Pith. sign in

REVIEW 3 major objections 6 minor 36 references

Integrating Learning-Based Manipulation and Physics-Based Locomotion for Whole-Body Badminton Robot Control

T0 review · 3 major / 6 minor · reviewed 2026-08-16 · deepseek-v4-flash

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

Pith's one-line read This paper claims that a badminton robot should split control: a model-based chassis for safe locomotion and a learning-based arm, warmed up by a privileged model-based teacher, for hitting, reaching 94.5% success against a serving…

desk verdict A real badminton robot with genuinely useful real-robot numbers, but the claims outrun the evidence on sparse rewards and cross-chassis zero-shot, and the chassis/arm coupling deserves sharper scrutiny. read the letter →

arxiv 2504.17771 v2 pith:U74HF3WJ submitted 2025-04-24 cs.RO cs.AIcs.LG

Haochen Wang , Zhiwei Shi , Chengxi Zhu , Yafei Qiao , Cheng Zhang , Fan Yang , Pengjie Ren , Lan Lu
show 1 more author
Dong Xuan
This is my paper · ORCID
classification cs.ROcs.AIcs.LG
keywords badmintonrobothybridcontrolmodel-basedimitationlearningreinforcementsim-to-realtransfermobilemanipulationwhole-body
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

This paper sets out to show that agile badminton play can be controlled by a hybrid whole-body system rather than by a pure model-based or pure learning-based approach. The chassis is driven by a model-based strategy that estimates a reachable base position, while the racket arm is driven by a neural-network policy trained with an imitation-learning-plus-reinforcement-learning recipe supervised by a privileged model-based teacher. The central claim is that this training recipe transfers zero-shot from simulation to the real robot and outperforms the model-based arm strategy, with a 94.5% success rate against a serving machine and a 90.7% return success rate against human players. A careful reader would care because the system shows where learned control earns its keep in fast, safety-critical sports robots: fine manipulation at the end effector, not high-speed locomotion.

What carries the argument

The load-bearing mechanism is the physics-informed 'IL+RL' training recipe centered on a model-based teacher $\hat{\pi}_{\mathrm{PM}}$ with privileged information. In the IL phase, the student maximizes $\mathbb{E}_{s_t \sim \pi_\theta}[\ln \pi_\theta(\hat{\pi}_{\mathrm{PM}}(s_t)\mid s_t)]$ while a critic $V_\phi$ is fit to discounted returns, and in the RL phase the PPO objective is augmented by the teacher-supervision term $\lambda J_{\mathrm{SUP}}(\theta)$. The recipe's role is to replace reward shaping and curriculum learning with a warm start and soft exploration boundaries, so that sparse rewards suffice. The other supporting object is the rigid transformation of the ball trajectory, translation by $p_{\mathrm{base}}$ and rotation by chassis orientation $\alpha$, which lets the arm policy act as though the chassis has already arrived at a known base position.

What would settle it

Run the serving-machine protocol again while measuring chassis end-position error for every rally, then compare that error distribution with the simulated noise used in training; test the arm with artificially shifted $p_{\mathrm{base}}$ values matching the tail of the measured distribution. If success rate drops below the reported 94.5% under those shifts, the zero-shot transfer claim depends on chassis accuracy rather than on the arm policy alone.

Watch

Extended reading notes

Core claim

On the paper's own terms, the discovery is that a strong model-based racquet-sports strategy, given privileged access to the complete badminton trajectory in simulation, can act as a teacher that makes a sparse-reward neural arm policy trainable. During the imitation phase, DAgger matches the teacher's actions on states visited by the student while a critic is trained on balanced positive and negative samples; during the reinforcement phase, PPO's objective is augmented with the teacher's supervision so exploration stays within soft boundaries. The resulting arm policy, fed with ball trajectories rigidly transformed into the robot frame using the chassis's estimated base position and orientation, hits 97% of balls and returns 94.5% into the designated area, versus 91% and 77.5% for the model-based arm strategy, and sustains up to a 40-shot rally with human players.

Load-bearing premise

The load-bearing premise is that the model-based chassis actually reaches its estimated base position and orientation closely enough that the ball trajectory, transformed into that frame, is what the arm policy expects; if chassis tracking errors exceed the noise injected during training, the policy acts on distorted inputs and the reported success rates could fall.

