PyTorchKR Papers

최신 논문

총 403편
태그·기간 등 상세 필터

카테고리 탐색

로그인 후 이 태그 구독

arXiv:2502.01546

Dynamic object goal pushing with mobile manipulators through model-free constrained reinforcement learning

Non-prehensile pushing to move and reorient objects to a goal is a versatile loco-manipulation skill. In the real world, the object's physical properties and friction with the floor contain significant uncertainties, which makes the task challenging for a mobile manipulator. In this paper, we develop a learning-based controller for a mobile manipulator to move an unknown object to a desired position and yaw orientation through a sequence of pushing actions. The proposed controller for the robotic arm and the mobile base motion is trained using a constrained Reinforcement Learning (RL) formulation. We demonstrate its capability in experiments with a quadrupedal robot equipped with an arm. The learned policy achieves a success rate of 91.35% in simulation and at least 80% on hardware in challenging scenarios. Through our extensive hardware experiments, we show that the approach demonstrates high robustness against unknown objects of different masses, materials, sizes, and shapes. It reactively discovers the pushing location and direction, thus achieving contact-rich behavior while observing only the pose of the object. Additionally, we demonstrate the adaptive behavior of the learned policy towards preventing the object from toppling.

arXiv:2501.17348

Better Slow than Sorry: Introducing Positive Friction for Reliable Dialogue Systems

While theories of discourse and cognitive science have long recognized the value of unhurried pacing, recent dialogue research tends to minimize friction in conversational systems. Yet, frictionless dialogue risks fostering uncritical reliance on AI outputs, which can obscure implicit assumptions and lead to unintended consequences. To meet this challenge, we propose integrating positive friction into conversational AI, which promotes user reflection on goals, critical thinking on system response, and subsequent re-conditioning of AI systems. We hypothesize systems can improve goal alignment, modeling of user mental states, and task success by deliberately slowing down conversations in strategic moments to ask questions, reveal assumptions, or pause. We present an ontology of positive friction and collect expert human annotations on multi-domain and embodied goal-oriented corpora. Experiments on these corpora, along with simulated interactions using state-of-the-art systems, suggest incorporating friction not only fosters accountable decision-making, but also enhances machine understanding of user beliefs and goals, and increases task success rates.

arXiv:2501.15742

Intuition and importance of feedback control through laboratory experiences

This work aims to raise awareness among engineering students from different disciplines on the importance of feedback control. The proposal consists in comparing the performance of different control strategies in a laboratory session, considering Matlab/Simulink simulations of the non-linear pendulum model. First, students attempt to make the pendulum stop at unstable equilibrium by controlling the torque input with a joystick connected to the computer via an Arduino board. Different friction scenarios are considered for students to explore the dissipation in the system response. Then, as a second task, the Arduino is used to introduce the position reference, while students implement different control strategies, such as Bang-Bang, PID (proportional-integral-derivative) and FPID (fractional PID) controllers, analyzing the system response by inspecting the signals in a scope and in a 3D animated model. The dynamic model results as an application of the laws of rotational motion, and the control methods are explained from an intuitive point of view, focusing on the meaning and motivation of the control actions, with the intention to develop intuition about PID and FPID control methods.

arXiv:2501.14586

A sub-structuring approach for model reduction of frictionally clamped thin-walled structures

Thin-walled structures clamped by friction joints, such as aircraft skin panels are exposed to bending-stretching coupling and frictional contact. We propose an original sub-structuring approach, where the system is divided into thin-walled and support regions, so that geometrically nonlinear behavior is relevant only in the former, and nonlinear contact behavior only in the latter. This permits to derive reduced component models, in principle, with available techniques. The Hurty-/Craig-Bampton method, combined with an interface reduction relying on an orthogonal polynomial series, is used to construct the reduction basis for each component. To model geometrically nonlinear behavior, implicit condensation is used, where an original, engineering-oriented proposition is made for the delicate scaling of the static load cases required to estimate the coefficients of the nonlinear terms. The proposed method is validated and its computational performance is assessed for the example of a plate with frictional clamping, using finite element analysis as reference. The numerical results shed light into an interesting mutual interaction: The extent of geometric hardening is limited by the reduced boundary stiffness when more sliding occurs in the clamping. On the other hand, the frictional dissipation is increased by the tangential loading induced by membrane stretching.

