Pith. sign in

REVIEW 5 major objections 6 minor 31 references

Refining Motion for Peak Performance: Identifying Optimal Gait Parameters for Energy-Efficient Quadrupedal Bounding

T0 review · 5 major / 6 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read The paper argues that in quadrupedal bounding, duty factor and stride duration — not speed alone — set the energy cost, with optimal values that shift with speed.

desk verdict Useful empirical sweep of bounding gait parameters, but the paper's own hardware data contradict its phase-shift conclusion and the validation needs more rigor. read the letter →

arxiv 2507.12751 v1 pith:4HI56BJL submitted 2025-07-17 cs.RO

Yasser G. Alqaham , Jing Cheng , Zhenyu Gan This is my paper
classification cs.RO
keywords quadrupedalboundingcostoftransportdutyfactorphaseshiftstridedurationenergy-efficientlocomotiongaitoptimizationA1quadrupedrobot
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 hypothesizes that duty factor, phase shift, and stride duration are the main gait-level levers for energy cost in quadrupedal bounding, and tests it on a simulated and physical quadruped. It finds that raising duty factor into the 0.22–0.30 range and lengthening stride duration lowers cost of transport at every speed tested, while phase shift has a smaller effect. If the finding holds on hardware, a bounding robot can extend battery life by adapting these timing parameters to its speed instead of running a fixed gait. The result suggests speed-dependent gait scheduling rather than one universal optimal setting.

What carries the argument

The load-bearing metric is the work-based cost of transport in Eq. (4): the sum of absolute mechanical power at all twelve joints over a stride, divided by weight times forward distance. The control machinery is a floating-base model of the A1 with point-contact feet, a gait generator that reduces bounding to three free parameters — duty factor $\gamma$, phase shift $\varphi$, and stride duration $T$ — and a controller in which each stance leg pair is treated as a spring-loaded inverted pendulum for hip height, ground reaction forces are solved by a quadratic program under a single-rigid-body approximation and friction-cone constraints, and swing legs follow a placement strategy. The controller deliberately lets the torso pitch rather than forcing a fixed posture, which is what makes long, efficient strides feasible.

What would settle it

Measure battery current on the A1 at 0.5 m/s while holding each gait parameter at the values in the paper's sweep for about 30 steady-state strides; if energy per meter is lower for duty factors below 0.22 or for shorter strides than for the recommended settings, the simulation-based ranking is wrong.

Watch

Extended reading notes

Core claim

The discovery is that the energy economy of a two-flight-phase bounding gait is largely set by two timing parameters: duty factor and stride duration. Across simulated speeds of 0.5, 1.5, and 2.5 m/s, cost of transport falls as duty factor rises into the 0.22–0.30 range and as stride duration lengthens toward the stability boundary, while mid-range phase shifts (0.44–0.52) keep cost low and consistent. Short strides are costly because they multiply touch-down events and power fluctuations; overly long strides destabilize the torso and cause slipping. Treadmill tests on the A1 at 0.5 m/s confirm the duty-factor trend and the benefit of longer strides, but put the best phase shift near 0.64 rather than 0.50, so the paper concludes the optimal parameters depend on speed and should be scheduled adaptively.

Load-bearing premise

The results assume the mechanical joint work measured in simulation ranks gaits the same way the robot's real battery consumption does, even though motor losses, damping, and friction are left out.

Editorial extensions

If this is right

  • A bounding controller can lower energy use by holding duty factor near 0.22–0.30 and using the longest stride duration that keeps the torso stable, with no hardware changes.
  • Optimal gait parameters depend on forward speed, so an adaptive scheduler that retunes $\gamma$, $\varphi$, and $T$ as speed changes should beat any single fixed gait.
  • Since mid-range phase shifts give low and stable COT in simulation, phase shift can be set coarsely without a major energy penalty, simplifying control.
  • On physical hardware, best efficiency sits at higher phase shift and longer stride than simulation predicts, so deployment should include hardware-in-the-loop tuning even if simulation guides the search.