Editorial extensions

If this is right

  • If the system works as reported, a robot can swap its mobile chassis without retraining the arm policy, because the arm only sees the ball trajectory in a chassis-centered frame.
  • The physics-informed teacher recipe offers a reusable alternative to reward shaping for other sparse-reward agile manipulation tasks, such as table tennis and high-speed catching.
  • The observed margin over the model-based arm strategy (94.5% vs 77.5% success) suggests that learned fine manipulation can compensate for modeling errors in ballistic and aerodynamic shuttlecock prediction.
  • Fielding a learned arm on a model-based chassis preserves the safety and stability of locomotion while gaining the adaptability of learning where it matters.
  • Because the arm policy is trained in simulation with chassis noise and then deployed without real-world fine-tuning, the system is a direct demonstration that zero-shot sim-to-real transfer can work for agile racket sports.

Reading between the lines

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

  • A testable extension implied by the paper is to deliberately bias $p_{\mathrm{base}}$ by offsets matching the real chassis error distribution; if success rate falls below the reported 94.5%, the zero-shot claim depends on chassis accuracy rather than on the arm policy alone.
  • The same base-relative frame transform could extend the recipe to other mobile manipulators, but only if the base pose estimate is accurate; the policy's inputs are already base-relative, so the estimate's precision gates performance.
  • The teacher's privileged trajectory sets the IL ceiling, so in regimes where the model-based strategy is inaccurate the RL phase must do more than fine-tune; the reported margin may not hold where ballistic prediction is poor.
  • A reader could probe whether critic warm-up alone, without teacher supervision, recovers the IL-to-RL drop on a simpler agile task; the paper's ablation suggests the teacher supervision is the stronger factor.
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

3 major / 6 minor

Summary. Hamlet combines a model-based chassis controller with a learned arm policy for a wheeled badminton robot. The arm policy is trained in simulation via a DAgger-based imitation phase followed by PPO, with a privileged model-based teacher supervising both phases and a critic warm-up during IL. The authors report real-robot results: 94.5% success rate against a serving machine, 90.7% average return success rate against humans, a maximum rally of 40, and superior performance over a model-based arm controller (97% vs 91% hit rate, 94.5% vs 77.5% success rate). They also claim zero-shot sim2real transfer and portability to different chassis without arm-policy retraining.

Significance. If the reported results hold, the architecture is a valuable demonstration of decoupling locomotion and manipulation for agile sports robots, and the IL+RL recipe with a privileged model-based teacher offers a practical route that avoids heavy reward shaping. The real-robot evaluation is a genuine strength: 200 serving-machine trials per controller and 300 human-robot rounds are substantive empirical evidence, and the comparison against a model-based arm policy is a meaningful baseline. The main weaknesses are statistical reporting and the sensitivity of the learned arm policy to chassis pose estimation, which currently prevent the strongest claims from being fully supported.

major comments (3)
  1. [§IV-D, Table I] The headline comparison reports 94.5% vs 77.5% success rate and 97% vs 91% hit rate as point estimates over 200 trials per strategy, with no confidence intervals or significance tests. Because the two strategies were tested under the same 20 scenarios, please report binomial confidence intervals for each rate and a paired test such as McNemar for the difference. Without this, the statement in §IV-B that the learning-based policy significantly outperforms the model-based strategy is not quantitatively supported.
  2. [§III-A.1, §IV-A] The arm policy input depends on an estimated chassis pose p_base and orientation alpha through Eq. (1), and §III-A.1 states that the policy operates on the premise that the robot has reached p_base. Yet the gap analysis in §IV-A reports that chassis end-position error is larger and more variable than arm error. The only stated mitigation, noise injection into the simulated chassis (§III-B.1), is not validated against the measured real error distribution, and no sensitivity analysis of policy success versus p_base error is provided. Because the policy never observes the actual chassis pose, systematic chassis tracking error directly warps the input trajectory; please add either a validation that injected noise covers the measured chassis error, a perturbation analysis of success rate versus p_base error, or a real-world ablation showing that the policy tolerates the observed chassis uncertainty.
  3. [§IV-C, Fig. 6] The ablation study reports only mean reward curves, with no variance, number of seeds, or quantitative final success rates. The text claims that variants without critic warm-up or model-based supervision exhibit a significant initial drop and greater volatility, but no statistical evidence is shown. Please report multiple seeds with mean and standard deviation, and final task success rates for each ablated variant, so the contribution of each component can be judged.
minor comments (6)
  1. [§IV-D, Fig. 4] The text says 20 scenarios are created from combinations of three positions, three angles, and two strengths (3×3×2=18), while the Fig. 4 caption says six positions; please reconcile these numbers.
  2. [Abstract, §I] There are typos in the abstract and introduction: 'excel control policies' should be 'excellent control policies', and 'In this train framework' should be 'In this training framework'.
  3. [§III-B.3] The sentence 'where λ is a The reward function' is incomplete and should be rewritten.
  4. [Table I, §IV-D] The caption and text do not make clear how many trials correspond to each entry and whether the 20 scenarios are balanced; state sample sizes and scenario breakdown explicitly in the table.
  5. [§IV-A, Fig. 5] The 194 control trajectories used for the gap analysis are not described; state how they were generated and how the end-distance errors in Fig. 5 were computed.
  6. [§IV-E, Fig. 7] Define what counts as a round and a successful return in human-robot play, and report per-set results for the three 100-round sets rather than only the aggregate average.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the learned arm policy is trained with a privileged model-based teacher, but the headline results are empirical comparisons against a non-privileged baseline and human/serving-machine opponents, not derivations from the inputs.

