arXiv subjects

Xiaoyi Hu

Publications and source records attributed to Xiaoyi Hu.

6 recordsLinked to original sources

arXiv2025

Keep on Going: Learning Robust Humanoid Motion Skills via Selective Adversarial Training

Humanoid robots are expected to operate reliably over long horizons while executing versatile whole-body skills. Yet Reinforcement Learning (RL) motion policies typically lose stability under prolonged operation, sensor/actuator noise, and real world disturbances. In this work, we propose a Selective Adversarial Attack for Robust Training (SA2RT) to enhance the robustness of motion skills. The adversary is learned to identify and sparsely perturb the most vulnerable states and actions under an attack-budget constraint, thereby exposing true weakness without inducing conservative overfitting. The resulting non-zero sum, alternating optimization continually strengthens the motion policy against the strongest discovered attacks. We validate our approach on the Unitree G1 humanoid robot across perceptive locomotion and whole-body control tasks. Experimental results show that adversarially trained policies improve the terrain traversal success rate by 40%, reduce the trajectory tracking error by 32%, and maintain long horizon mobility and tracking performance. Together, these results demonstrate that selective adversarial attacks are an effective driver for learning robust, long horizon humanoid motion skills.

cs.RO↗

arXiv2025

Catching Spinning Table Tennis Balls in Simulation with End-to-End Curriculum Reinforcement Learning

The game of table tennis is renowned for its extremely high spin rate, but most table tennis robots today struggle to handle balls with such rapid spin. To address this issue, we have contributed a series of methods, including: 1. Curriculum Reinforcement Learning (RL): This method helps the table tennis robot learn to play table tennis progressively from easy to difficult tasks. 2. Analysis of Spinning Table Tennis Ball Collisions: We have conducted a physics-based analysis to generate more realistic trajectories of spinning table tennis balls after collision. 3. Definition of Trajectory States: The definition of trajectory states aids in setting up the reward function. 4. Selection of Valid Rally Trajectories: We have introduced a valid rally trajectory selection scheme to ensure that the robot's training is not influenced by abnormal trajectories. 5. Reality-to-Simulation (Real2Sim) Transfer: This scheme is employed to validate the trained robot's ability to handle spinning balls in real-world scenarios. With Real2Sim, the deployment costs for robotic reinforcement learning can be further reduced. Moreover, the trajectory-state-based reward function is not limited to table tennis robots; it can be generalized to a wide range of cyclical tasks. To validate our robot's ability to handle spinning balls, the Real2Sim experiments were conducted. For the specific video link of the experiment, please refer to the supplementary materials.

cs.RO↗

arXiv2025

Impact of Static Friction on Sim2Real in Robotic Reinforcement Learning

In robotic reinforcement learning, the Sim2Real gap remains a critical challenge. However, the impact of Static friction on Sim2Real has been underexplored. Conventional domain randomization methods typically exclude Static friction from their parameter space. In our robotic reinforcement learning task, such conventional domain randomization approaches resulted in significantly underperforming real-world models. To address this Sim2Real challenge, we employed Actuator Net as an alternative to conventional domain randomization. While this method enabled successful transfer to flat-ground locomotion, it failed on complex terrains like stairs. To further investigate physical parameters affecting Sim2Real in robotic joints, we developed a control-theoretic joint model and performed systematic parameter identification. Our analysis revealed unexpectedly high friction-torque ratios in our robotic joints. To mitigate its impact, we implemented Static friction-aware domain randomization for Sim2Real. Recognizing the increased training difficulty introduced by friction modeling, we proposed a simple and novel solution to reduce learning complexity. To validate this approach, we conducted comprehensive Sim2Sim and Sim2Real experiments comparing three methods: conventional domain randomization (without Static friction), Actuator Net, and our Static friction-aware domain randomization. All experiments utilized the Rapid Motor Adaptation (RMA) algorithm. Results demonstrated that our method achieved superior adaptive capabilities and overall performance.

