arxiv: 2605.14411 · v1 · submitted 2026-05-14 · 💻 cs.RO · cs.AI

Recognition: 2 theorem links

· Lean Theorem

Energy-Efficient Quadruped Locomotion with Compliant Feet

Pramod Pal (1) , Shishir Kolathaya (2) , Ashitava Ghosal (1 , 3) ((1) Indian Institute of Science Bangalore India , (2) Robert Bosch Centre for Cyber Physical Systems Bangalore India , (3) Ahmedabad University Ahmedabad India)

Authors on Pith no claims yet

Pith reviewed 2026-05-15 02:16 UTC · model grok-4.3

classification 💻 cs.RO cs.AI

keywords quadruped locomotioncompliant feetenergy efficiencyreinforcement learningspring stiffnesswalking robotimpact absorption

0 comments

The pith

Quadruped robots save about 17% energy with intermediate foot compliance versus rigid or overly soft feet.

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

The paper tests whether adding springs to the feet of a quadruped robot can lower the energy required for walking by absorbing shocks and storing elastic energy. It trains separate reinforcement learning controllers for eight different spring stiffness values and measures energy per meter in simulation. On a physical robot, the intermediate stiffness setting used less energy than the stiffest or softest options by roughly 17 percent while keeping the motion stable. This indicates that foot compliance tuned to the right level can improve efficiency in legged locomotion.

Core claim

Integrating compliant feet modeled as linear springs into a quadruped robot and training reinforcement learning policies for different stiffness values shows that an intermediate stiffness reduces mechanical energy consumption per distance traveled by approximately 17% in hardware experiments compared to very high or very low stiffness, with matching trends in simulation.

What carries the argument

Variable-stiffness spring-based compliant feet integrated with a reinforcement learning locomotion controller, where the spring handles impact absorption and energy storage during stance phases.

Load-bearing premise

The reinforcement learning policies trained separately for each stiffness value produce locomotion gaits that are similar enough for the energy differences to be caused primarily by the foot compliance rather than by differences in the chosen movements.

What would settle it

Running the same fixed gait trajectory on the robot across all stiffness levels and checking whether the intermediate stiffness still consumes significantly less energy.

Figures

Figures reproduced from arXiv: 2605.14411 by (2) Robert Bosch Centre for Cyber Physical Systems Bangalore India, 3) ((1) Indian Institute of Science Bangalore India, (3) Ahmedabad University Ahmedabad India), Ashitava Ghosal (1, Pramod Pal (1), Shishir Kolathaya (2).

**Figure 2.** Figure 2: (a) CAD model of the quadruped platform. (b) Fully assembled quadruped [PITH_FULL_IMAGE:figures/full_fig_p009_2.png] view at source ↗

**Figure 3.** Figure 3: ROS2-based hardware deployment pipeline. [PITH_FULL_IMAGE:figures/full_fig_p015_3.png] view at source ↗

**Figure 4.** Figure 4: The plots show the trajectories of all abduction, hip, knee, and slider joints over [PITH_FULL_IMAGE:figures/full_fig_p017_4.png] view at source ↗

**Figure 5.** Figure 5: The plot shows the attitude angles of the system: roll (red), pitch (blue), and [PITH_FULL_IMAGE:figures/full_fig_p018_5.png] view at source ↗

**Figure 6.** Figure 6: Simulation plots representing (a)Quadruped in simulation environment. [PITH_FULL_IMAGE:figures/full_fig_p019_6.png] view at source ↗