Reading between the lines

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

  • The authors leave implicit that the same sweep could be run online: a controller that nudges duty factor and stride duration while measuring cost of transport could find the speed-dependent optimum without a precomputed table.
  • Because the controller already treats legs as springs and lets the torso rotate freely, the optimal stride duration may track a mechanical resonance; plotting optima against Froude number could collapse all three speeds onto one curve.
  • The simulation-to-hardware gap in phase shift and stride duration suggests that adding motor electrical losses and joint damping to the model would shift predicted optima toward the hardware values.
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

5 major / 6 minor

Summary. This paper investigates how three bounding gait parameters (duty factor, phase shift, and stride duration) affect the cost of transport of a Unitree A1 quadrupedal robot. The authors build on their prior SLIP-based controller and trajectory-optimized default gait, implement it in Gazebo, and sweep each parameter at 0.5, 1.5, and 2.5 m/s, reporting COT computed from absolute mechanical joint power. They then perform hardware validation on a treadmill at 0.5 m/s. The paper claims that higher duty factors (in the two-flight-phase bounding regime) and longer stride durations lower COT, that phase-shift variations have subtler effects, that the optimal parameter combination depends on speed, and that the hardware results show similar trends to simulation.

Significance. If the claims held, the paper would provide a useful, systematic mapping of gait-parameter sensitivity for bounding on a commercially available quadruped, with the additional value of a direct hardware comparison. The simulation sweep is systematic and internally consistent, and the public code/video links are a reproducibility strength. The hardware duty-factor trend (optimal near 0.30) is consistent with the simulation's low-cost range of 0.22-0.30, which is a genuine point in favor of the approach. However, the paper's central validation claim is weakened by the phase-shift and stride-duration discrepancies between simulation and hardware, and by the absence of error bars and statistical analysis in the hardware section.

major comments (5)
  1. [§III.C, Fig. 5(B)] The hardware phase-shift data contradict the statement in the Fig. 3 caption and Section III.A that phase-shift variations have 'more subtle effects' and that hardware shows 'similar trends' to simulation. In simulation at 0.5 m/s (Fig. 3(D)) COT decreases to a minimum near phi=0.44-0.52 and then rises slightly, whereas on hardware (Fig. 5(B)) COT decreases through the largest tested phi=0.64 and varies by about a factor of four. The paper itself reports the hardware optimum at phi=0.64 versus 0.50 in simulation. Because the interpretation of phase shift is load-bearing for the central claim, this discrepancy must be addressed, either by adding motor losses/friction to the simulation, by reporting electrical-power measurements, or by explicitly reframing the phase-shift conclusion as simulation-only.
  2. [§III.C, Fig. 5(C)] The hardware stride-duration data do not show the simulation's optimal region. Simulation at 0.5 m/s finds a COT minimum near T=0.22-0.26 s and reports instability beyond 0.34 s, while Fig. 5(C) shows the hardware's most efficient tested point at T=0.36 s with no clear optimum inside the range. The text in Section III.C notes that 0.36 s was 'less stable,' but the conclusion still states that longer stride durations lower COT at all speeds. This overgeneralizes a hardware trend measured at one speed and apparently confounded by stability, so the claim must be restricted and the discrepancy explained.
  3. [§III.C] The hardware validation reports no number of repeated trials, no error bars, and no statistical tests for the COT data in Fig. 5. Given the variability visible in the simulation box plots and the small differences between some parameter settings, the reader cannot judge whether the hardware optima (e.g., phi=0.64 versus 0.50) are significant. At minimum, the authors should report per-condition trial counts, means and standard deviations, and ideally a paired comparison of the claimed optimal settings against neighboring settings.
  4. [§III.A, §III.C] The experimental validation is performed only at 0.5 m/s, while the paper's conclusions state that the optimal combination varies with the robot's average speed and that longer stride durations lower COT at all speeds. The three-speed dependence is therefore supported only by simulation; the hardware data cannot validate the speed-dependence claim. The authors should either add hardware tests at 1.5 and 2.5 m/s or explicitly limit the experimental conclusions to the low-speed case.
  5. [§II.D, §III.C] The work-based COT in Eq. (4) uses absolute mechanical joint power, while Section III.C concedes that the simulation omits motor dynamics, mechanical damping, and friction. The hardware mismatches appear in exactly the parameters (phase shift and stride duration) where these losses could re-order the ranking. A concrete fix is to compare Eq. (4) with electrical energy measured from the motor drivers on a subset of the swept settings; if the ranking changes, the optimal-parameter claims must be revised.
minor comments (6)
  1. [Abstract] The word 'demonstrate' in the abstract is too strong given the partial validation; 'suggest' or 'indicate' would be more accurate.
  2. [§II.C, Eq. (2)] The fitted coefficients a1-a4 in Eq. (2) are not reported numerically and no fitting data or procedure is given; this limits reproducibility of the controller. Adding the values or a reference to the fitting details would help.
  3. [§II.C, Eq. (3)] The QP weights W1, W2, W3 and scalars alpha and beta are not given numerically, and the friction coefficient mu is stated only as 0.6. These parameters should be listed in the paper rather than only in the code release.
  4. [§III.C] The text says data sets were omitted when the robot failed to reach a steady state, but it does not report how many configurations were omitted or at which parameter values; this information is needed to assess the completeness of the parameter sweep.
  5. [§III.B, Fig. 4] Fig. 4 shows absolute power for three single strides, but the text does not say whether these are representative individual strides or averages; clarifying this would improve the interpretation.
  6. [References] Reference [28] is listed as 'under review' and should be updated to its published status if available, and the access dates for the Unitree and ROS web references should be consistent.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation; COT is measured from simulation and hardware, not constructed from the gait-parameter inputs.

