PyTorchKR Papers

최신 논문

총 403편
태그·기간 등 상세 필터

카테고리 탐색

로그인 후 이 태그 구독

arXiv:2407.13257

Predictive control for nonlinear stochastic systems: Closed-loop guarantees with unbounded noise

We present a stochastic model predictive control framework for nonlinear systems subject to unbounded process noise with closed-loop guarantees. First, we provide a conceptual shrinking-horizon framework that utilizes general probabilistic reachable sets and minimizes the expected cost. Then, we provide a tractable receding-horizon formulation that uses a nominal state to minimize a deterministic quadratic cost and satisfy tightened constraints. Our theoretical analysis demonstrates recursive feasibility, satisfaction of chance constraints, and bounds on the expected cost for the resulting closed-loop system. We provide a constructive design for probabilistic reachable sets of nonlinear continuously differentiable systems using stochastic contraction metrics and an assumed bound on the covariance matrices. Numerical simulations highlight the computational efficiency and theoretical guarantees of the proposed method. Overall, this paper provides a framework for computationally tractable stochastic predictive control with closed-loop guarantees for nonlinear systems with unbounded noise.

arXiv:2407.12989

Trajectory Planning Using Tire Thermodynamics for Automated Drifting

Automated vehicles need to estimate tire-road friction information, as it plays a key role in safe trajectory planning and vehicle dynamics control. Notably, friction is not solely dependent on road surface conditions, but also varies significantly depending on the tire temperature. However, tire parameters such as the friction coefficient have been conventionally treated as constant values in automated vehicle motion planning. This paper develops a simple thermodynamic model that captures tire friction temperature variation. To verify the model, it is implemented into trajectory planning for automated drifting - a challenging application that requires leveraging an unstable, drifting equilibrium at the friction limits. The proposed method which captures the hidden tire dynamics provides a dynamically feasible trajectory, leading to more precise tracking during experiments with an LQR (Linear Quadratic Regulator) controller.

arXiv:2407.11805

Friction and Road Condition Estimation by Combining Cause- and Effect-Based Methods using Bayesian Networks

Knowledge about the maximum tire-road friction potential is an important factor to ensure the driving stability and traffic safety of the vehicle. Many authors proposed systems that either measure friction related parameters or estimate the friction coefficient directly via a mathematical model. However these systems can be negatively impacted by environmental factors or require a sufficient level of excitation in the form of tire slip, which is often too low under practical conditions. Therefore, this work investigates, if a more robust estimation can be achieved by fusing the information of multiple systems using a Bayesian network, which models the statistical relationship between the sensors and the maximum friction coefficient. First, the Bayesian network is evaluated over its entire domain to compare the inference process to all possible road conditions. After that, the algorithm is applied to data from a test vehicle to demonstrate the performance under real conditions.

arXiv:2407.12804

Modulating Language Model Experiences through Frictions

Language models are transforming the ways that their users engage with the world. Despite impressive capabilities, over-consumption of language model outputs risks propagating unchecked errors in the short-term and damaging human capabilities for critical thinking in the long-term. How can we develop scaffolding around language models to curate more appropriate use? We propose selective frictions for language model experiences, inspired by behavioral science interventions, to dampen misuse. Frictions involve small modifications to a user's experience, e.g., the addition of a button impeding model access and reminding a user of their expertise relative to the model. Through a user study with real humans, we observe shifts in user behavior from the imposition of a friction over LLMs in the context of a multi-topic question-answering task as a representative task that people may use LLMs for, e.g., in education and information retrieval. We find that frictions modulate over-reliance by driving down users' click rates while minimally affecting accuracy for those topics. Yet, frictions may have unintended effects. We find marked differences in users' click behaviors even on topics where frictions were not provisioned. Our contributions motivate further study of human-AI behavioral interaction to inform more effective and appropriate LLM use.

arXiv:2407.13760

Neural Network Tire Force Modeling for Automated Drifting

Automated drifting presents a challenge problem for vehicle control, requiring models and control algorithms that can precisely handle nonlinear, coupled tire forces at the friction limits. We present a neural network architecture for predicting front tire lateral force as a drop-in replacement for physics-based approaches. With a full-scale automated vehicle purpose-built for the drifting application, we deploy these models in a nonlinear model predictive controller tuned for tracking a reference drifting trajectory, for direct comparisons of model performance. The neural network tire model exhibits significantly improved path tracking performance over the brush tire model in cases where front-axle braking force is applied, suggesting the neural network's ability to express previously unmodeled, latent dynamics in the drifting condition.

