Searcharxiv⌕ Search

arXiv subjects

Peng Xu

Publications and source records attributed to Peng Xu.

At least 235 records · Page 13Linked to original sources
arXiv2023

RBG-Maxwell Framework: Simulation of Collisional Plasma Systems via Coupled Boltzmann-Maxwell equations on GPU

This paper presents the RBG-Maxwell framework, a relativistic collisional plasma simulator on GPUs. We provide detailed discussions on the fundamental equations, numerical algorithms, implementation specifics, and key testing outcomes. The RBG-Maxwell framework is a robust numerical code designed for simulating the evolution of plasma systems through a kinetic approach on large-scale GPUs. It offers easy adaptability to a wide range of physical systems. Given the appropriate initial distributions, particle masses, charges, differential cross-sections, and external forces (which are not confined to electromagnetic forces), the RBG-Maxwell framework can direct the evolution of a particle system from a non-equilibrium state to a thermal state.

physics.plasm-ph↗
arXiv2023

When Super-Resolution Meets Camouflaged Object Detection: A Comparison Study

Super Resolution (SR) and Camouflaged Object Detection (COD) are two hot topics in computer vision with various joint applications. For instance, low-resolution surveillance images can be successively processed by super-resolution techniques and camouflaged object detection. However, in previous work, these two areas are always studied in isolation. In this paper, we, for the first time, conduct an integrated comparative evaluation for both. Specifically, we benchmark different super-resolution methods on commonly used COD datasets, and meanwhile, we evaluate the robustness of different COD models by using COD data processed by SR methods. Our goal is to bridge these two domains, discover novel experimental phenomena, summarize new experim.

cs.CV↗
arXiv2023

A Systematic Approach for Inertial Sensor Calibration of Gravity Recovery Satellites and Its Application to Taiji-1 Mission

High-precision inertial sensors or accelerometers can provide us references of free-falling motions in gravitational field in space. They serve as the key payloads for gravity recovery missions such as the CHAMP, the GRACE-type missions, and the planned Next Generation Gravity Missions. In this work, a systematic method of electrostatic inertial sensor calibrations for gravity recovery satellites is suggested, which is applied to and verified with the Taiji-1 mission. With this method, the complete operating parameters including the scale factors, the center of mass offset vector and the intrinsic biased acceleration can be precisely calibrated with only two sets of short-term in-orbit experiments. Taiji-1 is the first technology demonstration satellite of the "Taiji Program in Space", which, in its final extended phase in 2022, could be viewed as operating in the mode of a high-low satellite-to-satellite tracking gravity mission. Based on the calibration principles, swing maneuvers with time span about 200 s and rolling maneuvers for 19 days were conducted by Taiji-1 in 2022. The inertial sensor's operating parameters are precisely re-calibrated with Kalman filters and are updated to the Taiji-1 science team. Data from one of the sensitive axis is re-processed with the updated operating parameters, and the performance is found to be slightly improved compared with former results. This approach could be of high reference value for the accelerometer or inertial sensor calibrations of the GFO, the Chinese GRACE-type mission, and the Next Generation Gravity Missions. This could also shed some light on the in-orbit calibrations of the ultra-precision inertial sensors for future GW space antennas because of the technological inheritance between these two generations of inertial sensors.

physics.ins-det↗
arXiv2023

RT-2: Vision-Language-Action Models Transfer Web Knowledge to Robotic Control

We study how vision-language models trained on Internet-scale data can be incorporated directly into end-to-end robotic control to boost generalization and enable emergent semantic reasoning. Our goal is to enable a single end-to-end trained model to both learn to map robot observations to actions and enjoy the benefits of large-scale pretraining on language and vision-language data from the web. To this end, we propose to co-fine-tune state-of-the-art vision-language models on both robotic trajectory data and Internet-scale vision-language tasks, such as visual question answering. In contrast to other approaches, we propose a simple, general recipe to achieve this goal: in order to fit both natural language responses and robotic actions into the same format, we express the actions as text tokens and incorporate them directly into the training set of the model in the same way as natural language tokens. We refer to such category of models as vision-language-action models (VLA) and instantiate an example of such a model, which we call RT-2. Our extensive evaluation (6k evaluation trials) shows that our approach leads to performant robotic policies and enables RT-2 to obtain a range of emergent capabilities from Internet-scale training. This includes significantly improved generalization to novel objects, the ability to interpret commands not present in the robot training data (such as placing an object onto a particular number or icon), and the ability to perform rudimentary reasoning in response to user commands (such as picking up the smallest or largest object, or the one closest to another object). We further show that incorporating chain of thought reasoning allows RT-2 to perform multi-stage semantic reasoning, for example figuring out which object to pick up for use as an improvised hammer (a rock), or which type of drink is best suited for someone who is tired (an energy drink).