full rationale

The paper's central dependent variable, cost of transport, is computed directly from simulated and experimental joint torques, velocities, and traveled distance via Eq. (4). It is not algebraically derived from the gait parameters themselves, so there is no equation-level circularity. The authors' prior work is used to choose the controller (Ref. [27]) and the default sweep center (Ref. [26]), but these are experimental inputs rather than conclusions. The reported optimal ranges (duty factor 0.22-0.30, stride duration 0.22-0.26 s, phase shift 0.44-0.52) are broader than the default point and are not equal to the default by construction; the hardware sweep even reports a different phase-shift optimum (0.64 versus 0.50). Thus the self-citations do not carry the load-bearing argument; the new simulations and real-robot experiments are the evidence. The most serious issue is a non-circular correctness problem: Section III.C claims 'we observe similar trends' while Fig. 5(B) shows monotonic COT decrease up to phase shift 0.64 with roughly fourfold variation, contradicting the stated 'subtle effects' claim for phase shift. This is an internal consistency issue, not a circularity, and therefore does not raise the circularity score. The acknowledged model discrepancies (missing motor dynamics, damping, friction) likewise weaken validity but not circularity.

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

The central COT landscape depends on the controller's fitted hip trajectory coefficients, the QP weights, the friction model, and the chosen energy metric. These are inherited from the authors' prior work or chosen by hand, and any of them could shift the reported optima.