full rationale

I walked the claimed derivation chain. The paper's method has three load-bearing pieces: (i) a model-based chassis controller, (ii) a learning-based arm policy whose inputs are the ball trajectory transformed by the estimated chassis pose pbase and orientation alpha, and (iii) an IL+RL training recipe in which a model-based strategy with privileged information (complete trajectory) supervises both the actor and critic during imitation and continues to supervise during PPO. None of these pieces reduces by construction to the paper's headline outputs. The arm policy is not defined in terms of the final success rate; it is a neural network trained to imitate the privileged teacher and then refine via PPO with sparse and dense rewards. The 94.5% serving-machine success rate, 90.7% human-opponent success rate, and 40-shot maximum rally are measured real-world outcomes, not quantities that follow algebraically from the training objective. The model-based teacher is taken from the authors' prior work [13], and the paper explicitly says the chassis strategy is 'suggested by the work [13]' and the privileged teacher is 'developed based on [13]'. This is a self-citation, but it is not load-bearing in a circular sense: the comparison in Table I evaluates the learned policy against a real-time model-based strategy that does not receive privileged trajectory information, and the learned policy is not forced to match the teacher exactly because the RL phase adds J_PPO and exploration beyond J_SUP. The only step that could look like a forced relationship is the observation in Section III-A.1 that 'the arm policy operates on the premise that the robot has reached pbase' and transforms the ball trajectory using pbase and alpha. That is an explicit modeling assumption about input coordinates, and the sim2real gap analysis in Section IV-A even concedes the chassis has larger and more variable error than the arm. But this is a robustness/assumption concern, not circularity: the policy's input being computed from an estimated pose does not make the policy's output identical to its input or to the fitted teacher. No parameter is fitted to the evaluation data and then renamed as a prediction; no uniqueness theorem is imported from the authors' prior work to forbid alternatives; no known empirical pattern is repackaged under new coordinates. The ablation study and the serving-machine/human experiments are independent empirical evidence for the central claim.

Assumptions & free parameters 5 free parameters · 4 assumptions · 0 invented entities

The central claims rest on three practical assumptions: the model-based chassis reliably delivers the base position, the simulator plus noise is faithful enough for zero-shot transfer, and the hand-designed reward and supervision weights are adequate. No free parameters are reported numerically, and the sparse-reward claim is undercut by the dense reward list. The system introduces no invented entities; its novelty is architectural.