cs.RO↗
arXiv2023

Extrinsic Manipulation on a Support Plane by Learning Regrasping

Extrinsic manipulation, a technique that enables robots to leverage extrinsic resources for object manipulation, presents practical yet challenging scenarios. Particularly in the context of extrinsic manipulation on a supporting plane, regrasping becomes essential for achieving the desired final object poses. This process involves sequential operation steps and stable placements of objects, which provide grasp space for the robot. To address this challenge, we focus on predicting diverse placements of objects on the plane using deep neural networks. A framework that comprises orientation generation, placement refinement, and placement discrimination stages is proposed, leveraging point clouds to obtain precise and diverse stable placements. To facilitate training, a large-scale dataset is constructed, encompassing stable object placements and contact information between objects. Through extensive experiments, our approach is demonstrated to outperform the start-of-the-art, achieving an accuracy rate of 90.4\% and a diversity rate of 81.3\% in predicted placements. Furthermore, we validate the effectiveness of our approach through real-robot experiments, demonstrating its capability to compute sequential pick-and-place steps based on the predicted placements for regrasping objects to goal poses that are not readily attainable within a single step. Videos and dataset are available at https://sites.google.com/view/pmvlr2022/.

cs.RO↗
arXiv2023

Advances in Deep Concealed Scene Understanding

Concealed scene understanding (CSU) is a hot computer vision topic aiming to perceive objects exhibiting camouflage. The current boom in terms of techniques and applications warrants an up-to-date survey. This can help researchers to better understand the global CSU field, including both current achievements and remaining challenges. This paper makes four contributions: (1) For the first time, we present a comprehensive survey of deep learning techniques aimed at CSU, including a taxonomy, task-specific challenges, and ongoing developments. (2) To allow for an authoritative quantification of the state-of-the-art, we offer the largest and latest benchmark for concealed object segmentation (COS). (3) To evaluate the generalizability of deep CSU in practical scenarios, we collect the largest concealed defect segmentation dataset termed CDS2K with the hard cases from diversified industrial scenarios, on which we construct a comprehensive benchmark. (4) We discuss open problems and potential research directions for CSU. Our code and datasets are available at https://github.com/DengPingFan/CSU, which will be updated continuously to watch and summarize the advancements in this rapidly evolving field.

cs.CV↗
arXiv2023

Probing Supermassive Black Hole Binaries with Orbital Resonances of Laser-Ranged Satellite

Coalescing supermassive black hole binaries (SMBHBs) are the primary source candidates for low frequency gravitational wave (GW) detections, which could bring us deep insights into galaxy evolutions over cosmic time and violent processes of spacetime dynamics. Promising candidates had been found based on optical and X-ray observations, which claims for new and ready-to-use GW detection approaches before the operations of space-borne antennas. We show that, satellite laser ranging (SLR) missions could serve as probes of coalescing SMBHBs through the GW-induced resonant effects. Lasting and characteristic imprints caused by such resonances in the residual distances or accelerations from SLR measurements are studied, and the detection SNR is analyzed with both the current and future improved ranging precisions. Within redshift $z \sim 1$, the threshold SNR=5 requires 1-2 years of accumulated data for the current precision and months of data for improved precision, which are workable for the data processing of SLR missions. Meanwhile, joint detections with multiple SLR missions could further improve the total SNR and the confidence level. Such a detection scheme could fulfill the requirement of a tentative SMBHB probe during the preparing stage of LISA and Taiji, and it requires no further investment to any new and advanced facilities. It is also worthwhile to look back and re-process the archived data from the past decades, in where resonant signals from SMBHBs might be hidden.

