PyTorchKR Papers

최신 논문

총 1,050편
태그·기간 등 상세 필터

카테고리 탐색

로그인 후 이 태그 구독

arXiv:2108.03936

3D Human Reconstruction in the Wild with Collaborative Aerial Cameras

Aerial vehicles are revolutionizing applications that require capturing the 3D structure of dynamic targets in the wild, such as sports, medicine, and entertainment. The core challenges in developing a motion-capture system that operates in outdoors environments are: (1) 3D inference requires multiple simultaneous viewpoints of the target, (2) occlusion caused by obstacles is frequent when tracking moving targets, and (3) the camera and vehicle state estimation is noisy. We present a real-time aerial system for multi-camera control that can reconstruct human motions in natural environments without the use of special-purpose markers. We develop a multi-robot coordination scheme that maintains the optimal flight formation for target reconstruction quality amongst obstacles. We provide studies evaluating system performance in simulation, and validate real-world performance using two drones while a target performs activities such as jogging and playing soccer. Supplementary video: https://youtu.be/jxt91vx0cns

arXiv:2108.03637공식 코드 · Papers with Code

Saliency-Associated Object Tracking

Most existing trackers based on deep learning perform tracking in a holistic strategy, which aims to learn deep representations of the whole target for localizing the target. It is arduous for such methods to track targets with various appearance variations. To address this limitation, another type of methods adopts a part-based tracking strategy which divides the target into equal patches and tracks all these patches in parallel. The target state is inferred by summarizing the tracking results of these patches. A potential limitation of such trackers is that not all patches are equally informative for tracking. Some patches that are not discriminative may have adverse effects. In this paper, we propose to track the salient local parts of the target that are discriminative for tracking. In particular, we propose a fine-grained saliency mining module to capture the local saliencies. Further, we design a saliency-association modeling module to associate the captured saliencies together to learn effective correlation representations between the exemplar and the search image for state estimation. Extensive experiments on five diverse datasets demonstrate that the proposed method performs favorably against state-of-the-art trackers.

arXiv:2108.03172

A General Regularized Distributed Solution for System State Estimation from Relative Measurements

This work presents a novel general regularized distributed solution for the state estimation problem in networked systems. Resting on the graph-based representation of sensor networks and adopting a multivariate least-squares approach, the designed solution exploits the set of the available inter-sensor relative measurements and leverages a general regularization framework, whose parameter selection is shown to control the estimation procedure convergence performance. As confirmed by the numerical results, this new estimation scheme allows (i) the extension of other approaches investigated in the literature and (ii) the convergence optimization in correspondence to any (undirected) graph modeling the given sensor network.

arXiv:2101.03457

Deep Ensemble Learning-based Approach to Real-time Power System State Estimation

Power system state estimation (PSSE) is commonly formulated as weighted least-square (WLS) algorithm and solved using iterative methods such as Gauss-Newton methods. However, iterative methods have become more sensitive to system operating conditions than ever before due to the deployment of intermittent renewable energy sources, low carbon technologies (e.g., electric vehicles), and demand response programs. Appropriate PSSE approaches are required to avoid pitfalls of the WLS-based PSSE computations for accurate prediction of operating conditions. This paper proposes a data-driven real-time PSSE using a deep ensemble learning algorithm. In the proposed approach, the ensemble learning setup is formulated with dense residual neural networks as base-learners and multivariate-linear regressor as meta-learner. Historical measurements and states are utilised to train and test the model. The trained model can be used in real-time to estimate power system states (voltage magnitudes and phase angles) using real-time measurements. Most of current data-driven PSSE methods assume the availability of a complete set of measurements, which may not be the case in real power system data-acquisition. This paper adopts multivariate linear regression to forecast system states for instants of missing measurements to assist the proposed PSSE technique. Case studies are performed on various IEEE standard benchmark systems to validate the proposed approach. The results show that the proposed approach outperforms existing data-driven PSSE methods techniques.

arXiv:2108.01306

Distributed Dynamic State-Input Estimation for Power Networks of Microgrids and Active Distribution Systems with Unknown Inputs