free parameters (5)
  • lambda (supervision weight in RL objective)
    Hand-chosen weight in Eq. (8) balancing the PPO loss and the model-based supervision loss; no value or tuning analysis is reported.
  • Reward coefficients for seven reward terms
    Section III-B.3 lists hitting, dynamic constraint, unsafe action, ball position, net crossing, net height, and landing position rewards; coefficients are not specified, and the dense terms contradict the claimed sparse-reward training.
  • Chassis timing constants T_b and T_init
    Section III-A.2 uses T_b to freeze target updates and T_init for the trust factor sigma; both are hand-chosen constants with no reported values.
  • Fixed hitting height H_b
    Section III-A.2 uses H_b to select the chassis target when plenty of time remains; value not reported.
  • Ball trajectory prediction parameters theta = learned from 10,000 collected trajectories
    The model-based predictor F_hat_o has learnable parameters theta used to compute hit time and position; these are fitted to real data, so the overall system depends on them.
assumptions (4)
  • domain assumption The arm policy can treat the chassis position p_base as fixed and correct at hit time.
    Section III-A.1 states 'The arm policy operates on the premise that the robot has reached pbase.' The model-based chassis controller must realize p_base accurately enough; the paper's own sim2real analysis shows the chassis has larger positional error than the arm.
  • domain assumption The PyBullet simulator with injected chassis noise captures real shuttlecock dynamics and racket contact sufficiently for zero-shot transfer.
    Training is done in simulation with real trajectories replayed and noise added (Section IV and III-B.1); no quantitative validation of the simulation gap is provided beyond end-position metrics that already show a chassis sim2real gap.
  • standard math DAgger and PPO converge to useful policies under the stated objectives with the pre-trained critic.
    Algorithm 1 assumes standard convergence and stability properties of DAgger and PPO; the paper offers no new theory for these guarantees.
  • domain assumption The seven reward terms induce correct return behavior without curriculum learning.
    Section III-B.3 defines the reward; the claim that no complex reward shaping is needed rests on this assumption, but dense rewards for ball position, net height, and landing position are already a form of shaping.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Integrating Learning-Based Manipulation and Physics-Based Locomotion for Whole-Body Badminton Robot Control." pith.science (2026). https://pith.science/paper/U74HF3WJ

@misc{pith2026250417771,
  author       = {Pith},
  title        = {Pith review of: Integrating Learning-Based Manipulation and Physics-Based Locomotion for Whole-Body Badminton Robot Control},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/U74HF3WJ}},
  note         = {Machine review of arXiv:2504.17771}
}
read the original abstract

Learning-based methods, such as imitation learning (IL) and reinforcement learning (RL), can produce excel control policies over challenging agile robot tasks, such as sports robot. However, no existing work has harmonized learning-based policy with model-based methods to reduce training complexity and ensure the safety and stability for agile badminton robot control. In this paper, we introduce Hamlet, a novel hybrid control system for agile badminton robots. Specifically, we propose a model-based strategy for chassis locomotion which provides a base for arm policy. We introduce a physics-informed "IL+RL" training framework for learning-based arm policy. In this train framework, a model-based strategy with privileged information is used to guide arm policy training during both IL and RL phases. In addition, we train the critic model during IL phase to alleviate the performance drop issue when transitioning from IL to RL. We present results on our self-engineered badminton robot, achieving 94.5% success rate against the serving machine and 90.7% success rate against human players. Our system can be easily generalized to other agile mobile manipulation tasks such as agile catching and table tennis. Our project website: https://dreamstarring.github.io/HAMLET/.

Figures

Figures reproduced from arXiv: 2504.17771 by the authors.

Figure 1
Figure 1. Agile badminton robot system. The badminton robot on the left, which consists of an omnidirectional chassis and a 5-DOF arm, is playing badminton against a human on the right. The robot can move flexibly in the court and rapidly swing the racket to return the ball. Controlling such robots broadly involves two strategies: model-based and learning-based. Model-based strategies [12], [13] offer both stability and safet… view at source ↗
Figure 2
Figure 2. Overview of the components of Hamlet.The green boxes represent the processing steps, including the detection and prediction of the ball trajectory. The blue boxes represent the model-based strategy for controlling the chassis, as discussed in III-A.2. The dark orange box represents the rigid transformation of ball trajectory coordinates, as discussed in III-A.1. The yellow box represents the learning-based policy fo… view at source ↗
Figure 5
Figure 5. Comparison of the control gap between simulation and real world of chassis and arm. We test the difference between the mobile platform and the robot’s end position in the simulation and real world under the same control commands. Box plot (a) visualizes the end distance error. We show the standard deviation of the end distance error for the chassis and arm in chart (b). The results show that the chassis is more vola… view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Serving machine experiments setup. We set 6 positions for the serving machine on the right side of the court. We configure 20 combinations of machine positions, power, and angle to ensure diverse trajectories. The circles on the left side of the court show the ball lan…
Figure 7
Figure 7. Figure 7: Robot vs. Human Performance. We collect three sets of data from robot-human matches. In each set, human and robot play 100 rounds, and we record the robot’s return success rate and the maximum rally. We find that our system performs well during play with humans. in 200…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