gr-qc↗
arXiv2023

Do Not Train It: A Linear Neural Architecture Search of Graph Neural Networks

Neural architecture search (NAS) for Graph neural networks (GNNs), called NAS-GNNs, has achieved significant performance over manually designed GNN architectures. However, these methods inherit issues from the conventional NAS methods, such as high computational cost and optimization difficulty. More importantly, previous NAS methods have ignored the uniqueness of GNNs, where GNNs possess expressive power without training. With the randomly-initialized weights, we can then seek the optimal architecture parameters via the sparse coding objective and derive a novel NAS-GNNs method, namely neural architecture coding (NAC). Consequently, our NAC holds a no-update scheme on GNNs and can efficiently compute in linear time. Empirical evaluations on multiple GNN benchmark datasets demonstrate that our approach leads to state-of-the-art performance, which is up to $200\times$ faster and $18.8\%$ more accurate than the strong baselines.

cs.LG↗
arXiv2023

Language to Rewards for Robotic Skill Synthesis

Large language models (LLMs) have demonstrated exciting progress in acquiring diverse new capabilities through in-context learning, ranging from logical reasoning to code-writing. Robotics researchers have also explored using LLMs to advance the capabilities of robotic control. However, since low-level robot actions are hardware-dependent and underrepresented in LLM training corpora, existing efforts in applying LLMs to robotics have largely treated LLMs as semantic planners or relied on human-engineered control primitives to interface with the robot. On the other hand, reward functions are shown to be flexible representations that can be optimized for control policies to achieve diverse tasks, while their semantic richness makes them suitable to be specified by LLMs. In this work, we introduce a new paradigm that harnesses this realization by utilizing LLMs to define reward parameters that can be optimized and accomplish variety of robotic tasks. Using reward as the intermediate interface generated by LLMs, we can effectively bridge the gap between high-level language instructions or corrections to low-level robot actions. Meanwhile, combining this with a real-time optimizer, MuJoCo MPC, empowers an interactive behavior creation experience where users can immediately observe the results and provide feedback to the system. To systematically evaluate the performance of our proposed method, we designed a total of 17 tasks for a simulated quadruped robot and a dexterous manipulator robot. We demonstrate that our proposed method reliably tackles 90% of the designed tasks, while a baseline using primitive skills as the interface with Code-as-policies achieves 50% of the tasks. We further validated our method on a real robot arm where complex manipulation skills such as non-prehensile pushing emerge through our interactive system.

cs.RO↗
arXiv2023

LVLM-eHub: A Comprehensive Evaluation Benchmark for Large Vision-Language Models

Large Vision-Language Models (LVLMs) have recently played a dominant role in multimodal vision-language learning. Despite the great success, it lacks a holistic evaluation of their efficacy. This paper presents a comprehensive evaluation of publicly available large multimodal models by building a LVLM evaluation Hub (LVLM-eHub). Our LVLM-eHub consists of $8$ representative LVLMs such as InstructBLIP and MiniGPT-4, which are thoroughly evaluated by a quantitative capability evaluation and an online arena platform. The former evaluates $6$ categories of multimodal capabilities of LVLMs such as visual question answering and embodied artificial intelligence on $47$ standard text-related visual benchmarks, while the latter provides the user-level evaluation of LVLMs in an open-world question-answering scenario. The study reveals several innovative findings. First, instruction-tuned LVLM with massive in-domain data such as InstructBLIP heavily overfits many existing tasks, generalizing poorly in the open-world scenario. Second, instruction-tuned LVLM with moderate instruction-following data may result in object hallucination issues (i.e., generate objects that are inconsistent with target images in the descriptions). It either makes the current evaluation metric such as CIDEr for image captioning ineffective or generates wrong answers. Third, employing a multi-turn reasoning evaluation framework can mitigate the issue of object hallucination, shedding light on developing an effective pipeline for LVLM evaluation. The findings provide a foundational framework for the conception and assessment of innovative strategies aimed at enhancing zero-shot multimodal techniques. Our LVLM-eHub will be available at https://github.com/OpenGVLab/Multi-Modality-Arena

cs.CV↗
arXiv2023

Enhanced measurement precision with continuous interrogation during dynamical decoupling

Dynamical decoupling (DD) is normally ineffective when applied to DC measurement. In its straightforward implementation, DD nulls out DC signal as well while suppressing noise. This work proposes a phase relay method (PRM) that is capable of continuously interrogating the DC signal over many DD cycles. We illustrate its efficacy when applied to measurement of a weak DC magnetic field with an atomic spinor Bose-Einstein condensate. Sensitivities approaching standard quantum limit (SQL) or Heisenberg limit (HL) are potentially realizable for a coherent spin state (CSS) or a squeezed spin state (SSS) of 10,000 atoms respectively, while ambient laboratory level noise is suppressed by DD. Our work offers a practical approach to mitigate the limitations of DD to DC measurement and will like find other applications for resorting coherence in quantum sensing and quantum information processing research.