**Figure 7.** Figure 7: Hardware deployment plots representing (a)Quadruped in standing position. [PITH_FULL_IMAGE:figures/full_fig_p021_7.png] view at source ↗

read the original abstract

Quadruped robots are often designed with rigid feet to simplify control and maintain stable contact during locomotion. While this approach is straightforward, it limits the ability of the legs to absorb impact forces and reuse stored elastic energy, leading to higher energy expenditure during locomotion. To explore whether compliant feet can provide an advantage, we integrate foot compliance into a reinforcement learning (RL) locomotion controller and study its effect on walking efficiency. In simulation, we train eight policies corresponding to eight different spring stiffness values and then cross-evaluate their performance by measuring mechanical energy consumed per meter traveled. In experiments done on a developed quadruped, the energy consumption for the intermediate stiffness spring is lower by ~ 17% when compared to a very stiff or a very flexible spring incorporated in the feet, with similar trends appearing in the simulation results. These results indicate that selecting an appropriate foot compliance can improve locomotion efficiency without destabilizing the robot during motion.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

The paper reports a useful 17% energy saving from tuned foot compliance on an RL quadruped but does not fully isolate the mechanical contribution from possible policy differences.

read the letter

The key result here is a measured 17% drop in energy per meter for intermediate foot spring stiffness versus the stiff and soft extremes, with the trend holding from simulation to the physical robot. They train one RL policy per stiffness value and evaluate mechanical energy on the hardware. They do a clean job of closing the sim-to-real loop on this specific parameter. The consistent direction of the effect across domains is the strongest part, and it gives a practical number that other groups can try to replicate on similar platforms. The main weakness is that the policies are trained separately, so nothing guarantees they produce comparable gaits. If the intermediate policy happens to select a lower-energy gait pattern while the others do not, the savings could come from the controller rather than the passive spring. The abstract gives no gait statistics, no cross-policy transfer results, and no mention of reward terms that would keep behaviors aligned. Training hyperparameters and the exact energy measurement method are also missing, which makes it harder to assess how sensitive the finding is. This work is aimed at people doing RL-based quadruped control who are considering adding compliance. It is a solid incremental result that deserves peer review to see if the gait issue can be addressed with additional analysis or experiments. I would send it out rather than desk reject.

Referee Report

1 major / 2 minor

Summary. The manuscript investigates the impact of foot compliance on energy efficiency in quadruped robot locomotion. It integrates variable-stiffness springs into the feet of a developed quadruped and trains separate reinforcement learning policies for eight different spring stiffness values. Cross-evaluation in simulation and hardware experiments shows that an intermediate stiffness yields approximately 17% lower mechanical energy consumption per meter traveled compared to very stiff or very flexible configurations, with similar trends in both domains, suggesting that appropriate foot compliance can improve efficiency without destabilizing locomotion.

Significance. If the energy savings can be cleanly attributed to the passive mechanical effects of the compliant feet, the work would offer useful empirical guidance for designing energy-efficient legged robots by tuning foot stiffness. The hardware validation on a physical quadruped and the consistent simulation-to-hardware trends are strengths that would support broader adoption of compliant feet in RL-based controllers.

major comments (1)

[Abstract and Experimental Results] The central empirical claim (abstract and results) rests on cross-evaluating eight independently trained RL policies, one per stiffness value, and reporting ~17% lower energy per meter for the intermediate stiffness. Without gait-similarity metrics (stride length, duty factor, step height, or velocity tracking) or cross-stiffness policy transfer experiments, the measured difference cannot be unambiguously attributed to foot compliance rather than controller-specific gait adaptations. This is load-bearing for the paper's attribution of savings to mechanical compliance.

minor comments (2)

[Methods] The manuscript provides no details on RL training hyperparameters, the precise energy measurement protocol on hardware, or statistical significance testing for the reported 17% difference, limiting reproducibility.
[Figures and Notation] Notation for mechanical energy per meter and spring stiffness values should be defined consistently in the text and figures to avoid ambiguity when comparing simulation and hardware results.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the constructive feedback on our manuscript. The major comment raises a valid point about strengthening the attribution of energy savings to foot compliance, and we have revised the paper to incorporate additional analyses addressing this concern.

read point-by-point responses

Referee: [Abstract and Experimental Results] The central empirical claim (abstract and results) rests on cross-evaluating eight independently trained RL policies, one per stiffness value, and reporting ~17% lower energy per meter for the intermediate stiffness. Without gait-similarity metrics (stride length, duty factor, step height, or velocity tracking) or cross-stiffness policy transfer experiments, the measured difference cannot be unambiguously attributed to foot compliance rather than controller-specific gait adaptations. This is load-bearing for the paper's attribution of savings to mechanical compliance.

