A5079927285- Fault Detection and Control Systems
- Adversarial Robustness in Machine Learning
- Energy Efficient Wireless Sensor Networks
- Model Reduction and Neural Networks
- Medical Image Segmentation Techniques
- Numerical Methods and Algorithms
- Formal Methods in Verification
- Opportunistic and Delay-Tolerant Networks
- Fuel Cells and Related Materials
- Control Systems and Identification
- Real-time simulation and control systems
- Mobile Ad Hoc Networks
- Sparse and Compressive Sensing Techniques
- Remote Sensing and LiDAR Applications
- Risk and Safety Analysis
- Interactive and Immersive Displays
- Augmented Reality Applications
- Vehicle Dynamics and Control Systems
- Peer-to-Peer Network Technologies
- Extremum Seeking Control Systems
- Bluetooth and Wireless Communication Technologies
- Advanced Neural Network Applications
- Business Process Modeling and Analysis
- Artificial Immune Systems Applications
- Context-Aware Activity Recognition Systems
Southwest University
2020-2024
Shanghai Key Laboratory of Trustworthy Computing
2016-2023
East China Normal University
2013-2023
Jiangxi Science and Technology Normal University
2014-2015
University of Georgia
1997-2002
This paper presents an approach of safety verification based on neural networks for continuous dynamical systems which are modeled as a system ordinary differential equations. We adopt the deductive methods barrier certificates. These functions over states with certain constraints existence entails under consideration. propose to represent function by and provide comprehensive synthesis framework. In particular, we devise new type activation functions, i.e., Bent-ReLU, networks; sampling...
Navigating mobile User Interface (UI) applications using large language and vision models based on high-level goal instructions is emerging as an important research field with significant practical implications, such digital assistants automated UI testing. To evaluate the effectiveness of existing in navigation, benchmarks are required widely used literature. Although multiple have been recently established for evaluating functional correctness being judged pass or fail, they fail to...
Benefit from less computational difficulty, barrier certificate based method has attracted much attention in safety verification of hybrid systems. Barrier certificates are inherent existences a system and may have different types. A set well-defined conditions is prerequisite for successfully identifying specific type. Therefore, how to define that can identify invisible existing becomes an essential problem verification. This paper proposes helps construct new type certificate, namely, the...
This paper presents a novel approach to safety verification based on neural barrier certificates synthesis for continuous dynamical systems. We construct the framework as an inductive loop between Learner and Verifier certificate learning counterexample guidance. Compared with counterexample-guided method SMT solver, we design learn functions special structure, use form convert generation into polynomial optimization problem obtaining optimal counterexample. In phase, task of identifying...
Bayesian neural networks (BNNs) retain NN structures with a probability distribution placed over their weights. With the introduced uncertainties and redundancies, BNNs are proper choices of robust controllers for safety-critical control systems. This paper considers problem verifying safety nonlinear closed-loop systems BNN unbounded-time horizon. In essence, we compute safe weight set such that as long controller is always applied weights sampled from set, controlled system guaranteed to...
For multilayer mobile sensor network, the issues of improving network coverage by use sensors are studied. A coverage-optimized deployment algorithm based on grid-division and bipartite graph matching is proposed. Firstly, area divided into many grids build distribution matrix nodes. Then construct a G solve maximum cardinality G. It corresponds to an optimal scheme which maximizes minimizes total movement cost. Results show that after optimized increases, with increase in percentage nodes...
We provide a novel approach to synthesize controllers for nonlinear continuous dynamical systems with control against safety properties. The are based on neural networks (NNs). To certify the property we utilize barrier functions, which represented by NNs as well. train controller-NN and barrier-NN simultaneously, achieving verification-in-the-loop synthesis. prototype tool nncontroller number of case studies. experiment results confirm feasibility efficacy our approach.
The problem of figure-ground separation is modeled as one energy minimization using the Ising system model from quantum physics. for makes explicit definition shape in terms attributes such cocircularity, smoothness, proximity and contrast based on formulation an function that incorporates pair wise interactions between local image features form edgels. paper explores a class stochastic optimization techniques evolutionary algorithms context model. Experimental results synthetic edgel maps...
An intrusion detection system (IDS) using agent technology was designed for wireless sensor network of clustering structure. IDS which includes two different agents deployed in every node network. One is local and another global agent. They complete tasks detection. Based on Bluetooth communication technology, scattering formation algorithm TPSF employed to construct the cluster layer finish task assignment agents. The improved by limiting role nodes lighten complexity nodes, so can work...
According to the multilayer sensor network structure of mobile nodes, proposing a coverage optimization control policy based on voronoi division. Based Sink nodes by division, essay proposes algorithm (COCA) hole calculating optimize initial random deployment and then eliminate holes improve performance network. Simulation results show that, compared with deployed network, COCA only requires fewer achieve higher ratio maintain in low energy consumption, which embodies optimum deployment.
For sparse overlay network built by static sensor nodes, this paper proposes the introduction of heterogeneous mobile and designs a coverage strategy to optimize performance network. At first, monitoring area is divided into several sub-regions iterative quartering are represented quad-tree structure. Secondly, planned based on idea divide conquers combing overall regional with local strategy. The go through all according breadth-first traversal quad-tree. in each sub-region designed...
We present a new adaptive denosing method using compressive sensing (CS) and genetic algorithm (GA). use Regularized Orthogonal Matching Pursuit (ROMP) to remove the noise of image. ROMP has advantage correct performance, stability fast speed. In order obtain optimal denoising effect, we determine values parameters by GA. Experimental results show that proposed can image effectively. Compared with other traditional methods, retains most abundant edge information important details Therefore,...
The control of mobile coverage model is the focus research wireless sensor network. If a node can balance network load according to needs, which greatly reduce consumption network, improve transmission efficiency and prolong life cycle. Accordingly, this paper uses probability principle multilevel model, improves enhances energy-saving effect nodes. In order verify effectiveness reliability MATLAB simulation platform test performance design program region division moving path for realizes...