quant-ph↗
arXiv2023

DiffRate : Differentiable Compression Rate for Efficient Vision Transformers

Token compression aims to speed up large-scale vision transformers (e.g. ViTs) by pruning (dropping) or merging tokens. It is an important but challenging task. Although recent advanced approaches achieved great success, they need to carefully handcraft a compression rate (i.e. number of tokens to remove), which is tedious and leads to sub-optimal performance. To tackle this problem, we propose Differentiable Compression Rate (DiffRate), a novel token compression method that has several appealing properties prior arts do not have. First, DiffRate enables propagating the loss function's gradient onto the compression ratio, which is considered as a non-differentiable hyperparameter in previous work. In this case, different layers can automatically learn different compression rates layer-wisely without extra overhead. Second, token pruning and merging can be naturally performed simultaneously in DiffRate, while they were isolated in previous works. Third, extensive experiments demonstrate that DiffRate achieves state-of-the-art performance. For example, by applying the learned layer-wise compression rates to an off-the-shelf ViT-H (MAE) model, we achieve a 40% FLOPs reduction and a 1.5x throughput improvement, with a minor accuracy drop of 0.16% on ImageNet without fine-tuning, even outperforming previous methods with fine-tuning. Codes and models are available at https://github.com/OpenGVLab/DiffRate.

cs.CV↗
arXiv2023

BiAIT*: Symmetrical Bidirectional Optimal Path Planning with Adaptive Heuristic

Adaptively Informed Trees (AIT*) is an algorithm that uses the problem-specific heuristic to avoid unnecessary searches, which significantly improves its performance, especially when collision checking is expensive. However, the heuristic estimation in AIT* consumes lots of computational resources, and its asymmetric bidirectional searching strategy cannot fully exploit the potential of the bidirectional method. In this article, we propose an extension of AIT* called BiAIT*. Unlike AIT*, BiAIT* uses symmetrical bidirectional search for both the heuristic and space searching. The proposed method allows BiAIT* to find the initial solution faster than AIT*, and update the heuristic with less computation when a collision occurs. We evaluated the performance of BiAIT* through simulations and experiments, and the results show that BiAIT* can find the solution faster than state-of-the-art methods. We also analyze the reasons for the different performances between BiAIT* and AIT*. Furthermore, we discuss two simple but effective modifications to fully exploit the potential of the adaptively heuristic method.

cs.RO↗
arXiv2023

Code as Policies: Language Model Programs for Embodied Control

Large language models (LLMs) trained on code completion have been shown to be capable of synthesizing simple Python programs from docstrings [1]. We find that these code-writing LLMs can be re-purposed to write robot policy code, given natural language commands. Specifically, policy code can express functions or feedback loops that process perception outputs (e.g.,from object detectors [2], [3]) and parameterize control primitive APIs. When provided as input several example language commands (formatted as comments) followed by corresponding policy code (via few-shot prompting), LLMs can take in new commands and autonomously re-compose API calls to generate new policy code respectively. By chaining classic logic structures and referencing third-party libraries (e.g., NumPy, Shapely) to perform arithmetic, LLMs used in this way can write robot policies that (i) exhibit spatial-geometric reasoning, (ii) generalize to new instructions, and (iii) prescribe precise values (e.g., velocities) to ambiguous descriptions ("faster") depending on context (i.e., behavioral commonsense). This paper presents code as policies: a robot-centric formulation of language model generated programs (LMPs) that can represent reactive policies (e.g., impedance controllers), as well as waypoint-based policies (vision-based pick and place, trajectory-based control), demonstrated across multiple real robot platforms. Central to our approach is prompting hierarchical code-gen (recursively defining undefined functions), which can write more complex code and also improves state-of-the-art to solve 39.8% of problems on the HumanEval [1] benchmark. Code and videos are available at https://code-as-policies.github.io