arXiv:2501.12833

A coupled FE-BE multi-scale method for the dynamics of jointed structures

The damping of built-up structures stems largely from the microscopic dry frictional interactions in the contact interfaces. The accurate prediction of friction damping has been an important scientific aim of the past several decades. Recent research indicates that very good agreement with vibration measurements is to be expected if the actual contact surface topography is sufficiently well known and finely resolved, and frictional-unilateral interactions are modeled in terms of the Coulomb-Signorini conditions. Resolving all relevant length scales in one finite element model leads to enormous or even prohibitive computation effort and regularization of the set-valued contact laws might be needed to ensure numerical stability. In this work, we propose a multi-scale approach: The stress and deformation field in the contact region is modeled using elastic half-space theory, implemented on a regular and fine grid of boundary elements (BE), so that the compliance matrix can be expressed in closed form. The vibration behavior of the remaining region is described using a relatively coarse finite element (FE) model, which is further reduced via component mode synthesis. The two models are coupled by enforcing compatibility and equilibrium conditions in the far field. The set-valued Coulomb-Signorini conditions are enforced robustly and efficiently using a projected over-relaxation scheme in conjunction with an appropriate active-set strategy. For the S4 beam benchmark, very good agreement with regard to the amplitude-dependent frequency and damping ratio of the first few modes is achieved, while the computation effort is reduced by several orders of magnitude compared to the full-FE reference. The proposed multi-scale method permits a very fine resolution of the contact surface topography without suffering from numerical instability.

arXiv:2501.13878

Eye Gaze as a Signal for Conveying User Attention in Contextual AI Systems

Advanced multimodal AI agents can now collaborate with users to solve challenges in the world. Yet, these emerging contextual AI systems rely on explicit communication channels between the user and system. We hypothesize that implicit communication of the user's interests and intent would reduce friction and improve user experience when collaborating with AI agents. In this work, we explore the potential of wearable eye tracking to convey signals about user attention. We measure the eye tracking signal quality requirements to effectively map gaze traces to physical objects, then conduct experiments that provide visual scanpath history as additional context when querying vision language models. Our results show that eye tracking provides high value as a user attention signal and can convey important context about the user's current task and interests, improving understanding of contextual AI agents.

arXiv:2501.10316공식 코드 · Papers with Code

Know Your Mistakes: Towards Preventing Overreliance on Task-Oriented Conversational AI Through Accountability Modeling

Recent LLMs have enabled significant advancements for conversational agents. However, they are also well known to hallucinate, producing responses that seem plausible but are factually incorrect. On the other hand, users tend to over-rely on LLM-based AI agents, accepting AI's suggestion even when it is wrong. Adding positive friction, such as explanations or getting user confirmations, has been proposed as a mitigation in AI-supported decision-making systems. In this paper, we propose an accountability model for LLM-based task-oriented dialogue agents to address user overreliance via friction turns in cases of model uncertainty and errors associated with dialogue state tracking (DST). The accountability model is an augmented LLM with an additional accountability head that functions as a binary classifier to predict the relevant slots of the dialogue state mentioned in the conversation. We perform our experiments with multiple backbone LLMs on two established benchmarks (MultiWOZ and Snips). Our empirical findings demonstrate that the proposed approach not only enables reliable estimation of AI agent errors but also guides the decoder in generating more accurate actions. We observe around 3% absolute improvement in joint goal accuracy (JGA) of DST output by incorporating accountability heads into modern LLMs. Self-correcting the detected errors further increases the JGA from 67.13 to 70.51, achieving state-of-the-art DST performance. Finally, we show that error correction through user confirmations (friction turn) achieves a similar performance gain, highlighting its potential to reduce user overreliance.

arXiv:2501.03653

Study of Frictional and Impact Transients in Active-Passive Mechanical Pair

We consider an active-passive mechanical pair in which the relative motion of the latter is constrained by the mechanical impact. The system dynamics is described by the previously introduced modeling frameworks of force transition and dissipation through the nonlinear Coulomb friction and structural damping, the later in accord with Hertzian contact theory. The focus of the recent study is on combining both interaction mechanisms, and the detailed experimental evaluation which discloses validity of the modeling assumptions. Such mechanical pair interactions can be found in various mechatronic systems and mechanisms, like for example clutches, backlash elements, sliding items on the shaking and inclining surfaces, conveyor belts and others. This practical study demonstrates and discusses the transients of a vibro-impact dynamics and shows theoretical developments in line with experimental evaluation.