36 extracted references · 30 canonical work pages

  1. [1]

    Human mimetic forearm design with radioul- nar joint using miniature bone-muscle modules and its applications,

    K. Kawaharazuka et al., “Human mimetic forearm design with radioul- nar joint using miniature bone-muscle modules and its applications,” in Proc. IEEE/RSJ Int. Conf. Intell. Robots Syst. , Sep. 2017, pp. 4956– 4962

    work page 2017

  2. [2]

    GROSSMAN

    D. GROSSMAN. (2024) Finally, A Robot That Can Beat You at Badminton. https://www.popularmechanics.com/technology/robots/ a21422/badminton-playing-robot/

    work page 2024

  3. [3]

    Agile Catching with Whole-Body MPC and Blackbox Policy Learning,

    S. Abeyruwan et al. , “Agile Catching with Whole-Body MPC and Blackbox Policy Learning,” in Proc. Learn. Dyn. Control Conf. , ser. Proceedings of Machine Learning Research, N. Matni, M. Morari, and G. J. Pappas, Eds., vol. 211, 2023, pp. 851–863

    work page 2023

  4. [4]

    i-Sim2Real: Reinforcement Learning of Robotic Policies in Tight Human-Robot Interaction Loops,

    S. W. Abeyruwan et al. , “i-Sim2Real: Reinforcement Learning of Robotic Policies in Tight Human-Robot Interaction Loops,” in Proc. Conf. Robot Learn. , ser. Proceedings of Machine Learning Research, K. Liu, D. Kulic, and J. Ichnowski, Eds., vol. 205, 2022, pp. 212–224

    work page 2022

  5. [5]

    Learning high speed precision table tennis on a physical robot,

    T. Ding et al. , “Learning high speed precision table tennis on a physical robot,” in Proc. IEEE/RSJ Int. Conf. Intell. Robots Syst. , Oct. 2022, pp. 10 780–10 787

    work page 2022

  6. [6]

    Learning to play table tennis from scratch using muscular robots,

    D. B ¨uchler, S. Guist, R. Calandra, V . Berenz, B. Sch ¨olkopf, and J. Peters, “Learning to play table tennis from scratch using muscular robots,” IEEE Trans. Robotics , vol. 38, no. 6, pp. 3850–3860, Dec. 2022

    work page 2022

  7. [7]

    Robotic table tennis with model-free reinforcement learning,

    W. Gao et al. , “Robotic table tennis with model-free reinforcement learning,” in Proc. IEEE/RSJ Int. Conf. Intell. Robots Syst. , 2020, pp. 5556–5563

    work page 2020

  8. [8]

    Achieving human level competitive robot table tennis,

    D. B. D’Ambrosio, S. Abeyruwan, L. Graesser, A. Iscen, H. B. Amor, A. Bewley, B. J. Reed, K. Reymann, L. Takayama, Y . Tassa, et al. , “Achieving human level competitive robot table tennis,”arXiv preprint arXiv:2408.03906, 2024

    arXiv 2024

Show all 36 references
  1. [9]

    Opponent hitting behavior prediction and ball location control for a table tennis robot,

    Y . Ji, Y . Mao, F. Suo, X. Hu, Y . Hou, and Y . Yuan, “Opponent hitting behavior prediction and ball location control for a table tennis robot,” Biomimetics, vol. 8, no. 2, p. 229, 2023

    2023

  2. [10]

    Real-time visual measurement with opponent hitting behavior for table tennis robot,

    K. Zhang, Z. Cao, J. Liu, Z. Fang, and M. Tan, “Real-time visual measurement with opponent hitting behavior for table tennis robot,” IEEE Transactions on Instrumentation and Measurement , vol. 67, no. 4, pp. 811–820, 2018

    2018

  3. [11]

    A table-tennis robot control strategy for returning high-speed spinning ball,

    Y . Wang, Y . Luo, H. Zhang, W. Zhang, K. Dong, Q. He, Q. Zhang, E. Cheng, Z. Sun, and B. Song, “A table-tennis robot control strategy for returning high-speed spinning ball,” IEEE/ASME Transactions on Mechatronics, 2023

    2023

  4. [12]