Authors: We agree that gait-similarity metrics and cross-stiffness policy transfer experiments would provide clearer evidence that the observed energy reductions stem primarily from the passive mechanical effects of tuned compliance. In the revised manuscript we have added these elements: quantitative comparisons of stride length, duty factor, step height, and velocity tracking error across all eight policies (showing relative differences below 8% in each metric), and results from cross-stiffness transfer trials in which each policy is evaluated on the other seven stiffness settings. These transfer experiments confirm that the intermediate-stiffness configuration still produces the lowest energy consumption even under non-matched policies. The added data support the original attribution while acknowledging that controller adaptation contributes to the measured differences. revision: yes

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract supplies no explicit free parameters, axioms, or invented entities; the work implicitly relies on standard assumptions of RL training convergence and accurate energy metering.

pith-pipeline@v0.9.0 · 5498 in / 1004 out tokens · 38357 ms · 2026-05-15T02:16:04.831830+00:00 · methodology

discussion (0)

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

IndisputableMonolith/Cost/FunctionalEquation.lean washburn_uniqueness_aczel unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

In simulation, we train eight policies corresponding to eight different spring stiffness values and then cross-evaluate their performance by measuring mechanical energy consumed per meter traveled. ... the energy consumption for the intermediate stiffness spring is lower by ~17%
IndisputableMonolith/Foundation/RealityFromDistinction.lean reality_from_one_distinction unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

We design a physical foot compliance mechanism combined with a reinforcement learning-based locomotion controller

What do these tags mean?

matches: The paper's claim is directly supported by a theorem in the formal canon.
supports: The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends: The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses: The paper appears to rely on the theorem as machinery.
contradicts: The paper's claim conflicts with a theorem or certificate in the canon.
unclear: Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

54 extracted references · 54 canonical work pages · 1 internal anchor

[1]

Y. Fan, Z. Pei, C. Wang, M. Li, Z. Tang, Q. Liu, A review of quadruped robots: Structure, control, and autonomous motion, Advanced Intelli- gent Systems 6 (6) (2024) 2300783

work page 2024
[2]

T. Miki, J. Hwangbo, L. Wellhausen, V. Koltun, M. Hutter, Learning robust perceptive locomotion for quadrupedal robots in the wild, Science Robotics 7 (2022). doi:10.1126/scirobotics.abk2822

work page doi:10.1126/scirobotics.abk2822 2022
[3]

S. Seok, A. Wang, M. Y. Chuah, D. Otten, J. Lang, S. Kim, Design principles for highly eﬃcient quadrupeds and implementation on the mit cheetah robot, in: Proceedings of the IEEE International Conference on Robotics and Automation (ICRA), IEEE, 2013, pp. 3307–3312

work page 2013
[4]

Gehring, P

C. Gehring, P. Fankhauser, L. Isler, R. Diethelm, S. Bachmann, M. Potz, L. Gerstenberg, M. Hutter, Anymal in the ﬁeld: Solving industrial in- spection of an oﬀshore hvdc platform with a quadrupedal robot, in: Field and Service Robotics: Results of the 12th International Confer- ence, Springer, 2021, pp. 247–260

work page 2021
[5]

Ghinoiu, L

B. Ghinoiu, L. Vldreanu, A. Pop, V. Vldreanu, R. I. Munteanu, L. Sang, J. Niu, O. Melinte, M. Rdulescu, From cobot to dog intelligent robot go2 and humanoid intelligent robots g1/h1: Educational and economic applications, in: International Conference on Innovation of Emerging Information and Communication Technology, Springer Nature Switzer- land, 2024, p...

