Adopting reinforcement learning (RL) for traffic signal control (TSC) is increasingly popular, and RL has become a promising solution for traffic signal control. However, several challenges still need to be overcome. Firstly, most RL methods use fixed action duration and select the green phase for the next state, which makes the phase duration less dynamic and flexible. Secondly, the phase sequence of RL methods can be arbitrary, affecting the real-world deployment which may require a cyclical phase structure. Lastly, the average travel time and throughput are not fair metrics to evaluate TSC performance. To address these challenges, we propose a multi-level traffic signal control framework, DynLight, which uses an optimization method Max-QueueLength (M-QL) to determine the phase and uses a deep Q-network to determine the duration of the corresponding phase. Based on DynLight, we further propose DynLight-C which adopts a well-trained deep Q-network of DynLight and replace M-QL with a cyclical control policy that actuates a set of phases in fixed cyclical order to realize cyclical phase structure. Comprehensive experiments on multiple real-world datasets demonstrate that DynLight achieves a new state-of-the-art. Furthermore, the deep Q-network of DynLight can learn well on determining the phase duration and DynLight-C demonstrates high performance for deployment.
相關內容
A fruitful collaboration is based on the mutual knowledge of each other skills and on the possibility of communicating their own limits and proposing alternatives to adapt the execution of a task to the capabilities of the collaborators. This paper aims at reproducing such a scenario in a human-robot collaboration setting by proposing a novel communication control architecture. Exploiting control barrier functions, the robot is made aware of its (dynamic) skills and limits and, thanks to a local predictor, it is able to assess if it is possible to execute a requested task and, if not, to propose alternative by relaxing some constraints. The controller is interfaced with a communication infrastructure that enables human and robot to set up a bidirectional communication about the task to execute and the human to take an informed decision on the behavior of the robot. A comparative experimental validation is proposed.
We propose entropy-preserving and entropy-stable partitioned Runge--Kutta (RK) methods. In particular, we extend the explicit relaxation Runge--Kutta methods to IMEX--RK methods and a class of explicit second-order multirate methods for stiff problems arising from scale-separable or grid-induced stiffness in a system. The proposed approaches not only mitigate system stiffness but also fully support entropy-preserving and entropy-stability properties at a discrete level. The key idea of the relaxation approach is to adjust the step completion with a relaxation parameter so that the time-adjusted solution discretely satisfies the entropy condition. The relaxation parameter is computed by solving a scalar nonlinear equation at each timestep in general; however, as for a quadratic entropy function, we theoretically derive the explicit form of the relaxation parameter and numerically confirm that the relaxation parameter work for the Burgers equation. Several numerical results for ordinary differential equations and the Burgers equation are presented to demonstrate the entropy-conserving/stable behavior of these methods. We also compare the relaxation approach and the incremental direction technique for the Burgers equation with and without a limiter in the presence of shocks.
Multi-agent deep reinforcement learning has been applied to address a variety of complex problems with either discrete or continuous action spaces and achieved great success. However, most real-world environments cannot be described by only discrete action spaces or only continuous action spaces. And there are few works having ever utilized deep reinforcement learning (drl) to multi-agent problems with hybrid action spaces. Therefore, we propose a novel algorithm: Deep Multi-Agent Hybrid Soft Actor-Critic (MAHSAC) to fill this gap. This algorithm follows the centralized training but decentralized execution (CTDE) paradigm, and extend the Soft Actor-Critic algorithm (SAC) to handle hybrid action space problems in Multi-Agent environments based on maximum entropy. Our experiences are running on an easy multi-agent particle world with a continuous observation and discrete action space, along with some basic simulated physics. The experimental results show that MAHSAC has good performance in training speed, stability, and anti-interference ability. At the same time, it outperforms existing independent deep hybrid learning method in cooperative scenarios and competitive scenarios.
Feature representations derived from models pre-trained on large-scale datasets have shown their generalizability on a variety of audio analysis tasks. Despite this generalizability, however, task-specific features can outperform if sufficient training data is available, as specific task-relevant properties can be learned. Furthermore, the complex pre-trained models bring considerable computational burdens during inference. We propose to leverage both detailed task-specific features from spectrogram input and generic pre-trained features by introducing two regularization methods that integrate the information of both feature classes. The workload is kept low during inference as the pre-trained features are only necessary for training. In experiments with the pre-trained features VGGish, OpenL3, and a combination of both, we show that the proposed methods not only outperform baseline methods, but also can improve state-of-the-art models on several audio classification tasks. The results also suggest that using the mixture of features performs better than using individual features.
The security of quantum key distribution (QKD) is severely threatened by discrepancies between realistic devices and theoretical assumptions. Recently, a significant framework called the reference technique was proposed to provide security against arbitrary source flaws, including pulse correlations. Here, we propose an efficient four-phase twin-field QKD using laser pulses adopting the reference technique for security against potential source imperfections. We present a characterization of source flaws and connect them to experiments, together with a finite-key analysis against coherent attacks. In addition, we demonstrate the feasibility of our protocol through a proof-of-principle experimental implementation and achieve a secure key rate of 258 bps with a 20 dB channel loss. Compared with previous QKD protocols with imperfect devices, our study considerably improves both the secure key rate and the transmission distance, and shows application potential in the practical deployment of secure QKD with device imperfections.