    Development of automatic badminton playing robot with distance image sensor,

    N. Mizuno et al., “Development of automatic badminton playing robot with distance image sensor,” IF AC-PapersOnLine, vol. 52, no. 8, pp. 67–72, 2019

    2019

  5. [13]

    Varsm: Ver- satile autonomous racquet sports machine,

    F. Yang, Z. Shi, S. Ye, J. Qian, W. Wang, and D. Xuan, “Varsm: Ver- satile autonomous racquet sports machine,” in Proc. 13th ACM/IEEE Int. Conf. Cyber-Phys. Syst. , 2022, pp. 203–214

    2022

  6. [14]

    Deep whole-body control: Learning a unified policy for manipulation and locomotion,

    Z. Fu, X. Cheng, and D. Pathak, “Deep whole-body control: Learning a unified policy for manipulation and locomotion,” in Conference on Robot Learning, CoRL 2022, 14-18 December 2022, Auckland, New Zealand, ser. Proceedings of Machine Learning Research, K. Liu, D. Kulic, and J....

    2022

  7. [15]

    Catch the Ball: Accurate High-Speed Motions for Mobile Manipulators via Inverse Dynamics Learning,

    K. Dong, K. Pereida, F. Shkurti, and A. P. Schoellig, “Catch the Ball: Accurate High-Speed Motions for Mobile Manipulators via Inverse Dynamics Learning,” in Proc. IEEE/RSJ Int. Conf. Intell. Robots Syst. , 2020, pp. 6718–6725

    2020

  8. [16]

    Robust Predictive Control for Quadrupedal Locomotion: Learning to Close the Gap Between Reduced- and Full-Order Models,

    A. Pandala, R. T. Fawcett, U. Rosolia, A. D. Ames, and K. A. Hamed, “Robust Predictive Control for Quadrupedal Locomotion: Learning to Close the Gap Between Reduced- and Full-Order Models,” IEEE Robotics Autom. Lett. , vol. 7, no. 3, pp. 6622–6629, Jul. 2022

    2022

  9. [17]

    Combining learning-based locomotion policy with model-based manipulation for legged mobile manipulators,

    Y . Ma, F. Farshidian, T. Miki, J. Lee, and M. Hutter, “Combining learning-based locomotion policy with model-based manipulation for legged mobile manipulators,” IEEE Robotics and Automation Letters , vol. 7, no. 2, pp. 2377–2384, 2022

    2022

  10. [18]

    PIRLNav: Pretraining with imitation and RL finetuning for OBJECTNA V ,

    R. Ramrakhya, D. Batra, E. Wijmans, and A. Das, “PIRLNav: Pretraining with imitation and RL finetuning for OBJECTNA V ,” in Proc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. Workshops, Jun. 2023, pp. 17 896–17 906

    2023

  11. [20]

    Aw-opt: Learning robotic skills with imitation andrein- forcement at scale,

    Y . Lu et al. , “Aw-opt: Learning robotic skills with imitation andrein- forcement at scale,” in Proc. Conf. Robot Learn. , ser. Proceedings of Machine Learning Research, A. Faust, D. Hsu, and G. Neumann, Eds., vol. 164, 2021, pp. 1078–1088

    2021

  12. [21]

    Scalable deep reinforcement learning for vision-based robotic manipulation,

    D. Kalashnikov et al. , “Scalable deep reinforcement learning for vision-based robotic manipulation,” in Proc. Conf. Robot Learn. , ser. Proceedings of Machine Learning Research, vol. 87, Oct. 2018, pp. 651–673

    2018

  13. [22]

    Learning complex dexterous manipulation with deep reinforcement learning and demonstrations,

    A. Rajeswaran et al., “Learning complex dexterous manipulation with deep reinforcement learning and demonstrations,” in Proc. Robot.: Sci. Syst., H. Kress-Gazit, S. S. Srinivasa, T. Howard, and N. Atanasov, Eds., 2018

    2018

  14. [23]

    Learning fine-grained bimanual manipulation with low-cost hardware,

    T. Z. Zhao, V . Kumar, S. Levine, and C. Finn, “Learning fine-grained bimanual manipulation with low-cost hardware,” in Proc. Robot.: Sci. Syst., K. E. Bekris, K. Hauser, S. L. Herbert, and J. Yu, Eds., 2023

    2023

  15. [24]

    Mobile ALOHA: learning bimanual mobile manipulation with low-cost whole-body teleoperation,

