A5072408261- Reinforcement Learning in Robotics
- Robotic Path Planning Algorithms
- Artificial Intelligence in Games
- AI-based Problem Solving and Planning
- Greenhouse Technology and Climate Control
- Business Process Modeling and Analysis
- Adversarial Robustness in Machine Learning
- Anomaly Detection Techniques and Applications
- Guidance and Control Systems
- Gambling Behavior and Treatments
- Irrigation Practices and Water Management
- Sports Analytics and Performance
- Evolutionary Algorithms and Applications
- Domain Adaptation and Few-Shot Learning
- Advanced Neural Network Applications
- Advanced Image and Video Retrieval Techniques
- Smart Agriculture and AI
- Spectroscopy and Chemometric Analyses
- Machine Learning and Data Classification
- Multi-Agent Systems and Negotiation
- Advanced Bandit Algorithms Research
- Face and Expression Recognition
- Network Security and Intrusion Detection
- Visual Attention and Saliency Detection
- Digital Games and Media
National University of Defense Technology
2014-2024
National Center of Biomedical Analysis
2024
Hubei Normal University
2024
Xidian University
2023
Defence Electronics Research Laboratory
2023
Tsinghua–Berkeley Shenzhen Institute
2020-2022
Tsinghua University
2020-2022
University Town of Shenzhen
2022
Tencent (China)
2021
Fuyang Normal University
2021
In the field of satellite imagery, remote sensing image captioning (RSIC) is a hot topic with challenge overfitting and difficulty text alignment. To address these issues, this paper proposes vision-language aligning paradigm for RSIC to jointly represent vision language. First, new dataset DIOR-Captions built augmenting object detection in optical (DIOR) images manually annotated Chinese English contents. Second, Vision-Language model Cross-modal Attention (VLCA) presented generate accurate...
Hexagonal grids use a hierarchical subdivision tessellation to cover the entire plane or sphere. Due 6-fold rotational symmetry, hexagonal have some advantages (e.g. isoperimetry, equidistant neighbors, and uniform connectivity) over quadrangular triangular girds, which makes them suitable tackle tasks of geospatial information processing intelligent decision-making. In this paper, we first introduce applications based on grids. Then, planer spherical analyze group representations for them,...
Due to the high efficiency and less weather dependency, autonomous greenhouses provide an ideal solution meet increasing demand for fresh food. However, managers are faced with some challenges in finding appropriate control strategies crop growth, since decision space of greenhouse problem is astronomical number. Therefore, intelligent closed-loop framework highly desired generate automatic policy. As a powerful tool optimal control, reinforcement learning (RL) algorithms can surpass human...
Moving objects detection in moving background is a prerequisite for many analyses and applications computer vision. The most challenging part of that motion induced by camera may dominate the observed motion, which makes existing methods cannot detect real from robustly computationally efficient. In this paper, we proposed very fast method can accurately without prior knowledge characteristics. Frame difference caused compensated Oriented FAST Rotated BRIEF (ORB) features matching....
The rapidly growing global population presents challenges and demands for efficient production of healthy fresh food. Autonomous greenhouse equipped with standard sensors actuators (such as heating lighting) which enables control indoor climate crop production, contributes to producing higher yields. However, it requires skilled expensive labor, well a large amount energy. An autonomous strategy, powered by AI algorithms optimizing the yields resource use simultaneously, offers an ideal...
Model-based deep reinforcement learning has achieved success in various domains that require high sample efficiencies, such as Go and robotics. However, there are some remaining issues, planning efficient explorations to learn more accurate dynamic models, evaluating the uncertainty of learned rational utilization models. To mitigate these we present MEEE, a model-ensemble method consists optimistic exploration weighted exploitation. During exploration, unlike prior methods directly...
Agriculture is the foundation of human civilization. However, rapid increase global population poses a challenge on this cornerstone by demanding more food. Modern autonomous greenhouses, equipped with sensors and actuators, provide promising solution to problem empowering precise control for high-efficient food production. optimal greenhouses challenging, requiring decision-making based high-dimensional sensory data, scaling production limited scarcity labor capable handling task. With...
This paper investigates cooperative trajectory planning of multiple unmanned combat aerial vehicles (multi-UCAV) in performing autonomous air-to-ground target attack missions. By integrating an approximate allowable region model, several constraint models, and a multicriterion objective function, the problem is formulated as optimal control (CTOCP). Then, virtual motion camouflage (VMC) for multi-UCAV, combining with differential flatness theory, Gauss pseudospectral method (GPM), nonlinear...
