A5005995567- Autonomous Vehicle Technology and Safety
- Robotics and Sensor-Based Localization
- Robotic Path Planning Algorithms
- Advanced Vision and Imaging
- Advanced Neural Network Applications
- Video Surveillance and Tracking Methods
- Remote Sensing and LiDAR Applications
- Cooperative Communication and Network Coding
- Advanced Image and Video Retrieval Techniques
- Vehicle Dynamics and Control Systems
- Automated Road and Building Extraction
- Mobile Ad Hoc Networks
- Advanced Optical Sensing Technologies
- Advanced Measurement and Detection Methods
- Spectroscopy and Chemometric Analyses
- Software-Defined Networks and 5G
- Catalytic Processes in Materials Science
- Domain Adaptation and Few-Shot Learning
- 3D Surveying and Cultural Heritage
- Control and Dynamics of Mobile Robots
- Wireless Networks and Protocols
- Water Quality Monitoring and Analysis
- Full-Duplex Wireless Communications
- Advanced Wireless Communication Technologies
- Simulation and Modeling Applications
Shihezi University
2012-2025
Huazhong University of Science and Technology
2010-2024
National University of Defense Technology
2015-2024
Systems Technology (United States)
2024
PLA Academy of Military Science
2018-2024
Xinjiang Production and Construction Corps
2012-2023
Zhejiang University of Technology
2023
Academy of Military Medical Sciences
2021-2022
EY Technologies (United States)
2022
Halliburton (United Kingdom)
2015-2021
Deep reinforcement learning (DRL) integrates the feature representation ability of deep with decision-making so that it can achieve powerful end-to-end control capabilities. In past decade, DRL has made substantial advances in many tasks require perceiving high-dimensional input and making optimal or near-optimal decisions. However, there are still challenging problems theory applications DRL, especially limited samples, sparse rewards, multiple agents. Researchers have proposed various...
This article presents a novel path planning algorithm for autonomous land vehicles. There are four main contributions: Firstly, an evaluation standard is introduced to measure the performance of different algorithms and select appropriate parameters proposed algorithm. Secondly, guideline generated by human or global employed develop heuristic function overcome shortcoming traditional A-Star algorithms. Thirdly, improving obstacle avoidance performance, key points around employed, which...
In this paper, we consider the multivariate Bernoulli distribution as a model to estimate structure of graphs with binary nodes. This is discussed in framework exponential family, and its statistical properties regarding independence nodes are demonstrated. Importantly can not only main effects pairwise interactions among but also capable modeling higher order interactions, allowing for existence complex clique effects. We compare existing graphical inference models - Ising Gaussian model,...
In this paper, we propose to fuse the LIDAR and monocular image in framework of conditional random field detect road robustly challenging scenarios. points are aligned with pixels by cross calibration. Then boosted decision tree based classifiers trained for point cloud respectively. The scores two kinds treated as unary potentials corresponding pixel nodes field. fused can be solved efficiently graph cut. Extensive experiments tested on KITTI-Road benchmark show that our method reaches...
In the current health care field, benefit from provision of medical big data, machine learning can be used to obtain knowledge data. Machine methods describe cases an objective perspective, and predictions diagnostic results generated a combination related pathological factors. The introduction for way diagnosis accuracy is major change inevitable direction future model. this paper, random forest algorithm analyze case breast cancer. combine characteristics multiple eigenvalues, combined...
Multi-camera 3D perception has emerged as a prominent research field in autonomous driving, offering viable and cost-effective alternative to LiDAR-based solutions. The existing multi-camera algorithms primarily rely on monocular 2D pre-training. However, the pre-training overlooks spatial temporal correlations among system. To address this limitation, we propose first unified framework, called UniScene, which involves initially reconstructing scene foundational stage subsequently...
Accurate maneuver prediction for surrounding vehicles enables intelligent to make safe and socially compliant decisions in advance, thus improving the safety comfort of driving. The main contribution this paper is proposing a practical, high-performance, low-cost maneuver-prediction approach vehicles. Our based on dynamic Bayesian network, which exploits multiple predictive features, namely, historical states predicting vehicles, road structures, as well traffic interactions inferring...