free parameters (4)
  • a1, a2, a3, a4 in Eq. (2) hip trajectory fit = not disclosed
    Coefficients of the sinusoidal SLIP-based hip dynamics are determined by curve fitting in prior work [27] and control the simulated motion that generates all COT values.
  • QP weights W1, W2, W3, alpha, beta = not reported
    Cost weights in Eq. (3) shape the ground reaction forces and thus the torques and COT; they are chosen by the authors without sensitivity analysis.
  • static friction coefficient mu = 0.6
    Used in the QP friction cone constraints; a different value would change feasible GRFs and possibly COT.
  • Default gait parameters (gamma=0.22, phi=0.50, T=0.22 s) = 0.22, 0.50, 0.22
    Center of the parameter sweeps, taken from the authors' prior trajectory optimization [26]; they anchor the reported optima.
assumptions (6)
  • standard math The floating-base Euler-Lagrange dynamics in Eq. (1) accurately describe the Unitree A1.
    Standard rigid-body dynamics; accepted but not separately validated in the paper.
  • domain assumption SLIP-based sinusoidal hip trajectory Eq. (2) with fitted coefficients produces the desired COM motion.
    This is a modeling assumption from prior work [27]; the paper uses it without revalidation.
  • domain assumption Single Rigid Body (SRB) approximation is sufficient for computing ground reaction forces in bounding.
    SRB ignores joint acceleration effects and elastic deformation of the body; acceptable for the controller but not error-quantified.
  • domain assumption Work-based COT in Eq. (4), using absolute mechanical joint power, is an adequate energy metric.
    Omits electrical losses, motor dynamics, and friction, as acknowledged in Sec. III.C; the paper assumes rankings are preserved.
  • domain assumption The controller from [27] can stably track the commanded bounding gaits for 30 strides on the real robot.
    Experimental feasibility relies on this; unstable configurations were omitted, confirming this is a binding assumption.
  • domain assumption Front and rear leg pairs share the same duty factor in the bounding gait.
    Reduces the gait parameter space from eight to three parameters (Sec. II.B); asymmetry is not explored.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Refining Motion for Peak Performance: Identifying Optimal Gait Parameters for Energy-Efficient Quadrupedal Bounding." pith.science (2026). https://pith.science/paper/4HI56BJL

@misc{pith2026250712751,
  author       = {Pith},
  title        = {Pith review of: Refining Motion for Peak Performance: Identifying Optimal Gait Parameters for Energy-Efficient Quadrupedal Bounding},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/4HI56BJL}},
  note         = {Machine review of arXiv:2507.12751}
}
read the original abstract

Energy efficiency is a critical factor in the performance and autonomy of quadrupedal robots. While previous research has focused on mechanical design and actuation improvements, the impact of gait parameters on energetics has been less explored. In this paper, we hypothesize that gait parameters, specifically duty factor, phase shift, and stride duration, are key determinants of energy consumption in quadrupedal locomotion. To test this hypothesis, we modeled the Unitree A1 quadrupedal robot and developed a locomotion controller capable of independently adjusting these gait parameters. Simulations of bounding gaits were conducted in Gazebo across a range of gait parameters at three different speeds: low, medium, and high. Experimental tests were also performed to validate the simulation results. The findings demonstrate that optimizing gait parameters can lead to significant reductions in energy consumption, enhancing the overall efficiency of quadrupedal locomotion. This work contributes to the advancement of energy-efficient control strategies for legged robots, offering insights directly applicable to commercially available platforms.

Figures

Figures reproduced from arXiv: 2507.12751 by the authors.

Figure 1
Figure 1. This figure illustrates the gait parameters in one stride for the [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. This control diagram illustrates the quadrupedal robot’s gait generation and control system. The Gait Generation block determines the key gait [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. This figure shows the cost of transport (COT) as a function of three key gait parameters: duty factor (red), phase shift (blue), and stride duration [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: This figure illustrates the absolute power consumption across [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

31 extracted references · 28 canonical work pages

  1. [1]

    Bigdog, the rough-terrain quadruped robot,

    M. Raibert, K. Blankespoor, G. Nelson, and R. Playter, “Bigdog, the rough-terrain quadruped robot,” in Proceedings of the 17th World Congress. IFAC, 2008, pp. 10 822–10 825

    work page 2008

  2. [2]

    High speed trot- running: Implementation of a hierarchical controller using propriocep- tive impedance control on the mit cheetah,

    D. J. Hyun, S. Seok, A. Wang, and S. Kim, “High speed trot- running: Implementation of a hierarchical controller using propriocep- tive impedance control on the mit cheetah,” The International Journal of Robotics Research , vol. 33, no. 11, pp. 1417–1445, 2014

    work page 2014

  3. [3]

    Design of the hydraulically actuated, torque- controlled quadruped robot hyq,

    C. Semini et al. , “Design of the hydraulically actuated, torque- controlled quadruped robot hyq,” IEEE/ASME Transactions on Mecha- tronics, vol. 17, no. 5, pp. 1081–1091, 2011

    work page 2011

  4. [4]