cs.RO↗
arXiv2023

Baseband control of superconducting qubits with shared microwave drives

Accurate control of qubits is the central requirement for building functional quantum processors. For the current superconducting quantum processor, high-fidelity control of qubits is mainly based on independently calibrated microwave pulses, which could differ from each other in frequencies, amplitudes, and phases. With this control strategy, the needed physical source could be challenging, especially when scaling up to large-scale quantum processors is considered. Inspired by Kane's proposal for spin-based quantum computing, here, we explore theoretically the possibility of baseband flux control of superconducting qubits with only shared and always-on microwave drives. In our strategy, qubits are by default far detuned from the drive during system idle periods, qubit readout and baseband flux-controlled two-qubit gates can thus be realized with minimal impacts from the always-on drive. By contrast, during working periods, qubits are tuned on resonance with the drive and single-qubit gates can be realized. Therefore, universal qubit control can be achieved with only baseband flux pulses and always-on shared microwave drives. We apply this strategy to the qubit architecture where tunable qubits are coupled via a tunable coupler, and the analysis shows that high-fidelity qubit control is possible. Besides, the baseband control strategy needs fewer physical resources, such as control electronics and cooling power in cryogenic systems, than that of microwave control. More importantly, the flexibility of baseband flux control could be employed for addressing the non-uniformity issue of superconducting qubits, potentially allowing the realization of multiplexing and cross-bar technologies and thus controlling large numbers of qubits with fewer control lines. We thus expect that baseband control with shared microwave drives can help build large-scale superconducting quantum processors.

quant-ph↗
arXiv2023

Scalable Heteronuclear Architecture of Neutral Atoms Based on EIT

Based on our recent paper [arXiv:2206.12176 (2022)], we propose a scalable heteronuclear architecture of parallel implementation of CNOT gates in arrays of alkali-metal neutral atoms for quantum information processing. We considered a scheme where we perform CNOT gates in a parallel manner within the array, while they are performed sequentially between the pairs of neighboring qubits by coherently transporting an array of atoms of one atomic species (ancilla qubits) using an array of mobile optical dipole traps generated by a 2D acousto-optic deflector (AOD). The atoms of the second atomic species (data qubits) are kept in the array of static optical dipole traps generated by spatial light modulator (SLM). The moving ancillas remain in the superposition of their logical ground states without loss of coherence, while their transportation paths avoid overlaps with the spatial positions of data atoms. We numerically optimized the system parameters to achieve the fidelity for parallelly implemented CNOT gates around $\mathcal{F}=95\%$ for the experimentally feasible conditions. Our design can be useful for implementation of surface codes for quantum error correction. Renyi entropy and mutual information are also investigated to characterize the gate performance.

quant-ph↗
arXiv2023

Floquet Dynamical Decoupling at Zero Bias

Dynamical decoupling (DD) is an efficient method to decouple systems from environmental noises and to prolong the coherence time of systems. In contrast to discrete and continuous DD protocols in the presence of bias field, we propose a Floquet DD at zero bias to perfectly suppress both the zeroth and first orders of noises according to the Floquet theory. Specifically, we demonstrate the effectiveness of this Floquet DD protocol in two typical systems including a spinor atomic Bose-Einstein condensate decohered by classical stray magnetic fields and a semiconductor quantum dot electron spin coupled to nuclear spins. Furthermore, our protocol can be used to sense high-frequency noises. The Floquet DD protocol we propose shines new light on low-cost and high-portable DD technics without bias field and with low controlling power, which may have wide applications in quantum computing, quantum sensing, nuclear magnetic resonance and magnetic resonance imaging.

cond-mat.quant-gas↗
arXiv2023

Unsupervised noise reductions for gravitational reference sensors or accelerometers based on Noise2Noise method

Onboard electrostatic suspension inertial sensors are important applications for gravity satellites and space gravitational wave detection missions, and it is important to suppress noise in the measurement signal. Due to the complex coupling between the working space environment and the satellite platform, the process of noise generation is extremely complex, and traditional noise modeling and subtraction methods have certain limitations. With the development of deep learning, applying it to high-precision inertial sensors to improve the signal-to-noise ratio is a practically meaningful task. Since there is a single noise sample and unknown true value in the measured data in orbit, odd-even sub-samplers and periodic sub-samplers are designed to process general signals and periodic signals, and adds reconstruction layers consisting of fully connected layers to the model. Experimental analysis and comparison are conducted based on simulation data, GRACE-FO acceleration data and Taiji-1 acceleration data. The results show that the deep learning method is superior to traditional data smoothing processing sol

physics.geo-ph↗