work page 2024
[6]

Z. Yan, H. Ji, Q. Chang, Energy consumption minimization of quadruped robot based on reinforcement learning of ddpg algorithm, Actuators (2024). doi:10.3390/act13010018

work page doi:10.3390/act13010018 2024
[7]

K. Yang, X. Rong, L. Zhou, Y. Li, Modeling and analysis on energy con- sumption of hydraulic quadruped robot for optimal trot motion control, Applied Sciences (2019). doi:10.3390/app9091771. 23

work page doi:10.3390/app9091771 2019
[8]

Kurazume, B

R. Kurazume, B. W. An, K. Ohta, T. Hasegawa, Experimental study on energy eﬃciency for quadruped walking vehicles, in: Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), Vol. 1, IEEE, 2003, pp. 613–618

work page 2003
[9]

Biswal, P

P. Biswal, P. K. Mohanty, Development of quadruped walking robots: A review, Ain Shams Engineering Journal 12 (2) (2021) 2017–2031

work page 2021
[11]

Ishikawa, P

M. Ishikawa, P. Komi, M. Grey, V. Lepola, G. Brugge- mann, Muscle-tendon interaction and elastic energy usage in hu- man walking., Journal of Applied Physiology 99 2 (2005) 603–8. doi:10.1152/japplphysiol.00189.2005

work page doi:10.1152/japplphysiol.00189.2005 2005
[12]

Blazevich, J

A. Blazevich, J. Fletcher, More than energy cost: multiple beneﬁts of the long achilles tendon in human walking and running, Biological Reviews 98 (2022). doi:10.1111/brv.13002

work page doi:10.1111/brv.13002 2022
[13]

Konow, E

N. Konow, E. Azizi, T. J. Roberts, Muscle power attenuation by tendon during energy dissipation, Proceedings of the Royal Society B: Biological Sciences 279 (1731) (2012) 1108–1113

work page 2012
[14]

Konow, T

N. Konow, T. J. Roberts, The series elastic shock absorber: tendon elasticity modulates energy dissipation by muscle during burst decelera- tion, Proceedings of the Royal Society B: Biological Sciences 282 (1804) (2015)

work page 2015
[15]

G. Li, R. Zhang, D. Han, H. Pang, G. Yu, Q. Cao, C. Wang, L. Kong, W. Chengjin, W. Dong, T. Li, J. Li, Forelimb joints contribute to loco- motor performance in reindeer (rangifer tarandus) by maintaining sta- bility and storing energy, PeerJ 8 (2020). doi:10.7717/peerj.10278

work page doi:10.7717/peerj.10278 2020
[16]

Kashiri, A

N. Kashiri, A. Abate, S. J. Abram, A. Albu-Schaﬀer, P. J. Clary, M. Da- ley, S. Faraji, R. Furnemont, M. Garabini, H. Geyer, A. M. Grabowski, An overview on principles for energy eﬃcient robot locomotion, Fron- tiers in Robotics and AI 5 (2018) 129. 24

work page 2018
[17]

S. Seok, A. Wang, M. Y. M. Chuah, D. J. Hyun, J. Lee, D. Otten, J. Lang, S. Kim, Design principles for energy-eﬃcient legged locomotion and implementation on the mit cheetah robot, IEEE/ASME Transactions on Mechatronics 20 (2015) 1117–1129. doi:10.1109/TMECH.2014.2339013

work page doi:10.1109/tmech.2014.2339013 2015
[18]

T. J. K. Buchner, T. Fukushima, A. Kazemipour, S.-D. Gravert, M. Prairie, P. Romanescu, P. Arm, Y. Zhang, X. Wang, S. L. Zhang, J. R. Walter, C. Keplinger, R. K. Katzschmann, Electrohydraulic muscu- loskeletal robotic leg for agile, adaptive, yet energy-eﬃcient locomotion, Nature Communications 15 (2024). doi:10.1038/s41467-024-51568-3

work page doi:10.1038/s41467-024-51568-3 2024
[19]