arXiv:2407.06920

Src Kinase Slows Collective Rotation of Confined Epithelial Cell Monolayers

Collective cell migration is key during development, wound healing and metastasis and relies on coordinated cell behaviors at the group level. Src kinase is a key signalling protein for physiological functions of epithelia, as it regulates many cellular processes, including adhesion, motility, and mechanotransduction. Its over-activation is associated to cancer aggressiveness. Here, we take advantage of optogenetics to precisely control Src activation in time and show that its pathological-like activation slows collective rotation of epithelial cells confined into circular adhesive patches. We interpret velocity, force and stress data during period of non-activation and period of activation of Src thanks to an hydrodynamic description of the cell assembly as a polar active fluid. Src activation leads to a 2-fold decrease in the ratio of polar angle to friction, which could result from increased adhesiveness at the cell-substrate interface. Measuring internal stress allows us to show that active stresses are subdominant compared to traction forces. Our work reveals the importance of fine-tuning the level of Src activity for coordinated collective behaviors.

arXiv:2407.06313

Data-Driven Multi-Modal Learning Model Predictive Control

We present a Learning Model Predictive Controller (LMPC) for multi-modal systems performing iterative control tasks. Assuming availability of historical data, our goal is to design a data-driven control policy for the multi-modal system where the current mode is unknown. First, we propose a novel method to select local data for constructing affine time-varying (ATV) models of a multi-modal system in the context of LMPC. Then we present how to build a sampled safe set from multi-modal historical data. We demonstrate the effectiveness of our method through simulation results of automated driving on a friction-varying track.

arXiv:2407.04869

A Defeasible Deontic Calculus for Resolving Norm Conflicts

When deciding how to act, we must consider other agents' norms and values. However, our norms are ever-evolving. We often add exceptions or change our minds, and thus norms can conflict over time. Therefore, to maintain an accurate mental model of other's norms, and thus to avoid social friction, such conflicts must be detected and resolved quickly. Formalizing this process has been the focus of various deontic logics and normative multi-agent systems. We aim to bridge the gap between these two fields here. We contribute a defeasible deontic calculus with inheritance and prove that it resolves norm conflicts. Through this analysis, we also reveal a common resolution strategy as a red herring. This paper thus contributes a theoretically justified axiomatization of norm conflict detection and resolution.

arXiv:2406.16093

Towards Natural Language-Driven Assembly Using Foundation Models

Large Language Models (LLMs) and strong vision models have enabled rapid research and development in the field of Vision-Language-Action models that enable robotic control. The main objective of these methods is to develop a generalist policy that can control robots with various embodiments. However, in industrial robotic applications such as automated assembly and disassembly, some tasks, such as insertion, demand greater accuracy and involve intricate factors like contact engagement, friction handling, and refined motor skills. Implementing these skills using a generalist policy is challenging because these policies might integrate further sensory data, including force or torque measurements, for enhanced precision. In our method, we present a global control policy based on LLMs that can transfer the control policy to a finite set of skills that are specifically trained to perform high-precision tasks through dynamic context switching. The integration of LLMs into this framework underscores their significance in not only interpreting and processing language inputs but also in enriching the control mechanisms for diverse and intricate robotic operations.

arXiv:2406.05124

Optimizing Exit Queues for Proof-of-Stake Blockchains: A Mechanism Design Approach

Byzantine fault-tolerant consensus protocols have provable safety and liveness properties for static validator sets. In practice, however, the validator set changes over time, potentially eroding the protocol's security guarantees. For example, systems with accountable safety may lose some of that accountability over time as adversarial validators exit. As a result, protocols must rate limit entry and exit so that the set changes slowly enough to ensure security. Here, the system designer faces a fundamental trade-off. Slower exits increase friction, making it less attractive to stake in the first place. Faster exits provide more utility to stakers but weaken the protocol's security. This paper provides the first systematic study of exit queues for Proof-of-Stake blockchains. Given a collection of validator-set consistency constraints imposed by the protocol, the social planner's goal is to provide a constrained-optimal mechanism that minimizes disutility for the participants. We introduce the MINSLACK mechanism, a dynamic capacity first-come-first-served queue in which the amount of stake that can exit in a period depends on the number of previous exits and the consistency constraints. We show that MINSLACK is optimal when stakers equally value the processing of their withdrawal. When stakers values are heterogeneous, the optimal mechanism resembles a priority queue with dynamic capacity. However, this mechanism must reserve exit capacity for the future in case a staker with a much higher need for liquidity arrives. We conclude with a survey of known consistency constraints and highlight the diversity of existing exit mechanisms.

arXiv:2406.02211

Novel pre-emptive control solutions for V2X connected electric vehicles

V2X technologies will become widespread in the next generation of passenger cars, and enable the development of novel vehicle control functionalities. Although a wide literature describes the energy efficiency benefits of V2X connectivity, e.g., in terms of vehicle speed profiling and platooning, there is a gap in the analysis of the potential of vehicle connectivity in enhancing the performance of active safety control systems. To highlight the impact vehicle connectivity could have on future active safety systems, this paper presents two novel control functions for connected vehicles, benefitting from the precise knowledge of the expected path and tire-road friction conditions ahead, as well as the current position of the ego vehicle. These functions, developed within recent and ongoing European projects, are: i) pre-emptive traction control; and ii) pre-emptive braking control.

