference on Robotics and Automation, 2010, pp. 2507–2514

2010
[12]

Humanoid motion planning for dual-arm manipulation and re-grasping tasks,

N. Vahrenkamp, D. Berenson, T. Asfour, J. Kuffner, and R. Dillmann, “Humanoid motion planning for dual-arm manipulation and re-grasping tasks,” inIEEE/RSJ International Conference on Intelligent Robots and Systems, 2009, pp. 2464–2470

2009
[13]

Planning biped navigation strategies in complex environments,

J. Chestnutt, J. Kuffner, K. Nishiwaki, and S. Kagami, “Planning biped navigation strategies in complex environments,” inIEEE-RAS International Conference on Humanoid Robots, 2003

2003
[14]

Biped footstep planning,

N. Perrin, “Biped footstep planning,” inHumanoid Robotics: A Refer- ence. Springer, 2018, pp. 1–21

2018
[15]

ARA*: Anytime A* with provable bounds on sub-optimality,

M. Likhachev, G. J. Gordon, and S. Thrun, “ARA*: Anytime A* with provable bounds on sub-optimality,” inAdvances in Neural Information Processing Systems 16, 2004, pp. 767–774

2004
[16]

Anytime search-based footstep planning with suboptimality bounds,

A. Hornung, A. Dornbush, M. Likhachev, and M. Bennewitz, “Anytime search-based footstep planning with suboptimality bounds,” inIEEE- RAS International Conference on Humanoid Robots, 2012, pp. 674–679

2012
[17]

Humanoid robot locomotion and manipulation step planning,

K. Bouyarmane and A. Kheddar, “Humanoid robot locomotion and manipulation step planning,”Advanced Robotics, vol. 26, no. 10, pp. 1099–1126, 2012

2012
[18]

An efficient acyclic contact planner for multiped robots,

S. Tonneau, A. Del Prete, J. Pettr ´e, C. Park, D. Manocha, and N. Mansard, “An efficient acyclic contact planner for multiped robots,” IEEE Transactions on Robotics, vol. 34, no. 3, pp. 586–601, 2018

2018
[19]

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

2014
[20]

Optimization-based locomotion planning, estimation, and control design for the atlas humanoid robot,

S. Kuindersma, R. Deits, M. Fallon, A. Valenzuela, H. Dai, F. Permenter, T. Koolen, P. Marion, and R. Tedrake, “Optimization-based locomotion planning, estimation, and control design for the atlas humanoid robot,” Autonomous Robots, vol. 40, no. 3, pp. 429–455, 2016

2016
[21]

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
[22]

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 J. Math., vol. 145, no. 2, pp. 367–393, 1990

1990
[23]

D* lite,

S. Koenig and M. Likhachev, “D* lite,” inEighteenth National Con- ference on Artificial Intelligence. American Association for Artificial Intelligence, 2002, pp. 476–483

2002
[24]

Anytime dynamic A*: An anytime, replanning algorithm,

M. Likhachev, D. Ferguson, G. Gordon, A. T. Stentz, and S. Thrun, “Anytime dynamic A*: An anytime, replanning algorithm,” inInterna- tional Conference on Automated Planning and Scheduling, 2005

2005
[25]

Accessed: 2020-9-30

SBPL. Accessed: 2020-9-30. [Online]. Available: https://github.com/ sbpl/sbpl

2020
[26]

Biped walking pattern generation by using preview control of zero-moment point,

S. Kajita, F. Kanehiro, K. Kaneko, K. Fujiwara, K. Harada, K. Yokoi, and H. Hirukawa, “Biped walking pattern generation by using preview control of zero-moment point,” inIEEE International Conference on Robotics and Automation, vol. 2, 2003, pp. 1620–1626

2003
[27]

Three-dimensional bipedal walking control based on divergent component of motion,

J. Englsberger, C. Ott, and A. Albu-Sch ¨affer, “Three-dimensional bipedal walking control based on divergent component of motion,”IEEE Transactions on Robotics, vol. 31, no. 2, pp. 355–368, 2015

2015
[28]

Kinematic control of redundant manipulators: Generalizing the task-priority framework to inequality task,

O. Kanoun, F. Lamiraux, and P. Wieber, “Kinematic control of redundant manipulators: Generalizing the task-priority framework to inequality task,”IEEE Transactions on Robotics, vol. 27, no. 4, pp. 785–792, 2011

2011
[29]

A strictly con- vex hull for computing proximity distances with continuous gradients,