    Z. Fu, T. Z. Zhao, and C. Finn, “Mobile ALOHA: learning bimanual mobile manipulation with low-cost whole-body teleoperation,” vol. abs/2401.02117, 2024. [Online]. Available: https://doi.org/10.48550/ arXiv.2401.02117

    2024 doi arXiv

  16. [25]

    Actor-critic alignment for offline-to-online reinforcement learning,

    Z. Yu and X. Zhang, “Actor-critic alignment for offline-to-online reinforcement learning,” in International Conference on Machine Learning. PMLR, 2023, pp. 40 452–40 474

    2023

  17. [26]

    Analyzing and overcoming degradation in warm-start reinforcement learning,

    B. Wexler, E. Sarafian, and S. Kraus, “Analyzing and overcoming degradation in warm-start reinforcement learning,” in 2022 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) . IEEE, 2022, pp. 4048–4055

    2022

  18. [27]

    Cherry-picking with reinforcement learning,

    Y . Zhang, L. Ke, A. Deshpande, A. Gupta, and S. S. Srinivasa, “Cherry-picking with reinforcement learning,” in Robotics: Science and Systems XIX, Daegu, Republic of Korea, July 10-14, 2023 , K. E. Bekris, K. Hauser, S. L. Herbert, and J. Yu, Eds., 2023. [Online]. Available: h...

    2023 doi

  19. [28]

    A Reduction of Imitation Learning and Structured Prediction to No-Regret Online Learning,

    S. Ross, G. J. Gordon, and D. Bagnell, “A Reduction of Imitation Learning and Structured Prediction to No-Regret Online Learning,” in Proc. 14th Int. Conf. Artif. Intell. Statist. JMLR Workshop Conf. Proc. , ser. JMLR Proceedings, G. J. Gordon, D. B. Dunson, and M. Dud ´ık, Ed...

    2011

  20. [30]

    Pybullet, a python module for physics simulation for games, robotics and machine learning,

    E. Coumans and Y . Bai, “Pybullet, a python module for physics simulation for games, robotics and machine learning,” 2016

    2016

  21. [31]

    Multi-camera asynchronous ball localization and trajectory prediction with factor graphs and human poses,

    Q. Xiao, Z. Zaidi, and M. C. Gombolay, “Multi-camera asynchronous ball localization and trajectory prediction with factor graphs and human poses,” vol. abs/2401.17185, 2024. [Online]. Available: https://doi.org/10.48550/arXiv.2401.17185

    2024 doi arXiv

  22. [32]

    TTNet: Real-time temporal and spatial video analysis of table tennis,

    R. V oeikov, N. Falaleev, and R. Baikulov, “TTNet: Real-time temporal and spatial video analysis of table tennis,” in Proc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. Workshops, Jun. 2020, pp. 3866–3874

    2020

  23. [33]

    Robotic table tennis: A case study into a high speed learning system,

    D. B. D’Ambrosio et al. , “Robotic table tennis: A case study into a high speed learning system,” 2023

    2023

  24. [34]

    Model-based trajectory prediction and hitting velocity control for a new table tennis robot,

    Y . Ji et al. , “Model-based trajectory prediction and hitting velocity control for a new table tennis robot,” in Proc. IEEE/RSJ Int. Conf. Intell. Robots Syst. , 2021, pp. 2728–2734

    2021

  25. [35]

    Impedance control and its effects on a humanoid robot playing table tennis,

    R. Xiong, Y . Sun, Q. Zhu, J. Wu, and J. Chu, “Impedance control and its effects on a humanoid robot playing table tennis,” Int. J. Adv. Robot. Syst. , vol. 9, no. 5, p. 178, Nov. 2012

    2012

  26. [36]

    Athletic Mobile Manipulator System for Robotic Wheelchair Tennis,

    Z. Zaidi et al. , “Athletic Mobile Manipulator System for Robotic Wheelchair Tennis,” IEEE Robotics Autom. Lett. , vol. 8, no. 4, pp. 2245–2252, Apr. 2023

    2023

  27. [2017]

    Available: http://arxiv.org/abs/1707.06347

    [Online]. Available: http://arxiv.org/abs/1707.06347

    arXiv

  28. [2020]

    Available: https://arxiv.org/abs/2006.09359

    [Online]. Available: https://arxiv.org/abs/2006.09359

    2006 arXiv

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.