arXiv:2501.04793

A Novel Observer Design for LuGre Friction Estimation and Control

Dynamic components of the friction may directly impact the stability and performance of the motion control systems. The LuGre model is a prevalent friction model utilized to express this dynamic behavior. Since the LuGre model is very comprehensive, friction compensation based on it might be challenging. Inspired by this, we develop a novel observer to estimate and compensate for LuGre friction. Furthermore, we present a Lyapunov stability analysis to show that observer dynamics are asymptotically stable under certain conditions. Compared to its counterparts, the proposed observer constitutes a simple and standalone scheme that can be utilized with arbitrary control inputs in a straightforward way. As a primary difference, the presented observer estimates velocity and uses the velocity error to estimate friction in addition to control input. The extensive simulations revealed that the introduced observer enhances position and velocity tracking performance in the presence of friction.

arXiv:2501.04276

Bridging Adaptivity and Safety: Learning Agile Collision-Free Locomotion Across Varied Physics

Real-world legged locomotion systems often need to reconcile agility and safety for different scenarios. Moreover, the underlying dynamics are often unknown and time-variant (e.g., payload, friction). In this paper, we introduce BAS (Bridging Adaptivity and Safety), which builds upon the pipeline of prior work Agile But Safe (ABS)(He et al.) and is designed to provide adaptive safety even in dynamic environments with uncertainties. BAS involves an agile policy to avoid obstacles rapidly and a recovery policy to prevent collisions, a physical parameter estimator that is concurrently trained with agile policy, and a learned control-theoretic RA (reach-avoid) value network that governs the policy switch. Also, the agile policy and RA network are both conditioned on physical parameters to make them adaptive. To mitigate the distribution shift issue, we further introduce an on-policy fine-tuning phase for the estimator to enhance its robustness and accuracy. The simulation results show that BAS achieves 50% better safety than baselines in dynamic environments while maintaining a higher speed on average. In real-world experiments, BAS shows its capability in complex environments with unknown physics (e.g., slippery floors with unknown frictions, unknown payloads up to 8kg), while baselines lack adaptivity, leading to collisions or. degraded agility. As a result, BAS achieves a 19.8% increase in speed and gets a 2.36 times lower collision rate than ABS in the real world. Videos: https://adaptive-safe-locomotion.github.io.

arXiv:2412.18840공식 코드 · Papers with Code

Implicit factorized transformer approach to fast prediction of turbulent channel flows

Transformer neural operators have recently become an effective approach for surrogate modeling of systems governed by partial differential equations (PDEs). In this paper, we introduce a modified implicit factorized transformer (IFactFormer-m) model which replaces the original chained factorized attention with parallel factorized attention. The IFactFormer-m model successfully performs long-term predictions for turbulent channel flow, whereas the original IFactFormer (IFactFormer-o), Fourier neural operator (FNO), and implicit Fourier neural operator (IFNO) exhibit a poor performance. Turbulent channel flows are simulated by direct numerical simulation using fine grids at friction Reynolds numbers $\text{Re}_{\tau}\approx 180,395,590$, and filtered to coarse grids for training neural operator. The neural operator takes the current flow field as input and predicts the flow field at the next time step, and long-term prediction is achieved in the posterior through an autoregressive approach. The results show that IFactFormer-m, compared to other neural operators and the traditional large eddy simulation (LES) methods including dynamic Smagorinsky model (DSM) and the wall-adapted local eddy-viscosity (WALE) model, reduces prediction errors in the short term, and achieves stable and accurate long-term prediction of various statistical properties and flow structures, including the energy spectrum, mean streamwise velocity, root mean square (rms) values of fluctuating velocities, Reynolds shear stress, and spatial structures of instantaneous velocity. Moreover, the trained IFactFormer-m is much faster than traditional LES methods. By analyzing the attention kernels, we elucidate the reasons why IFactFormer-m converges faster and achieves a stable and accurate long-term prediction compared to IFactFormer-o. Code and data are available at: https://github.com/huiyu-2002/IFactFormer-m.

arXiv:2412.15756

Probabilistic Latent Variable Modeling for Dynamic Friction Identification and Estimation