This paper proposes a joint input and state dynamic estimation scheme for power networks in microgrids and active distribution systems with unknown inputs. The conventional dynamic state estimation of power networks in the transmission system relies on the forecasting methods to obtain the state-transition model of state variables. However, under highly dynamic conditions in the operation of microgrids and active distribution networks, this approach may become ineffective as the forecasting accuracy is not guaranteed. To overcome such drawbacks, this paper employs the power networks model derived from the physical equations of branch currents. Specifically, the power network model is a linear state-space model, in which the state vector consists of branch currents, and the input vector consists of bus voltages. To estimate both state and input variables, we propose linear Kalman-based dynamic filtering algorithms in batch-mode regression form, considering the cross-correlation between states and inputs. For the scalability of the proposed scheme, the distributed implementation is also presented. Complementarily, the predicted state and input vectors are leveraged for bad data detection. Results carried out on a 13-bus microgrid system in real-time Opal-RT platform demonstrate the effectiveness of the proposed method in comparison with the traditional weighted least square and tracking state estimation methods.

arXiv:2108.00366공식 코드 · Papers with Code

Agent-aware State Estimation in Autonomous Vehicles

Autonomous systems often operate in environments where the behavior of multiple agents is coordinated by a shared global state. Reliable estimation of the global state is thus critical for successfully operating in a multi-agent setting. We introduce agent-aware state estimation -- a framework for calculating indirect estimations of state given observations of the behavior of other agents in the environment. We also introduce transition-independent agent-aware state estimation -- a tractable class of agent-aware state estimation -- and show that it allows the speed of inference to scale linearly with the number of agents in the environment. As an example, we model traffic light classification in instances of complete loss of direct observation. By taking into account observations of vehicular behavior from multiple directions of traffic, our approach exhibits accuracy higher than that of existing traffic light-only HMM methods on a real-world autonomous vehicle data set under a variety of simulated occlusion scenarios.

논문 ID: document/6155073

Unscented Kalman Filter: Aspects and Adaptive Setting of Scaling Parameter

This technical note deals with the unscented Kalman filter for state estimation of nonlinear stochastic dynamic systems with a special focus on the scaling parameter of the filter. Its standard choice is analyzed and its impact on the estimation quality is discussed. On the basis of the analysis, a novel method for adaptive setting of the parameter in the unscented Kalman filter is proposed. The results are illustrated in a numerical example.

arXiv:2107.12674

Vision-Guided Forecasting -- Visual Context for Multi-Horizon Time Series Forecasting

Autonomous driving gained huge traction in recent years, due to its potential to change the way we commute. Much effort has been put into trying to estimate the state of a vehicle. Meanwhile, learning to forecast the state of a vehicle ahead introduces new capabilities, such as predicting dangerous situations. Moreover, forecasting brings new supervision opportunities by learning to predict richer a context, expressed by multiple horizons. Intuitively, a video stream originated from a front-facing camera is necessary because it encodes information about the upcoming road. Besides, historical traces of the vehicle's states give more context. In this paper, we tackle multi-horizon forecasting of vehicle states by fusing the two modalities. We design and experiment with 3 end-to-end architectures that exploit 3D convolutions for visual features extraction and 1D convolutions for features extraction from speed and steering angle traces. To demonstrate the effectiveness of our method, we perform extensive experiments on two publicly available real-world datasets, Comma2k19 and the Udacity challenge. We show that we are able to forecast a vehicle's state to various horizons, while outperforming the current state-of-the-art results on the related task of driving state estimation. We examine the contribution of vision features, and find that a model fed with vision features achieves an error that is 56.6% and 66.9% of the error of a model that doesn't use those features, on the Udacity and Comma2k19 datasets respectively.

arXiv:2107.10043공식 코드 · Papers with Code

KalmanNet: Neural Network Aided Kalman Filtering for Partially Known Dynamics

State estimation of dynamical systems in real-time is a fundamental task in signal processing. For systems that are well-represented by a fully known linear Gaussian state space (SS) model, the celebrated Kalman filter (KF) is a low complexity optimal solution. However, both linearity of the underlying SS model and accurate knowledge of it are often not encountered in practice. Here, we present KalmanNet, a real-time state estimator that learns from data to carry out Kalman filtering under non-linear dynamics with partial information. By incorporating the structural SS model with a dedicated recurrent neural network module in the flow of the KF, we retain data efficiency and interpretability of the classic algorithm while implicitly learning complex dynamics from data. We demonstrate numerically that KalmanNet overcomes non-linearities and model mismatch, outperforming classic filtering methods operating with both mismatched and accurate domain knowledge.