arXiv:2406.02206

Nonlinear Model Predictive Control for Preview-Based Traction Control

This study presents a nonlinear model predictive control (NMPC) formulation for preview-based traction control, which uses the information on the expected tire-road friction coefficient ahead to enhance the wheel slip control performance, in the context of connected vehicles with V2X features. Proof-of-concept experiments on an electric vehicle prototype highlight the real-time capability of the controller, and the wheel slip control performance improvement brought by the tire-road friction coefficient preview. Finally, an experimentally validated simulation model is used in sensitivity analyses, to evaluate the performance benefit of the preview-based controller for different dynamic characteristics (e.g., time constant and pure time delays) of the electric powertrains.

arXiv:2406.03245

Reconfiguring Participatory Design to Resist AI Realism

The growing trend of artificial intelligence (AI) as a solution to social and technical problems reinforces AI Realism -- the belief that AI is an inevitable and natural order. In response, this paper argues that participatory design (PD), with its focus on democratic values and processes, can play a role in questioning and resisting AI Realism. I examine three concerning aspects of AI Realism: the facade of democratization that lacks true empowerment, demands for human adaptability in contrast to AI systems' inflexibility, and the obfuscation of essential human labor enabling the AI system. I propose resisting AI Realism by reconfiguring PD to continue engaging with value-centered visions, increasing its exploration of non-AI alternatives, and making the essential human labor underpinning AI systems visible. I position PD as a means to generate friction against AI Realism and open space for alternative futures centered on human needs and values.

arXiv:2405.17199

Physically Consistent Modeling & Identification of Nonlinear Friction with Dissipative Gaussian Processes

Friction modeling has always been a challenging problem due to the complexity of real physical systems. Although a few state-of-the-art structured data-driven methods show their efficiency in nonlinear system modeling, deterministic passivity as one of the significant characteristics of friction is rarely considered in these methods. To address this issue, we propose a Gaussian Process based model that preserves the inherent structural properties such as passivity. A matrix-vector physical structure is considered in our approaches to ensure physical consistency, in particular, enabling a guarantee of positive semi-definiteness of the damping matrix. An aircraft benchmark simulation is employed to demonstrate the efficacy of our methodology. Estimation accuracy and data efficiency are increased substantially by considering and enforcing more structured physical knowledge. Also, the fulfillment of the dissipative nature of the aerodynamics is validated numerically.

arXiv:2405.18537

Augmented Conversation with Embedded Speech-Driven On-the-Fly Referencing in AR

This paper introduces the concept of augmented conversation, which aims to support co-located in-person conversations via embedded speech-driven on-the-fly referencing in augmented reality (AR). Today computing technologies like smartphones allow quick access to a variety of references during the conversation. However, these tools often create distractions, reducing eye contact and forcing users to focus their attention on phone screens and manually enter keywords to access relevant information. In contrast, AR-based on-the-fly referencing provides relevant visual references in real-time, based on keywords extracted automatically from the spoken conversation. By embedding these visual references in AR around the conversation partner, augmented conversation reduces distraction and friction, allowing users to maintain eye contact and supporting more natural social interactions. To demonstrate this concept, we developed \system, a Hololens-based interface that leverages real-time speech recognition, natural language processing and gaze-based interactions for on-the-fly embedded visual referencing. In this paper, we explore the design space of visual referencing for conversations, and describe our our implementation -- building on seven design guidelines identified through a user-centered design process. An initial user study confirms that our system decreases distraction and friction in conversations compared to smartphone searches, while providing highly useful and relevant information.

arXiv:2405.18515

Atlas3D: Physically Constrained Self-Supporting Text-to-3D for Simulation and Fabrication

Existing diffusion-based text-to-3D generation methods primarily focus on producing visually realistic shapes and appearances, often neglecting the physical constraints necessary for downstream tasks. Generated models frequently fail to maintain balance when placed in physics-based simulations or 3D printed. This balance is crucial for satisfying user design intentions in interactive gaming, embodied AI, and robotics, where stable models are needed for reliable interaction. Additionally, stable models ensure that 3D-printed objects, such as figurines for home decoration, can stand on their own without requiring additional supports. To fill this gap, we introduce Atlas3D, an automatic and easy-to-implement method that enhances existing Score Distillation Sampling (SDS)-based text-to-3D tools. Atlas3D ensures the generation of self-supporting 3D models that adhere to physical laws of stability under gravity, contact, and friction. Our approach combines a novel differentiable simulation-based loss function with physically inspired regularization, serving as either a refinement or a post-processing module for existing frameworks. We verify Atlas3D's efficacy through extensive generation tasks and validate the resulting 3D models in both simulated and real-world environments.

