A5102925222- EEG and Brain-Computer Interfaces
- Robotic Path Planning Algorithms
- Functional Brain Connectivity Studies
- Indoor and Outdoor Localization Technologies
- Speech and Audio Processing
- Autonomous Vehicle Technology and Safety
- 3D Surveying and Cultural Heritage
- Advanced Adaptive Filtering Techniques
- Advanced Neural Network Applications
- Neural dynamics and brain function
- 3D Shape Modeling and Analysis
- Topic Modeling
- Human Motion and Animation
- Data Quality and Management
- Integrated Energy Systems Optimization
- Time Series Analysis and Forecasting
- Power Line Communications and Noise
- Customer churn and segmentation
- Reinforcement Learning in Robotics
- Traffic Prediction and Management Techniques
- Air Traffic Management and Optimization
- Millimeter-Wave Propagation and Modeling
- Smart Grid and Power Systems
- Solar Radiation and Photovoltaics
- Adversarial Robustness in Machine Learning
Shenyang Jianzhu University
2023
Nanjing University of Information Science and Technology
2023
Beijing Jiaotong University
2022-2023
InternetLab
2023
University College London
2023
Horizon Robotics (China)
2022
University of California, San Diego
2020-2021
Chinese University of Hong Kong
2015
University of Hong Kong
2015
Modern autonomous driving system is characterized as modular tasks in sequential order, i.e., perception, prediction, and planning. In order to perform a wide diversity of achieve advanced-level intelligence, contemporary approaches either deploy standalone models for individual tasks, or design multi-task paradigm with separate heads. However, they might suffer from accumulative errors deficient task coordination. Instead, we argue that favorable framework should be devised optimized...
There have been two streams in the 3D detection from point clouds: single-stage methods and two-stage methods. While former is more computationally efficient, latter usually provides better accuracy. By carefully examining approaches, we found that if appropriately designed, first stage can produce accurate box regression. In this scenario, second mainly rescores boxes such with localization get selected. From observation, devised a anchor-free network fulfill these requirements. This...
To probe deeply into the underlying mechanism of flight delay propagation, we construct propagation networks among airports via refined nonlinear Granger causality, which is more fitting on airport data and obtain precise causal estimation. Specifically, analysis method used to determine relationship between airports. The these pairs constitutes an network, in nodes edges respectively, represent specific Complex network theory then applied analyse global structure networks, indicating that...
This paper presents our 2nd place solution for the NuPlan Challenge 2023. Autonomous driving in real-world scenarios is highly complex and uncertain. Achieving safe planning multimodal a challenging task. Our approach, Imitation with Spatial-Temporal Heatmap, adopts learning form of behavior cloning, innovatively predicts future states heatmap representation, uses trajectory refinement techniques to ensure final safety. The experiment shows that method effectively balances vehicle's progress...
China's demand for solar energy has been growing rapidly to meet transformation targets. However, the potential of is affected by weather conditions and expected change under climate warming. Here, we project photovoltaic (PV) power over China low high emission scenarios 2060s, taking advantage meteorological variables from 24 CMIP6 models 4 PV with varied formats. The ensemble mean these yields an average 277.2 KWh m−2 yr−1 during 2004–2014, a decreasing tendency west east. By 2054–2064,...
In this report, we introduce our solution to the Occupancy and Flow Prediction challenge in Waymo Open Dataset Challenges at CVPR 2022, which ranks 1st on leaderboard. We have developed a novel hierarchical spatial-temporal network featured with encoders, multi-scale aggregator enriched latent variables, recursive 3D decoder. use multiple losses including focal loss modified flow trace efficiently guide training process. Our method achieves Flow-Grounded AUC of 0.8389 outperforms all other teams
We present a system for localizing sound sources in room with several <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">ad hoc</i> microphone arrays. Each circular array performs direction of arrival (DOA) estimation independently using commercial software. The DOAs are fed to fusion center, concatenated, and used perform the localization based on two proposed methods, which require only few labeled source locations (anchor points) training....
α Oscillations in sensory cortex, under frontal control, desynchronize during attentive preparation. Here, a selective attention study with simultaneous EEG humans of either sex, we first demonstrate that diminished anticipatory synchrony between the mid-frontal region dorsal network and ventral visual cortex [frontal-sensory (FSS)] significantly correlates greater task performance. Then, double-blind, randomized controlled healthy adults, implement closed-loop neurofeedback (NF) FSS signal...
Phase-pure synthesis has been a major challenge for metal oxynitrides due to their sensitivity conditions and the limited understanding of underlying thermodynamics. The beta-phase tantalum oxynitride (beta-TaON), promising material applications in photocatalysis energy storage, is particularly difficult synthesize reproducible, phase-pure form. In this work, computational thermodynamic model with experimental validation presented evaluate beta-TaON via ammonolysis reactions....
As autonomous driving systems being deployed to millions of vehicles, there is a pressing need improving the system's scalability, safety and reducing engineering cost. A realistic, scalable, practical simulator world highly desired. In this paper, we present an efficient solution based on generative models which learns dynamics scenes. With model, can not only simulate diverse futures given scenario but also generate variety scenarios conditioned various prompts. Our innovative design...
Alpha oscillations in sensory cortex, under frontal control, desynchronize during attentive preparation. Here, a selective attention study with simultaneous EEG, we first demonstrate that diminished anticipatory alpha synchrony between the mid-frontal region of dorsal network and ventral visual cortex (frontal-sensory (FSS)) significantly correlates greater task performance. Then, double-blind, randomized controlled healthy adults, implement closed-loop neurofeedback FSS signal over ten days...
The traditional K-means clustering center initial value algorithm does not perform well in massive data. In order to improve the initialization efficiency of center, this article adopted idea dynamic for inter region data classify data, and selected dense grids as based on distribution characteristics This method can ensure that cluster is located a high-density area rather than being overly concentrated, allowing K-Means quickly efficiently initialize when processing Finally, compared...
The construction of new power system with energy as the principal part is being promoted, which poses challenges to safety, economy, and stability system. It requires more regulatory resources stronger capabilities. Based on integrated grid operation smart (OS2) China Southern Power Grid, a deployment architecture for source-grid-load-storage collaborative control proposed. In this architecture, platform deployed in Zone III, can expand functions OS2 through data model interaction, realize...
Modern autonomous driving system is characterized as modular tasks in sequential order, i.e., perception, prediction, and planning. In order to perform a wide diversity of achieve advanced-level intelligence, contemporary approaches either deploy standalone models for individual tasks, or design multi-task paradigm with separate heads. However, they might suffer from accumulative errors deficient task coordination. Instead, we argue that favorable framework should be devised optimized...
There have been two streams in the 3D detection from point clouds: single-stage methods and two-stage methods. While former is more computationally efficient, latter usually provides better accuracy. By carefully examining approaches, we found that if appropriately designed, first stage can produce accurate box regression. In this scenario, second mainly rescores boxes such with localization get selected. From observation, devised a anchor-free network fulfill these requirements. This...