Precise identification of dynamic models in robotics is essential to support control design, friction compensation, output torque estimation, etc. A longstanding challenge remains in the identification of friction models for robotic joints, given the numerous physical phenomena affecting the underlying friction dynamics which result into nonlinear characteristics and hysteresis behaviour in particular. These phenomena proof difficult to be modelled and captured accurately using physical analogies alone. This has motivated researchers to shift from physics-based to data-driven models. Currently, these methods are still limited in their ability to generalize effectively to typical industrial robot deployement, characterized by high- and low-velocity operations and frequent direction reversals. Empirical observations motivate the use of dynamic friction models but these remain particulary challenging to establish. To address the current limitations, we propose to account for unidentified dynamics in the robot joints using latent dynamic states. The friction model may then utilize both the dynamic robot state and additional information encoded in the latent state to evaluate the friction torque. We cast this stochastic and partially unsupervised identification problem as a standard probabilistic representation learning problem. In this work both the friction model and latent state dynamics are parametrized as neural networks and integrated in the conventional lumped parameter dynamic robot model. The complete dynamics model is directly learned from the noisy encoder measurements in the robot joints. We use the Expectation-Maximisation (EM) algorithm to find a Maximum Likelihood Estimate (MLE) of the model parameters. The effectiveness of the proposed method is validated in terms of open-loop prediction accuracy in comparison with baseline methods, using the Kuka KR6 R700 as a test platform.

arXiv:2412.08286

Towards Precision in Bolted Joint Design: A Preliminary Machine Learning-Based Parameter Prediction

Bolted joints are critical in engineering for maintaining structural integrity and reliability. Accurate prediction of parameters influencing their function and behavior is essential for optimal performance. Traditional methods often fail to capture the non-linear behavior of bolted joints or require significant computational resources, limiting accuracy and efficiency. This study addresses these limitations by combining empirical data with a feed-forward neural network to predict load capacity and friction coefficients. Leveraging experimental data and systematic preprocessing, the model effectively captures nonlinear relationships, including rescaling output variables to address scale discrepancies, achieving 95.24% predictive accuracy. While limited dataset size and diversity restrict generalizability, the findings demonstrate the potential of neural networks as a reliable, efficient alternative for bolted joint design. Future work will focus on expanding datasets and exploring hybrid modeling techniques to enhance applicability.

arXiv:2412.06809

Generating Diverse Synthetic Datasets for Evaluation of Real-life Recommender Systems

Synthetic datasets are important for evaluating and testing machine learning models. When evaluating real-life recommender systems, high-dimensional categorical (and sparse) datasets are often considered. Unfortunately, there are not many solutions that would allow generation of artificial datasets with such characteristics. For that purpose, we developed a novel framework for generating synthetic datasets that are diverse and statistically coherent. Our framework allows for creation of datasets with controlled attributes, enabling iterative modifications to fit specific experimental needs, such as introducing complex feature interactions, feature cardinality, or specific distributions. We demonstrate the framework's utility through use cases such as benchmarking probabilistic counting algorithms, detecting algorithmic bias, and simulating AutoML searches. Unlike existing methods that either focus narrowly on specific dataset structures, or prioritize (private) data synthesis through real data, our approach provides a modular means to quickly generating completely synthetic datasets we can tailor to diverse experimental requirements. Our results show that the framework effectively isolates model behavior in unique situations and highlights its potential for significant advancements in the evaluation and development of recommender systems. The readily-available framework is available as a free open Python package to facilitate research with minimal friction.

arXiv:2411.15143

dafny-annotator: AI-Assisted Verification of Dafny Programs

Formal verification has the potential to drastically reduce software bugs, but its high additional cost has hindered large-scale adoption. While Dafny presents a promise to significantly reduce the effort to write verified programs, users are often required to provide logical annotations to aid the verifier. Here, we explore using a combination of Large Language Models and search to build dafny-annotator: a tool that adds logical annotations to a Dafny method until the verifier can prove it correct. On a test set from the DafnyBench collection of programs, greedy search guided by LLaMa 3.1 8B successfully annotates only 15.7% of the methods. Since this data-driven approach is hindered by the lack of large-scale training data, we propose a method for open-ended synthesis of new Dafny programs in a flexible pipeline where LLMs formulate high-level ideas, implement them, and incrementally propose changes to existing programs, which Dafny validates. This gives us a synthetic dataset, DafnySynth, which we use to augment DafnyBench for training. Fine-tuning on both datasets boosts LLaMa 8B's success rate to 50.6% -- significantly better than the base model, or training on either dataset alone. Our results suggest a path towards capable AI assistants for languages that don't yet have large-scale human-generated examples. In turn, such assistants might reduce friction for users and ultimately drive adoption.

arXiv:2411.12296

Modeling and Analysis of Terahertz Wave Propagation in Charged Dust Using Extended Mie Scattering Theory