A. Escande, S. Miossec, M. Benallegue, and A. Kheddar, “A strictly con- vex hull for computing proximity distances with continuous gradients,” IEEE Transactions on Robotics, vol. 30, no. 3, pp. 666–678, 2014

2014
[30]

ROS: an open-source Robot Operating Sys- tem,

M. Quigley, K. Conley, B. Gerkey, J. Faust, T. Foote, J. Leibs, R. Wheeler, and A. Ng, “ROS: an open-source Robot Operating Sys- tem,” inIEEE International Conference on Robotics and Automation Workshop on Open Source Software, 2009

2009
[31]

The Open Motion Planning Library,

I. A. S ¸ucan, M. Moll, and L. E. Kavraki, “The Open Motion Planning Library,”IEEE Robotics & Automation Magazine, vol. 19, no. 4, pp. 72–82, 2012

2012
[32]

Humanoid Robot HRP-5P: An electrically actuated humanoid robot with high-power and wide-range joints,

K. Kaneko, H. Kaminaga, T. Sakaguchi, S. Kajita, M. Morisawa, I. Kumagai, and F. Kanehiro, “Humanoid Robot HRP-5P: An electrically actuated humanoid robot with high-power and wide-range joints,”IEEE Robotics and Automation Letters, vol. 4, no. 2, pp. 1431–1438, 2019

2019
[33]

On the efficiency of certain quasi-random sequences of points in evaluating multi-dimensional integrals,

J. H. Halton, “On the efficiency of certain quasi-random sequences of points in evaluating multi-dimensional integrals,”Numerische Mathe- matik, vol. 2, pp. 84–90, 1960

1960

Pith tools

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

[1] [1]

Manipulation strategy decision and execution based on strategy proving operation for carrying large and heavy objects,

M. Murooka, S. Noda, S. Nozawa, Y . Kakiuchi, K. Okada, and M. Inaba, “Manipulation strategy decision and execution based on strategy proving operation for carrying large and heavy objects,” inIEEE International Conference on Robotics and Automation, 2014, pp. 3425–3432

2014

[2] [2]

Perception based locomotion system for a humanoid robot with adaptive footstep compensation under task constraints,

I. Kumagai, M. Morisawa, S. Nakaoka, T. Sakaguchi, H. Kaminaga, K. Kaneko, and F. Kanehiro, “Perception based locomotion system for a humanoid robot with adaptive footstep compensation under task constraints,” inIEEE/RSJ International Conference on Intelligent Robots and Systems, 2018, pp. 713–719

work page 2018

[3] [3]

Humanoid teleoperation for whole body manipulation,

M. Stilman, K. Nishiwaki, and S. Kagami, “Humanoid teleoperation for whole body manipulation,” inIEEE International Conference on Robotics and Automation, 2008, pp. 3175–3180

work page 2008

[4] [4]

Pivoting based manipulation by a humanoid robot,

E. Yoshida, M. Poirier, J.-P. Laumond, O. Kanoun, F. Lamiraux, R. Alami, and K. Yokoi, “Pivoting based manipulation by a humanoid robot,”Autonomous Robots, vol. 28, pp. 77–88, 2009

2009

[5] [5]

Hu- manoid full-body controller adapting constraints in structured objects through updating task-level reference force,

S. Nozawa, I. Kumagai, Y . Kakiuchi, K. Okada, and M. Inaba, “Hu- manoid full-body controller adapting constraints in structured objects through updating task-level reference force,” inIEEE/RSJ International Conference on Intelligent Robots and Systems, 2012, pp. 3417–3424

2012

[6] [6]

Finding locomanipulation plans quickly in the locomotion constrained manifold,

S. J. Jorgensen, M. Vedantam, R. Gupta, H. Cappel, and L. Sentis, “Finding locomanipulation plans quickly in the locomotion constrained manifold,” inIEEE International Conference on Robotics and Automa- tion, 2020, pp. 6611–6617

work page 2020

[7] [7]

Manipulation of documented objects by a walking humanoid robot,

S. Dalibard, A. Nakhaei, F. Lamiraux, and J. Laumond, “Manipulation of documented objects by a walking humanoid robot,” inIEEE-RAS International Conference on Humanoid Robots, 2010, pp. 518–523

work page 2010

[8] [8]

Humanoid whole-body planning for loco-manipulation tasks,

P. Ferrari, M. Cognetti, and G. Oriolo, “Humanoid whole-body planning for loco-manipulation tasks,” inIEEE International Conference on Robotics and Automation, 2017, pp. 4741–4746

work page 2017