arXiv:2107.07707공식 코드 · Papers with Code

Probabilistic Appearance-Invariant Topometric Localization with New Place Awareness

Probabilistic state-estimation approaches offer a principled foundation for designing localization systems, because they naturally integrate sequences of imperfect motion and exteroceptive sensor data. Recently, probabilistic localization systems utilizing appearance-invariant visual place recognition (VPR) methods as the primary exteroceptive sensor have demonstrated state-of-the-art performance in the presence of substantial appearance change. However, existing systems 1) do not fully utilize odometry data within the motion models, and 2) are unable to handle route deviations, due to the assumption that query traverses exactly repeat the mapping traverse. To address these shortcomings, we present a new probabilistic topometric localization system which incorporates full 3-dof odometry into the motion model and furthermore, adds an "off-map" state within the state-estimation framework, allowing query traverses which feature significant route detours from the reference map to be successfully localized. We perform extensive evaluation on multiple query traverses from the Oxford RobotCar dataset exhibiting both significant appearance change and deviations from routes previously traversed. In particular, we evaluate performance on two practically relevant localization tasks: loop closure detection and global localization. Our approach achieves major performance improvements over both existing and improved state-of-the-art systems.

arXiv:2107.05891

Dynamic State Estimation for Integrated Natural Gas and Electric Power Systems

A dynamic state estimation method of integrated natural gas and electric power systems (IGESs) in proposed. Firstly, the coupling model of gas pipeline networks and power systems by gas turbine units (GTUs) is established. Secondly, the Kalman filter based linear DSE model for the IGES is built. The gas density and mass flow rate, as well as the real and imaginary parts of bus voltages are taken as states, which are predicted by the linearized fluid dynamic equations of gases and exponential smoothing techniques. Boundary conditions of pipeline networks are used as supplementary constraints in the system model. At last, the proposed method is applied to an IGES including a 30-node pipeline network and IEEE 39-bus system coupled by two GTUs. Two indexes are used to evaluate the DSE performance under three measurement error conditions, and the results show that the DSE can obtain the accurate dynamic states in different conditions.

arXiv:2106.13461

Detectability Conditions and State Estimation for Linear Time-Varying and Nonlinear Systems

This work proposes a detectability condition for linear time-varying systems based on the exponential dichotomy spectrum. The condition guarantees the existence of an observer, whose gain is determined only by the unstable modes of the system. This allows for an observer design with low computational complexity compared to classical estimation approaches. An extension of this observer design to a class of nonlinear systems is proposed and local convergence of the corresponding estimation error dynamics is proven. Numerical results show the efficacy of the proposed observer design technique.

arXiv:2106.13084

Real Time State Estimation of Power Grids Using Convolutional Neural Networks and State Forecasting Via Recurrent Neural Networks

Power grids play a very important role in delivering electrical energy to homes, industries and other places that require it. Because of this increased demand they are facing a great challenge of voltage variations. This happens due to varied use of energy-consuming devices and appliances like electric vehicles, industrial consumption, occasional peak in energy demands etc. For these fluctuations in demands, it becomes extremely important to monitor the conditions at which the power grid operates. Once these conditions are known, the energy production can be manipulated to meet the demand. It has been found that the existing Power System State Estimation (PSSE) techniques may not be good in producing optimal Performance. Moreover, they are also expensive in terms of computational processing. To address this problem, this research proposes a state estimation method for power grids using Convolutional Neural Networks (CNN). It was found that the model produced an RMSE of 2.57 x 10-4, which was comparatively accurate than one of previous studies involved in making the estimation using a Prox Linear Model (2.97 x 10-4). Furthermore, the research also proposes Power System State Forecasting for improving system awareness and resilience. The forecasting is carried out using a model of Recurrent Neural Network (RNN). This model helps in accounting for long-term nonlinear aspects present in data and based on that it does the forecasting. The proposed model forecasted with a RMSE of 2.53 x 10-3, which is comparatively equal to the previous study mentioned above (2.59 x 10-3).