arXiv:2405.11669

Do No Harm: A Counterfactual Approach to Safe Reinforcement Learning

Reinforcement Learning (RL) for control has become increasingly popular due to its ability to learn rich feedback policies that take into account uncertainty and complex representations of the environment. When considering safety constraints, constrained optimization approaches, where agents are penalized for constraint violations, are commonly used. In such methods, if agents are initialized in, or must visit, states where constraint violation might be inevitable, it is unclear how much they should be penalized. We address this challenge by formulating a constraint on the counterfactual harm of the learned policy compared to a default, safe policy. In a philosophical sense this formulation only penalizes the learner for constraint violations that it caused; in a practical sense it maintains feasibility of the optimal control problem. We present simulation studies on a rover with uncertain road friction and a tractor-trailer parking environment that demonstrate our constraint formulation enables agents to learn safer policies than contemporary constrained RL methods.

arXiv:2405.03901

OmniActions: Predicting Digital Actions in Response to Real-World Multimodal Sensory Inputs with LLMs

The progression to "Pervasive Augmented Reality" envisions easy access to multimodal information continuously. However, in many everyday scenarios, users are occupied physically, cognitively or socially. This may increase the friction to act upon the multimodal information that users encounter in the world. To reduce such friction, future interactive interfaces should intelligently provide quick access to digital actions based on users' context. To explore the range of possible digital actions, we conducted a diary study that required participants to capture and share the media that they intended to perform actions on (e.g., images or audio), along with their desired actions and other contextual information. Using this data, we generated a holistic design space of digital follow-up actions that could be performed in response to different types of multimodal sensory inputs. We then designed OmniActions, a pipeline powered by large language models (LLMs) that processes multimodal sensory inputs and predicts follow-up actions on the target information grounded in the derived design space. Using the empirical data collected in the diary study, we performed quantitative evaluations on three variations of LLM techniques (intent classification, in-context learning and finetuning) and identified the most effective technique for our task. Additionally, as an instantiation of the pipeline, we developed an interactive prototype and reported preliminary user feedback about how people perceive and react to the action predictions and its errors.

arXiv:2405.01738

Question Suggestion for Conversational Shopping Assistants Using Product Metadata

Digital assistants have become ubiquitous in e-commerce applications, following the recent advancements in Information Retrieval (IR), Natural Language Processing (NLP) and Generative Artificial Intelligence (AI). However, customers are often unsure or unaware of how to effectively converse with these assistants to meet their shopping needs. In this work, we emphasize the importance of providing customers a fast, easy to use, and natural way to interact with conversational shopping assistants. We propose a framework that employs Large Language Models (LLMs) to automatically generate contextual, useful, answerable, fluent and diverse questions about products, via in-context learning and supervised fine-tuning. Recommending these questions to customers as helpful suggestions or hints to both start and continue a conversation can result in a smoother and faster shopping experience with reduced conversation overhead and friction. We perform extensive offline evaluations, and discuss in detail about potential customer impact, and the type, length and latency of our generated product questions if incorporated into a real-world shopping assistant.

arXiv:2405.00567

Remote Sensing Data Assimilation with a Chained Hydrologic-hydraulic Model for Flood Forecasting

A chained hydrologic-hydraulic model is implemented using predicted runoff from a large-scale hydrologic model (namely ISBA-CTRIP) as inputs to local hydrodynamic models (TELEMAC-2D) to issue forecasts of water level and flood extent. The uncertainties in the hydrological forcing and in friction parameters are reduced by an Ensemble Kalman Filter that jointly assimilates in-situ water levels and flood extent maps derived from remote sensing observations. The data assimilation framework is cycled in a real-time forecasting configuration. A cycle consists of a reanalysis and a forecast phase. Over the analysis, observations up to the present are assimilated. An ensemble is then initialized from the last analyzed states and issued forecasts for next 36 hr. Three strategies of forcing data for this forecast are investigated: (i) using CTRIP runoff for reanalysis and forecast, (ii) using observed discharge for analysis, then CTRIP runoff for forecast and (iii) using observed discharge for reanalysis and keep a persistent discharge value for forecast. It was shown that the data assimilation strategy provides a reliable reanalysis in hindcast mode. The combination of observed discharge and CTRIP runoff provides the most accurate results. For all strategies, the quality of the forecast decreases as the lead time increases. When the errors in CTRIP forcing are non-stationary, the forecast capability may be reduced. This work demonstrates that the forcing provided by a hydrologic model, while imperfect, can be efficiently used as input to a hydraulic model to issue reanalysis and forecasts, thanks to the assimilation of in-situ and remote sensing observations.