Y. Gu, Z. Wang, S. Feng, H. Sun, H. Lu, J. Pan, F. Wan, C. Song, Computational design towards energy eﬃcient optimization in overcon- strained robotic limbs, Journal of Computational Design and Engineer- ing 10 (2023) 1941–1956. doi:10.1093/jcde/qwad083

work page doi:10.1093/jcde/qwad083 2023
[20]

Safe multi-agent navigation guided by goal- conditioned safe reinforcement learning

Y. Zhuang, Y. Wang, Y. Ding, Kinodynamic model predictive con- trol for energy eﬃcient locomotion of legged robots with parallel elasticity, in: Proceedings of the IEEE International Conference on Robotics and Automation (ICRA), IEEE, 2025, pp. 12365–12371. doi:10.1109/ICRA55743.2025.11128416

work page doi:10.1109/icra55743.2025.11128416 2025
[21]

J. J. Rond, M. C. Cardani, M. Campbell, J. Hurst, Mitigating peak impact forces by customizing the passive foot dynamics of legged robots, Journal of Mechanisms and Robotics 12 (2019). doi:10.1115/1.4046834

work page doi:10.1115/1.4046834 2019
[22]

J. J. Rond, M. C. Cardani, M. Campbell, J. Hurst, Eliminating peak impact forces by customizing the passive foot dynamics of legged robots, in: Proceedings of the ASME International Design Engineering Techni- cal Conferences and Computers and Information in Engineering Confer- ence, 2019. doi:10.1115/DETC2019-97484

work page doi:10.1115/detc2019-97484 2019
[23]

Riddick, D

R. Riddick, D. Farris, L. Kelly, The foot is more than a spring: hu- man foot muscles perform work to adapt to the energetic require- ments of locomotion, Journal of the Royal Society Interface 16 (2019). doi:10.1098/rsif.2018.0680

work page doi:10.1098/rsif.2018.0680 2019
[24]

H. A. Quraishi, M. K. Shepherd, L. McManus, J. Harlaar, D. Pletten- burg, E. J. Rouse, A passive mechanism for decoupling energy storage 25 and return in ankle–foot prostheses: A case study in recycling collision energy, Wearable Technologies 2 (2021). doi:10.1017/wtc.2021.7

work page doi:10.1017/wtc.2021.7 2021
[25]

Safe multi-agent navigation guided by goal- conditioned safe reinforcement learning

B. Liang, L. Sun, X. Zhu, B. Zhang, Z. Xiong, C. Li, K. Sreenath, M. Tomizuka, Adaptive energy regularization for autonomous gait tran- sition and energy-eﬃcient quadruped locomotion, in: Proceedings of the IEEE International Conference on Robotics and Automation (ICRA), IEEE, 2024, pp. 5350–5356. doi:10.1109/ICRA55743.2025.11128812

work page doi:10.1109/icra55743.2025.11128812 2024
[26]

Kormushev, B

P. Kormushev, B. Ugurlu, D. Caldwell, N. Tsagarakis, Learning to exploit passive compliance for energy-eﬃcient gait generation on a compliant humanoid, Autonomous Robots 43 (2018) 79–95. doi:10.1007/s10514-018-9697-6

work page doi:10.1007/s10514-018-9697-6 2018
[27]

Spröwitz, A

A. Spröwitz, A. Tuleu, M. Vespignani, M. Ajallooeian, E. Badri, A. Ijspeert, Towards dynamic trot gait locomotion: Design, control, and experiments with cheetah-cub, a compliant quadruped robot, The International Journal of Robotics Research 32 (2013) 932–950

work page 2013
[28]

Y. Sun, C. Zong, F. Pancheri, T. Chen, T. C. Lueth, Design of topology optimized compliant legs for bio-inspired quadruped robots, Scientiﬁc Reports 13 (1) (2023) 4875

work page 2023
[29]