arXiv:2106.09826

Intrusion Detection and Localization for Networked Embedded Control Systems

Closed-loop control systems employ continuous sensing and actuation to maintain controlled variables within preset bounds and achieve the desired system output. Intentional disturbances in the system, such as in the case of cyberattacks, can compromise reachability of control goals, and in several cases jeopardize safety. The increasing connectivity and exposure of networked control to external networks has enabled attackers to compromise these systems by exploiting security vulnerabilities. Attacks against safety-critical control loops can not only drive the system over a trajectory different from the desired, but also cause fatal consequences to humans. In this paper we present a physics-based Intrusion Detection System (IDS) aimed at increasing the security in control systems. In addition to conventional process state estimation for intrusion detection, since the controller cannot be trusted, we introduce a controller state estimator. Additionally, we make our detector context-aware by utilizing sensor measurements from other control loops, which allows to distinguish and characterize disturbances from attacks. We introduce adaptive thresholding and adaptive filtering as means to achieve context-awareness. Together, these methodologies allow detection and localization of attacks in closed-loop controls. Finally, we demonstrate feasibility of the approach by mounting a series of attacks against a networked Direct Current (DC) motor closed-loop speed control deployed on an ECU testbed, as well as on a simulated automated lane keeping system. Among other application domains, this set of approaches is key to support security in automotive systems, and ultimately increase road and passenger safety.

arXiv:2106.04219

Time-series Imputation of Temporally-occluded Multiagent Trajectories

In multiagent environments, several decision-making individuals interact while adhering to the dynamics constraints imposed by the environment. These interactions, combined with the potential stochasticity of the agents' decision-making processes, make such systems complex and interesting to study from a dynamical perspective. Significant research has been conducted on learning models for forward-direction estimation of agent behaviors, for example, pedestrian predictions used for collision-avoidance in self-driving cars. However, in many settings, only sporadic observations of agents may be available in a given trajectory sequence. For instance, in football, subsets of players may come in and out of view of broadcast video footage, while unobserved players continue to interact off-screen. In this paper, we study the problem of multiagent time-series imputation, where available past and future observations of subsets of agents are used to estimate missing observations for other agents. Our approach, called the Graph Imputer, uses forward- and backward-information in combination with graph networks and variational autoencoders to enable learning of a distribution of imputed trajectories. We evaluate our approach on a dataset of football matches, using a projective camera module to train and evaluate our model for the off-screen player state estimation setting. We illustrate that our method outperforms several state-of-the-art approaches, including those hand-crafted for football.

arXiv:2106.03066

Efficient Secure State Estimation against Sparse Integrity Attack for System with Non-derogatory Dynamics

We consider the problem of estimating the state of a time-invariant linear Gaussian system in the presence of integrity attacks. The attacker can compromise $p$ out of $m$ sensors, the set of which is fixed over time and unknown to the system operator, and manipulate the measurements arbitrarily. Under the assumption that all the unstable eigenvalues of system matrix $A$ have geometric multiplicity 1 (unstable part of $A$ is non-derogatory), we propose a secure estimation scheme that is resilient to integrity attack as long as the system is $2p$-sparse detectable, which is proved to be the fundamental limit of secure dynamic estimation. In the absence of attack, the proposed estimation coincides with Kalman estimation with a certain probability that can be adjusted to trade-off between performance with and without attack. Furthermore, the detectability condition checking in the designing phase and the estimation computing in the online operating phase are both computationally efficient. A numerical example is provided to corroborate the results and illustrate the performance of the proposed estimator.

arXiv:2106.03287공식 코드 · Papers with Code

Stein ICP for Uncertainty Estimation in Point Cloud Matching

Quantification of uncertainty in point cloud matching is critical in many tasks such as pose estimation, sensor fusion, and grasping. Iterative closest point (ICP) is a commonly used pose estimation algorithm which provides a point estimate of the transformation between two point clouds. There are many sources of uncertainty in this process that may arise due to sensor noise, ambiguous environment, and occlusion. However, for safety critical problems such as autonomous driving, a point estimate of the pose transformation is not sufficient as it does not provide information about the multiple solutions. Current probabilistic ICP methods usually do not capture all sources of uncertainty and may provide unreliable transformation estimates which can have a detrimental effect in state estimation or decision making tasks that use this information. In this work we propose a new algorithm to align two point clouds that can precisely estimate the uncertainty of ICP's transformation parameters. We develop a Stein variational inference framework with gradient based optimization of ICP's cost function. The method provides a non-parametric estimate of the transformation, can model complex multi-modal distributions, and can be effectively parallelized on a GPU. Experiments using 3D kinect data as well as sparse indoor/outdoor LiDAR data show that our method is capable of efficiently producing accurate pose uncertainty estimates.