Balancing safety and performance is one of the predominant challenges in modern control system design. Moreover, it is crucial to robustly ensure safety without inducing unnecessary conservativeness that degrades performance. In this work we present a constructive approach for safety-critical control synthesis via Control Barrier Functions (CBF). By filtering a hand-designed controller via a CBF, we are able to attain performant behavior while providing rigorous guarantees of safety. In the face of disturbances, robust safety and performance are simultaneously achieved through the notion of Input-to-State Safety (ISSf). We take a tutorial approach by developing the CBF-design methodology in parallel with an inverted pendulum example, making the challenges and sensitivities in the design process concrete. To establish the capability of the proposed approach, we consider the practical setting of safety-critical design via CBFs for a connected automated vehicle (CAV) in the form of a class-8 truck without a trailer. Through experimentation we see the impact of unmodeled disturbances in the truck's actuation system on the safety guarantees provided by CBFs. We characterize these disturbances and using ISSf, produce a robust controller that achieves safety without conceding performance. We evaluate our design both in simulation, and for the first time on an automotive system, experimentally.
Detection of object anomalies is crucial in industrial processes, but unsupervised anomaly detection and localization is particularly important due to the difficulty of obtaining a large number of defective samples and the unpredictable types of anomalies in real life. Among the existing unsupervised anomaly detection and localization methods, the NF-based scheme has achieved better results. However, the two subnets (complex functions) $s_{i}(u_{i})$ and $t_{i}(u_{i})$ in NF are usually multilayer perceptrons, which need to squeeze the input visual features from 2D flattening to 1D, destroying the spatial location relationship in the feature map and losing the spatial structure information. In order to retain and effectively extract spatial structure information, we design in this study a complex function model with alternating CBAM embedded in a stacked $3\times3$ full convolution, which is able to retain and effectively extract spatial structure information in the normalized flow model. Extensive experimental results on the MVTec AD dataset show that CAINNFlow achieves advanced levels of accuracy and inference efficiency based on CNN and Transformer backbone networks as feature extractors, and CAINNFlow achieves a pixel-level AUC of $98.64\%$ for anomaly detection in MVTec AD.
Timed automata are a common formalism for the verification of concurrent systems subject to timing constraints. They extend finite-state automata with clocks, that constrain the system behavior in locations, and to take transitions. While timed automata were originally designed for safety (in the wide sense of correctness w.r.t. a formal property), they were progressively used in a number of works to guarantee security properties. In this work, we review works studying security properties for timed automata in the last two decades. We notably review theoretical works, with a particular focus on opacity, as well as more practical works, with a particular focus on attack trees and their extensions. We derive main conclusions concerning open perspectives, as well as tool support.
Artificial neural networks thrive in solving the classification problem for a particular rigid task, acquiring knowledge through generalized learning behaviour from a distinct training phase. The resulting network resembles a static entity of knowledge, with endeavours to extend this knowledge without targeting the original task resulting in a catastrophic forgetting. Continual learning shifts this paradigm towards networks that can continually accumulate knowledge over different tasks without the need to retrain from scratch. We focus on task incremental classification, where tasks arrive sequentially and are delineated by clear boundaries. Our main contributions concern 1) a taxonomy and extensive overview of the state-of-the-art, 2) a novel framework to continually determine the stability-plasticity trade-off of the continual learner, 3) a comprehensive experimental comparison of 11 state-of-the-art continual learning methods and 4 baselines. We empirically scrutinize method strengths and weaknesses on three benchmarks, considering Tiny Imagenet and large-scale unbalanced iNaturalist and a sequence of recognition datasets. We study the influence of model capacity, weight decay and dropout regularization, and the order in which the tasks are presented, and qualitatively compare methods in terms of required memory, computation time, and storage.
The goal of few-shot learning is to learn a classifier that generalizes well even when trained with a limited number of training instances per class. The recently introduced meta-learning approaches tackle this problem by learning a generic classifier across a large number of multiclass classification tasks and generalizing the model to a new task. Yet, even with such meta-learning, the low-data problem in the novel classification task still remains. In this paper, we propose Transductive Propagation Network (TPN), a novel meta-learning framework for transductive inference that classifies the entire test set at once to alleviate the low-data problem. Specifically, we propose to learn to propagate labels from labeled instances to unlabeled test instances, by learning a graph construction module that exploits the manifold structure in the data. TPN jointly learns both the parameters of feature embedding and the graph construction in an end-to-end manner. We validate TPN on multiple benchmark datasets, on which it largely outperforms existing few-shot learning approaches and achieves the state-of-the-art results.
小貼士
登錄享
相關主題
注冊地址： 北京市海淀區羊坊店路18號2幢3層301-191