Terahertz (THz) band (0.1-10 THz) possesses multi-gigahertz continuous bandwidth resources, making it a promising frequency band for high-speed wireless communications and environment sensing. The interaction between the THz wave and the external environment has been studied for various scenarios. However, it has recently been revealed that the friction forces in dust storms as well as the irradiation of sunlight and solar wind lead to the electrification of dust particles on Earth and the Moon. The THz wave propagation in these charged dust has not been fully investigated, which is essential for THz aerial communications in dust storms and lunar communications. In this paper, a channel model for THz wave propagation in charged dust is developed for wireless communications. Specifically, an extended Mie scattering model for charged dust is first introduced, which captures the electrodynamic feature of the interaction between THz wave and charged particles. Then, the diameter and density distributions of dust particles are modeled, based on which the propagation loss of THz wave in charged dust is modeled and elaborated. Finally, numerical results on the additional loss caused by these charged dust with different sizes in the THz band are evaluated and compared. Extensive results demonstrate that as the number of dust charges increases, the extinction cross section of smaller-sized particles significantly increases, and the overall attenuation led by charged dust increases by at most 50% at 0.3 THz.

arXiv:2411.08622공식 코드 · Papers with Code

Precision-Focused Reinforcement Learning Model for Robotic Object Pushing

Non-prehensile manipulation, such as pushing objects to a desired target position, is an important skill for robots to assist humans in everyday situations. However, the task is challenging due to the large variety of objects with different and sometimes unknown physical properties, such as shape, size, mass, and friction. This can lead to the object overshooting its target position, requiring fast corrective movements of the robot around the object, especially in cases where objects need to be precisely pushed. In this paper, we improve the state-of-the-art by introducing a new memory-based vision-proprioception RL model to push objects more precisely to target positions using fewer corrective movements.

arXiv:2411.08027

LLMPhy: Complex Physical Reasoning Using Large Language Models and World Models

Physical reasoning is an important skill needed for robotic agents when operating in the real world. However, solving such reasoning problems often involves hypothesizing and reflecting over complex multi-body interactions under the effect of a multitude of physical forces and thus learning all such interactions poses a significant hurdle for state-of-the-art machine learning frameworks, including large language models (LLMs). To study this problem, we propose a new physical reasoning task and a dataset, dubbed TraySim. Our task involves predicting the dynamics of several objects on a tray that is given an external impact -- the domino effect of the ensued object interactions and their dynamics thus offering a challenging yet controlled setup, with the goal of reasoning being to infer the stability of the objects after the impact. To solve this complex physical reasoning task, we present LLMPhy, a zero-shot black-box optimization framework that leverages the physics knowledge and program synthesis abilities of LLMs, and synergizes these abilities with the world models built into modern physics engines. Specifically, LLMPhy uses an LLM to generate code to iteratively estimate the physical hyperparameters of the system (friction, damping, layout, etc.) via an implicit analysis-by-synthesis approach using a (non-differentiable) simulator in the loop and uses the inferred parameters to imagine the dynamics of the scene towards solving the reasoning task. To show the effectiveness of LLMPhy, we present experiments on our TraySim dataset to predict the steady-state poses of the objects. Our results show that the combination of the LLM and the physics engine leads to state-of-the-art zero-shot physical reasoning performance, while demonstrating superior convergence against standard black-box optimization methods and better estimation of the physical parameters.

arXiv:2411.03736

Static structural behaviour of wire bearings: comparison with conventional bearings and study of design and operational parameters

In wire bearings the rolling process occurs on raceways machined on steel wires, and the rings are made of light materials such as aluminium. This particular architecture provides both weight and inertia savings, but also significantly different behaviour with respect to conventional bearings. For this reason, specific design and analysis tools must be developed; as a first step, this work uses Finite Element models to study the influence of different parameters on the static structural response of wire bearings. Thus, bearing stiffness, load capacity and contact status (contact force and angle, and ellipse truncation) have been evaluated for several combinations of conformity, friction coefficient and boundary conditions. The results have been compared with an equivalent conventional bearing, shedding light on the main structural features of wire bearings.

arXiv:2411.02905

Load distribution and friction torque in four-point contact slewing bearings considering manufacturing errors and ring flexibility

This work introduces a methodology for the calculation of the load distribution in four-point contact slewing bearings considering ball preload, manufacturing errors and ring flexibility. The model is built by the formulation and minimization of the potential energy of the bearing. Comparing with the rigid rings assumption, the results show that ring deformations involve lower interferences in idling conditions, and have a great effect in the load distribution, but not under external loads. Additionally, a new approach has been proposed for the calculation of the friction torque, which has lower computational cost in comparison to a previous approach by the authors, so more accurate results can be obtained due to refined calculations with no significant cost increase.