    Selecting gaits for economical locomotion of legged robots,

    W. Xi, Y . Yesilevskiy, and C. D. Remy, “Selecting gaits for economical locomotion of legged robots,” The International Journal of Robotics Research , vol. 35, no. 9, p. 1140–1154, Nov. 2015. [Online]. Available: http://dx.doi.org/10.1177/0278364915612572

    work page doi:10.1177/0278364915612572 2015

  5. [5]

    Energy-efficient gait planning for a hexapod robot,

    D. Belter and P. Skrzypczy ´nski, “Energy-efficient gait planning for a hexapod robot,” IEEE Transactions on Robotics , vol. 32, no. 6, pp. 1342–1357, 2016

    work page 2016

  6. [6]

    Efficient gait selection for quadrupedal robots on the moon and mars,

    Kolvenbach, Hendrik, Bellicoso, Dario, Jenelten, Fabian, Wellhausen, Lorenz, and Hutter, Marco, “Efficient gait selection for quadrupedal robots on the moon and mars,” 2018. [Online]. Available: http://hdl.handle.net/20.500.11850/261939

    work page 2018

  7. [7]

    Optimization and gaits in the locomotion of vertebrates,

    R. M. Alexander, “Optimization and gaits in the locomotion of vertebrates,” Physiological Reviews , vol. 69, no. 4, pp. 1199–1227, 1989

    work page 1989

  8. [8]

    Gait and the energetics of locomotion in horses,

    D. F. Hoyt and C. R. Taylor, “Gait and the energetics of locomotion in horses,” Nature, vol. 292, no. 5820, pp. 239–240, 1981

    work page 1981

Show all 31 references
  1. [9]

    Control of locomotion in bipeds, tetrapods, and fish,

    S. Grillner, “Control of locomotion in bipeds, tetrapods, and fish,” Handbook of Physiology: The Nervous System II , pp. 1179–1236, 1981

    1981

  2. [10]

    Energetics and mechanics of terrestrial locomotion. ii. kinetic energy changes of the limbs and body as a function of speed and body size in birds and mammals,

    N. C. Heglund and C. R. Taylor, “Energetics and mechanics of terrestrial locomotion. ii. kinetic energy changes of the limbs and body as a function of speed and body size in birds and mammals,” Journal of Experimental Biology , vol. 97, no. 1, pp. 23–40, 1982

    1982

  3. [11]

    Symmetrical gaits of horses,

    M. Hildebrand, “Symmetrical gaits of horses,” Science, vol. 150, no. 3697, pp. 701–708, Nov. 1965. [Online]. Available: https://doi.org/10.1126/science.150.3697.701

    1965 doi

  4. [12]

    Functional anatomy of the cheetah (acinonyx jubatus) hindlimb,

    P. E. Hudson, S. A. Corr, and A. M. Wilson, “Functional anatomy of the cheetah (acinonyx jubatus) hindlimb,” Journal of Anatomy , vol. 221, no. 6, pp. 632–644, 2012

    2012

  5. [13]

    Gaits of bipedal and quadrupedal animals,

    R. M. Alexander, “Gaits of bipedal and quadrupedal animals,” The International Journal of Robotics Research , vol. 3, no. 2, pp. 49–59, 1984

    1984

  6. [14]

    Abernethy et al

    B. Abernethy et al. , The Biophysical Foundations of Human Move- ment. Human Kinetics, 2016

    2016

  7. [15]

    Biomechanics: forces and move- ments in muscular locomotion,

    J. R. Hutchinson and M. Garcia, “Biomechanics: forces and move- ments in muscular locomotion,” Encyclopedia of Life Sciences , pp. 1–9, 2006

    2006

  8. [16]

    Scaling stride frequency and gait to animal size: mice to horses,

    N. C. Heglund, C. R. Taylor, and T. A. McMahon, “Scaling stride frequency and gait to animal size: mice to horses,” Science, vol. 186, no. 4169, pp. 1112–1113, 1974

    1974

  9. [17]

    Legged robots that balance,

    M. H. Raibert, “Legged robots that balance,” in The MIT Press, 1986

    1986

  10. [18]

    Trotting and galloping strategies of quadruped robots via analogue neural oscillators,