Semini, J

C. Semini, J. Goldsmith, D. Manfredi, F. Calignano, E. P. Ambrosio, J. Pakkanen, D. G. Caldwell, Additive manufacturing for agile legged robots with hydraulic actuation, in: International Conference on Ad- vanced Robotics (ICAR), IEEE, 2015, pp. 123–129

work page 2015
[30]

J. Luo, S. Ye, J. Su, B. Jin, Prismatic quasi-direct-drives for dynamic quadruped locomotion with high payload capacity, International Journal of Mechanical Sciences 235 (2022) 107698

work page 2022
[31]

H. Zong, J. Zhang, L. Jiang, K. Zhang, J. Shen, Z. Lu, K. Wang, Y. Wang, B. Xu, Bionic lightweight design of limb leg units for hydraulic quadruped robots by additive manufacturing and topology optimization, Bio-Design and Manufacturing 7 (2023) 1–13

work page 2023
[32]

H. S. Sucuoğlu, Development of topologically optimized mobile robotic system with machine learning-based energy-eﬃcient path planning structure, Machines 13 (8) (2025) 638. 26

work page 2025
[33]

E. A. Parra Ricaurte, J. Pareja, S. Domínguez, C. Rossi, Comparison of leg dynamic models for quadrupedal robots with compliant backbone, Scientiﬁc Reports 12 (2022) 18536

work page 2022
[34]

X. Zhou, X. Zhang, Q. Zhang, Hac-loco: Learning hierarchical active compliance control for quadruped locomotion under continuous external disturbances, arXiv preprint arXiv:2507.02447 (2025)

work page arXiv 2025
[35]

J. Ding, P. Posthoorn, V. Atanassov, F. Boekel, J. Kober, C. Della Santina, Quadrupedal locomotion with parallel compliance: E- go design, modeling, and control, IEEE/ASME Transactions on Mecha- tronics 29 (4) (2024) 2839–2848

work page 2024
[36]

Lakatos, K

D. Lakatos, K. Ploeger, F. Loeﬄ, D. Seidel, F. Schmidt, T. Gumpert, F. John, T. Bertram, A. Albu-Schäﬀer, Dynamic locomotion gaits of a compliantly actuated quadruped with slip-like articulated legs embodied in the mechanical design, IEEE Robotics and Automation Letters 3 (2018) 3908–3915

work page 2018
[37]

Risbourg, T

F. Risbourg, T. Corbères, P. A. Léziart, T. Flayols, N. Mansard, S. Ton- neau, Real-time footstep planning and control of the solo quadruped robot in 3d environments, in: Proceedings of the IEEE/RSJ Interna- tional Conference on Intelligent Robots and Systems (IROS), IEEE, 2022, pp. 12950–12956

work page 2022
[38]

M. H. Raibert, H. B. Brown Jr, M. Chepponis, Experiments in balance with a 3d one-legged hopping machine, The International Journal of Robotics Research 3 (2) (1984) 75–92

work page 1984
[39]

Farshidian, M

F. Farshidian, M. Neunert, A. W. Winkler, G. Rey, J. Buchli, An eﬃ- cient optimal planning and control framework for quadrupedal locomo- tion, in: Proceedings of the IEEE International Conference on Robotics and Automation (ICRA), IEEE, 2017, pp. 93–100

work page 2017
[40]

O. Cebe, C. Tiseo, G. Xin, H. C. Lin, J. Smith, M. Mistry, Online dynamic trajectory optimization and control for a quadruped robot, in: Proceedings of the IEEE International Conference on Robotics and Automation (ICRA), IEEE, 2021, pp. 12773–12779

work page 2021
[41]

Gehring, C

C. Gehring, C. D. Bellicoso, P. Fankhauser, S. Coros, M. Hutter, Quadrupedal locomotion using trajectory optimization and hierarchical 27 whole body control, in: Proceedings of the IEEE International Confer- ence on Robotics and Automation (ICRA), IEEE, 2017, pp. 4788–4794

work page 2017
[42]