In this paper, we investigate the exploration-exploitation dilemma of reinforcement learning algorithms. We adapt information directed sampling, an exploration framework that measures gain a policy, to continuous learning. To stabilize off-policy process and further improve sample efficiency, propose use randomized target dynamically adjust update-to-data ratio for different parts neural network model. Experiments show our approach significantly improves over existing methods successfully...
This study explored the effects of varying time pulsed ultrasound (PUS) treatment on physicochemical and textural properties chicken myofibrillar protein (CMP) gel. The solubility rapidly increased at ≤ 6 min then steadily decreased, while particle size showed opposite trend. At longer PUS times, total sulfhydryl（-SH）and reactive SH content CMP gel all decreased. absolute value zeta potential surface hydrophobicity were higher. most hydrogen bonds formed. G' G″ also optimal, indicating that...
Abstract Agriculture is the foundation of human civilization. However, rapid increase and aging global population pose challenges on this cornerstone by demanding more healthy fresh food. Internet Things (IoT) technology makes modern autonomous greenhouse a viable reliable engine food production. educated skilled labor capable overseeing high-tech greenhouses scarce. Artificial intelligence (AI) cloud computing technologies are promising solutions for precision control high-efficiency...
This study addressed a problem of rapid velocity consensus within swarm unmanned aerial vehicles. Our analytical framework was based on tools using matrix theory and algebraic graph theory. We established connections between connectivity the speed converging velocity. The relationship communication cost established. To deal with trade-off among connectivity, convergence cost, we propose distributed small world network construction method. characteristics expedite toward in vehicle swarm....
Principal Component Analysis (PCA) is a popular tool for dimension reduction and feature extraction in data analysis. Probabilistic PCA (PPCA) extends the standard by using probabilistic model. However, both PPCA are not robust, as they sensitive to outliers. To alleviate this problem, we propose novel method called Self-Paced (SP-PPCA) introducing Learning mechanism into PPCA. Furthermore, design corresponding optimization algorithm based on an alternative search strategy...
Real-time strategy (RTS) games can be seen as simulating real and complex dynamic environments in a limited small world, posing important challenges for the development of artificial intelligence. Existing applications intelligence technology RTS are not yet able to compete with professional human players. But there already ways control macro games, they amateur an excellent platform testing technology, more smarter methods being used overall it. The purpose this paper is systematically...
The vast amount of data in the video requires people to check and identify, that is, it is necessary extract a short possible image segment from long-time or massive data. Technicians are prone fatigue long-term high-load work, there inevitable omissions, computers can continue work reliably as long they have electricity. paper aimed at dangerous targets may appear data, especially unidentified personnel, vehicles, aircraft, etc. computer semantic recognition system automatically quickly...
The Monte Carlo Tree Search (MCTS) has demonstrated excellent performance in solving many planning problems. However, the state space and branching factors are huge, horizon is long practical applications, especially adversarial environment. It computationally expensive to cover a sufficient number of rewarded states that far away from root flat non-hierarchical MCTS. Therefore, MCTS inefficient for dealing with problems horizon, huge space, factors. In this work, we propose novel...
Imperfect information game in multiplayer no-limit Texas Hold'em is a critical challenge AI research. Recent advanced solving approaches, such as deep CounterFactual Value networks(CFVnet) combined with continual resolving, provide way to conduct depth-limited search imperfect-information games. However, CFVnet has limited deployment Heads-Up No-Limit Hold'em, and hard scale setting. In this paper, we propose novel algorithm, mean approximation, that effectively converting multi-agent...
Abstract In this paper, the robustness of global exponentially stable (ES) linear time‐invariant (LTI) systems with feedback control protocols is examined under influence random disturbances (RDs). For a given LTI system ES, state (SF) protocol and output (OF) are proposed to keep discussed system, respectively. Meanwhile, how much intensity RDs that can withstand obtained by transcendental equation, in which gains also designed calculated iterative algorithm. The theoretical analysis...
In addressing the challenge of tracking moving targets at sea, our focus has been directed towards development a reconstruction methodology founded upon satellite orbital manoeuvres. This endeavour led us to devise predictive model for manoeuvres within geographic coordinate system, alongside creation three-phase manoeuvre model. A Non-dominant Sorting Adaptive Memetic (NSAM) algorithm is proposed in this paper, which two-layer multi-objective optimization that retains advantages...