    D. Kim et al., “Trotting and galloping strategies of quadruped robots via analogue neural oscillators,” in Bioinspiration & Biomimetics , vol. 6, no. 1. IOP Publishing, 2011, p. 016002

    2011

  11. [19]

    Balance recovery with out-of-plane foot placement strategies for humanoid robots,

    J. G. D. Karssen, M. Wisse, and F. Gritti, “Balance recovery with out-of-plane foot placement strategies for humanoid robots,” in 2011 IEEE International Conference on Robotics and Automation . IEEE, 2011, pp. 2111–2116

    2011

  12. [20]

    Compliant leg behaviour explains basic dynamics of walking and running,

    H. Geyer, A. Seyfarth, and R. Blickhan, “Compliant leg behaviour explains basic dynamics of walking and running,” Proceedings of the Royal Society B , vol. 273, no. 1603, pp. 2861–2867, 2006

    2006

  13. [21]

    Development of a novel quadruped robot with parallel leg mechanism,

    F. Fei, F. Gao, and K. Wei, “Development of a novel quadruped robot with parallel leg mechanism,” in 2018 IEEE International Conference on Mechatronics and Automation (ICMA) . IEEE, 2018, pp. 1691– 1696

    2018

  14. [22]

    M. F. Silva and J. A. T. Machado, Energy Efficiency of Quadruped Gaits. Springer Berlin Heidelberg, 2006, p. 735–742. [Online]. Available: http://dx.doi.org/10.1007/3-540-26415-9 88

    2006 doi

  15. [23]

    High-speed bounding with the mit cheetah 2: Control design and experiments,

    H.-W. Park, P. M. Wensing, and S. Kim, “High-speed bounding with the mit cheetah 2: Control design and experiments,” The International journal of robotics research , vol. 36, no. 2, pp. 167–192, 2017

    2017

  16. [24]

    Mini cheetah: A platform for pushing the limits of dynamic quadruped control,

    B. Katz, J. Di Carlo, and S. Kim, “Mini cheetah: A platform for pushing the limits of dynamic quadruped control,” arXiv preprint arXiv:1909.06586, 2019

    1909 arXiv

  17. [25]

    Mit cheetah 3: Design and control of a robust, dynamic quadruped robot,

    G. Bledt, M. J. Powell, B. Katz, J. Di Carlo, P. M. Wensing, and S. Kim, “Mit cheetah 3: Design and control of a robust, dynamic quadruped robot,” in 2018 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) . IEEE, 2018, pp. 2245–2252

    2018

  18. [26]

    Energy-optimal asymmetrical gait selection for quadrupedal robots,

    Y . G. Alqaham, J. Cheng, and Z. Gan, “Energy-optimal asymmetrical gait selection for quadrupedal robots,” IEEE Robotics and Automation Letters, 2024

    2024

  19. [27]

    Harnessing natural oscillations for high-speed, efficient asymmetrical locomotion in quadrupedal robots,

    J. Cheng, Y . G. Alqaham, and Z. Gan, “Harnessing natural oscillations for high-speed, efficient asymmetrical locomotion in quadrupedal robots,” in 2024 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) . IEEE, Oct. 2024, p. 6864–6869. [Online]. Availa...

    2024

  20. [28]

    All common bipedal gaits emerge from a single passive model,

    Z. Gan, Y . Yesilevskiy, P. Zaytsev, and C. D. Remy, “All common bipedal gaits emerge from a single passive model,” Journal of the Royal Society, Interface , vol. under review, 2018

    2018

  21. [29]

    Unitree Robotics, 2023, https://www.unitree.com/en/a1, Last accessed on 2023-2-28

    2023

  22. [30]

    Robot Operating System, 2022, https://www.ros.org/, Last accessed on 2022-9-23

    2022

  23. [31]

    Design and use paradigms for gazebo, an open-source multi-robot simulator,

    N. Koenig and A. Howard, “Design and use paradigms for gazebo, an open-source multi-robot simulator,” in 2004 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) , vol. 3. IEEE, 2004, pp. 2149–2154

    2004

Pith tools

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