cs.RO↗

arXiv2024

Gas-phase formation of fullerene/9-hydroxyfluorene cluster cations

In interstellar environment, fullerene species readily react with large molecules (e.g., PAHs and their derivatives) in the gas phase, which may be the formation route of carbon dust grains in space. In this work, the gas-phase ion-molecule collision reaction between fullerene cations (Cn+, n=32, 34, ..., 60) and functionalized PAH molecules (9-hydroxyfluorene, C13H10O) are investigated both experimentally and theoretically. The experimental results show that fullerene/9-hydroxyfluorene cluster cations are efficiently formed, leading to a series of large fullerene/9-hydroxyfluorene cluster cations (e.g., [(C13H10O)C60]+, [(C13H10O)3C58+, and [(C26H18O)(C13H10O)2C48]+). The binding energies and optimized structures of typical fullerene/9-hydroxyfluorene cluster cations were calculated. The bonding ability plays a decisive role in the cluster formation processes. The reaction surfaces, modes and combination reaction sites can result in different binding energies, which represent the relative chemical reactivity. Therefore, the geometry and composition of fullerene/9-hydroxyfluorene cluster cations are complicated. In addition, there is an enhanced chemical reactivity for smaller fullerene cations, which is mainly attributed to the newly formed deformed carbon rings (e.g., 7 C-ring). As part of the coevolution network of interstellar fullerene chemistry, our results suggest that ion-molecule collision reactions contribute to the formation of various fullerene/9-hydroxyfluorene cluster cations in the ISM, providing insights into different chemical reactivity caused by oxygenated functional groups (e.g., hydroxyl, OH, or ether, C-O-C) on the cluster formations.

astro-ph.GA↗

arXiv2024

Gas-phase hydrogenation of large, astronomically relevant PAH cations

To investigate the gas-phase hydrogenation processes of large, astronomically relevant cationic polycyclic aromatic hydrocarbon (PAH) molecules under the interstellar environments, the ion-molecule collision reaction between six PAH cations and H-atoms is studied. The experimental results show that the hydrogenated PAH cations are efficiently formed, and no even-odd hydrogenated mass patterns are observed in the hydrogenation processes. The structure of newly formed hydrogenated PAH cations and the bonding energy for the hydrogenation reaction pathways are investigated with quantum theoretical calculations. The exothermic energy for each reaction pathway is relatively high, and the competition between hydrogenation and dehydrogenation is confirmed. From the theoretical calculation, the bonding ability plays an important role in the gas-phase hydrogenation processes. The factors that affect the hydrogenation chemical reactivity are discussed, including the effect of carbon skeleton structure, the side-edged structure, the molecular size, the five- and six-membered C-ring structure, the bay region structure, and the neighboring hydrogenation. The IR spectra of hydrogenated PAH cations are also calculated. These results we obtain once again validate the complexity of hydrogenated PAH molecules, and provide the direction for the simulations and observations under the coevolution interstellar chemistry network. We infer that if we do not consider other chemical evolution processes (e.g., photo-evolution), then the hydrogenation states and forms of PAH compounds are intricate and complex in the interstellar medium (ISM).

astro-ph.GA↗

arXiv2020

Laboratory study of the formation of fullerene (from smaller to larger, C$_{44}$ to C$_{70}$)/anthracene cluster cations in the gas phase

The formation and evolution mechanism of fullerenes in the planetary nebula or in the interstellar medium are still not understood. Here we present the study on the cluster formation and the relative reactivity of fullerene cations (from smaller to larger, C$_{44}$ to C$_{70}$) with anthracene molecule (C$_{14}$H$_{10}$). The experiment is performed in the apparatus that combines a quadrupole ion trap with a time-of-flight mass spectrometer. By using a 355 nm laser beam to irradiate the trapped fullerenes cations (C$_{60}$$^+$ or C$_{70}$$^+$), smaller fullerene cations C$_{(60-2n)}$$^+$, n=1-8 or C$_{(70-2m)}$$^+$, m=1-11 are generated, respectively. Then reacting with anthracene molecules, series of fullerene/anthracene cluster cations are newly formed (e.g., (C$_{14}$H$_{10}$)C$_{(60-2n)}$$^+$, n=1-8 and (C$_{14}$H$_{10}$)C$_{(70-2m)}$$^+$, m=1-11), and slight difference of the reactivity within the smaller fullerene cations are observed. Nevertheless, smaller fullerenes show obviously higher reactivity when comparing to fullerene C$_{60}$$^+$ and C$_{70}$$^+$. A successive loss of C$_2$ fragments mechanism is suggested to account for the formation of smaller fullerene cations, which then undergo addition reaction with anthracene molecules to form the fullerene-anthracene cluster cations. It is found that the higher laser energy and longer irradiation time are key factors that affect the formation of smaller fullerene cations. This may indicate that in the strong radiation field environment (such as photon-dominated regions) in space, fullerenes are expected to follow the top-down evolution route, and then form small grain dust (e.g., clusters) through collision reaction with co-existing molecules, here, smaller PAHs.

physics.atm-clus↗