arXiv:2106.03142

A Physics-Informed Deep Learning Paradigm for Traffic State and Fundamental Diagram Estimation

Traffic state estimation (TSE) bifurcates into two categories, model-driven and data-driven (e.g., machine learning, ML), while each suffers from either deficient physics or small data. To mitigate these limitations, recent studies introduced a hybrid paradigm, physics-informed deep learning (PIDL), which contains both model-driven and data-driven components. This paper contributes an improved version, called physics-informed deep learning with a fundamental diagram learner (PIDL+FDL), which integrates ML terms into the model-driven component to learn a functional form of a fundamental diagram (FD), i.e., a mapping from traffic density to flow or velocity. The proposed PIDL+FDL has the advantages of performing the TSE learning, model parameter identification, and FD estimation simultaneously. We demonstrate the use of PIDL+FDL to solve popular first-order and second-order traffic flow models and reconstruct the FD relation as well as model parameters that are outside the FD terms. We then evaluate the PIDL+FDL-based TSE using the Next Generation SIMulation (NGSIM) dataset. The experimental results show the superiority of the PIDL+FDL in terms of improved estimation accuracy and data efficiency over advanced baseline TSE methods, and additionally, the capacity to properly learn the unknown underlying FD relation.

arXiv:2106.02254

State Estimation in Unobservable Power Systems via Graph Signal Processing Tools

We consider the problem of estimating the states in an unobservable power system. To this end, we propose novel graph signal processing (GSP) methods. For simplicity, we start with analyzing the DC power flow (DC-PF) model and then extend our algorithms to the AC power flow (AC-PF) model. The main assumption behind the proposed GSP approach is that the grid states, which include the vector of phases and the vector of the magnitudes of the voltages in the system, is a smooth graph signal with respect to the system admittance matrix that represents the underlying graph. Thus, the first step in this paper is to validate the graph-smoothness assumption of the states, both empirically and theoretically. Then, we develop the regularized GSP weighted least squares (GSP-WLS) state estimator, which does not require observability of the network. We propose a sensor placement strategy that aims to optimize the estimation performance of the GSP-WLS estimator. Finally, we extend the GSP-WLS estimator method to the AC-PF model by integrating a smoothness regularization term into the Gauss-Newton algorithm. Numerical results on the IEEE 118-bus system demonstrate that the new GSP methods outperform commonly-used estimation approaches and are robust to missing data.

arXiv:2106.01993

Real-time Grid and DER Co-simulation Platform for Validating Large-scale DER Control Schemes

Distributed energy resources (DERs) such as responsive loads and energy storage systems are valuable resources available to grid operators for balancing supply-demand mismatches via load coordination. However, consumer acceptance of load coordination schemes depends on ensuring quality of service (QoS), which embodies device-level constraints. Since each device has its own internal energy state, the effect of QoS on the fleet can be cast as fleet-wide energy limits within which the aggregate "state of charge" (SoC) must be actively maintained. This requires coordination of DERs that is cognizant of the SoC, responsive to grid conditions, and depends on fast communication networks. To that effect, this paper presents a novel real-time grid-and-DER co-simulation platform for validating advanced DER coordination schemes and characterizing the capability of such a DER fleet. In particular, we present how the co-simulation platform is suitable for: i) testing real-time performance of a large fleet of DERs in delivering advanced grid services, including frequency regulation; ii) online state estimation to characterize the corresponding SoC of a large fleet of DERs; and iii) incorporating practical limitations of DERs and communications and analyzing the effects on fleet-wide performance. To illustrate these benefits of the presented grid-DER co-simulation platform, we employ the advanced DER coordination scheme called packetized energy management (PEM), which is a novel device-driven, asynchronous, and randomizing control paradigm for DERs. A fleet of thousands of PEM-enabled DERs are then added to a realistic and dynamical model of the Vermont transmission system to complete validation of the co-simulation platform.