Q. Li, L. Qian, S. Wang, P. Sun, X. Luo, Towards generation and transition of diverse gaits for quadrupedal robots based on trajectory optimization and whole-body impedance control, IEEE Robotics and Automation Letters 8 (4) (2023) 2389–2396

work page 2023
[43]

Humphreys, J

J. Humphreys, J. Li, Y. Wan, H. Gao, C. Zhou, Bio-inspired gait transi- tions for quadruped locomotion, IEEE Robotics and Automation Letters 8 (10) (2023) 6131–6138

work page 2023
[44]

J. Park, J. H. Park, Impedance control of quadruped robot and its impedance characteristic modulation for trotting on irregular terrain, in: Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), IEEE, 2012, pp. 175–180

work page 2012
[45]

H. Xie, M. Ahmadi, J. Shang, Z. Luo, An intuitive approach for quadruped robot trotting based on virtual model control, Proceedings of the Institution of Mechanical Engineers, Part I: Journal of Systems and Control Engineering 229 (4) (2015) 342–355

work page 2015
[46]

Boaventura, J

T. Boaventura, J. Buchli, C. Semini, D. G. Caldwell, Model-based hy- draulic impedance control for dynamic robots, IEEE Transactions on Robotics 31 (6) (2015) 1324–1336

work page 2015
[47]

Horvat, K

T. Horvat, K. Melo, A. J. Ijspeert, Model predictive control based frame- work for com control of a quadruped robot, in: Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), IEEE, 2017, pp. 3372–3378

work page 2017
[48]

Neunert, M

M. Neunert, M. Stäuble, M. Giftthaler, C. D. Bellicoso, J. Carius, C. Gehring, M. Hutter, J. Buchli, Whole-body nonlinear model pre- dictive control through contacts for quadrupeds, IEEE Robotics and Automation Letters 3 (3) (2018) 1458–1465

work page 2018
[49]

Pinosky, I

A. Pinosky, I. Abraham, A. Broad, B. Argall, T. D. Murphey, Hybrid control for combining model-based and model-free reinforcement learn- ing, The International Journal of Robotics Research 42 (6) (2023) 337– 355. 28

work page 2023
[50]

Zhang, L

H. Zhang, L. He, D. Wang, Deep reinforcement learning for real- world quadrupedal locomotion: A comprehensive review, Intelligence & Robotics 2 (3) (2022) 275–297

work page 2022
[51]

G. B. Margolis, P. Agrawal, Walk these ways: Tuning robot control for generalization with multiplicity of behavior, in: Conference on Robot Learning (CoRL), PMLR, 2023, pp. 22–31

work page 2023
[52]

J. E. McKenzie, Design of robotic quadruped legs, Master’s thesis, Mas- sachusetts Institute of Technology, Cambridge, MA, USA (2012). URL https://dspace.mit.edu/handle/1721.1/70444

work page 2012
[53]

Isaac Gym: High Performance GPU-Based Physics Simulation For Robot Learning

V. Makoviychuk, L. Wawrzyniak, Y. Guo, M. Lu, K. Storey, M. Mack- lin, D. Hoeller, N. Rudin, A. Allshire, A. Handa, G. State, Isaac gym: High performance gpu-based physics simulation for robot learning, arXiv preprint arXiv:2108.10470 (2021)

work page internal anchor Pith review Pith/arXiv arXiv 2021
[54]

Rudin, D

N. Rudin, D. Hoeller, P. Reist, M. Hutter, Learning to walk in minutes using massively parallel deep reinforcement learning, in: Conference on Robot Learning (CoRL), PMLR, 2022, pp. 91–100

work page 2022
[55]

Open Robotics, Ros 2 humble hawksbill, https://docs.ros.org/en/ humble/, accessed: 2026-03-20 (2022). 29

work page 2026