Multi-object tracking is a crucial problem for autonomous vehicle. Most state-of-the-art approaches adopt the tracking-by-detection strategy, which two-step procedure consisting of detection module and module. In this paper, we improve both steps. We by incorporating temporal information, beneficial detecting small objects. For module, propose novel compressed deep Convolutional Neural Network (CNN) feature based Correlation Filter tracker. By carefully integrating these two modules,...
Freespace detection is an essential component of autonomous driving technology and plays important role in trajectory planning. In the last decade, deep learning based freespace methods have been proved feasible. However, these efforts were focused on urban road environments few specifically designed for off-road due to lack dataset benchmark. this paper, we present ORFD dataset, which, our knowledge, first dataset. The was collected different scenes (woodland, farmland, grassland...
In this paper, we summarize the 1st NTIRE challenge on stereo image super-resolution (restoration of rich details in a pair low-resolution images) with focus new solutions and results. This has 1 track aiming at problem under standard bicubic degradation. total, 238 participants were successfully registered, 21 teams competed final testing phase. Among those participants, 20 submitted results PSNR (RGB) scores better than baseline. establishes benchmark for SR.
Current perception models in autonomous driving heavily rely on large-scale labelled 3D data, which is both costly and time-consuming to annotate. This work proposes a solution reduce the dependence training data by leveraging pre-training unlabeled outdoor LiDAR point clouds using masked autoencoders (MAE). While existing autoencoding methods mainly focus small-scale indoor or pillar-based our approach introduces new self-supervised occupancy method called Occupancy-MAE, specifically...
To address the challenge of users selecting rich tourism resources, this study proposes a model for cultural attraction recommendation using an optimized weighted association rule algorithm. This includes time and season weight tourist recommendations. improvement methods to some inherent issues in traditional models. Firstly, it constructed attractions, then algorithm by incorporating dynamic weights. It takes into account user's intended outcome. Moreover, incorporated seasonal weights...
Density functional theory (DFT) calculations were used to study the mechanism for hydrochlorination of acetylene catalyzed by MCl x (M = Hg, Au, Ru; 2, 3). For three catalysts, reaction occurs via a one-shift chlorine atom transfer, which avoids formation highly stable complex species. The adsorbed HCl acts as donor, while C 2 H favors abstraction. calculated real activation barrier decreases in order: HgCl > AuCl 3 RuCl , suggests that would be good candidate catalyst acetylene.
Dynamic vehicle detection and tracking is crucial for self-driving in urban environments. The main problem of the previous beam-model-based algorithms that they cannot detect track dynamic vehicles are occluded by other objects. In this paper, we develop a novel algorithm to solve our autonomous land (ALV), which equipped with Velodyne LIDAR GPS-aid inertial navigation system. For detection, improved two-dimensional virtual scan presented potential differencing operation. Then, each vehicle,...
Deep learning has recently demonstrated its promising performance for vision-based parking-slot detection. However, very few existing methods explicitly take into account the link information of marking-points, resulting in complex post-processing and erroneous In this paper, we propose an attentional graph neural network based detection method, which refers marking-points around-view image as graph-structured data utilize to aggregate neighboring between marking-points. Without any manually...
Abstract For autonomous driving, traversability analysis is one of the most basic and essential tasks. In this paper, we propose a novel LiDAR‐based terrain modeling approach, which could output stable, complete, accurate models results. As an inherent property environment that does not change with different view angles, our approach adopts multiframe information fusion strategy for modeling. Specifically, normal distributions transform mapping adopted to accurately model by fusing from...
Ru single‐atom catalysts hold great promise for the robust synthesis of vinyl chloride through acetylene hydrochlorination. However, easy over‐chlorination atoms during reaction suppress catalytic activity and stability. Herein, we have synthesized an oxygen doped catalyst by a sequential etching strategy, which delivers remarkable yield monomer (>99.38%) stability (>900 h, 180 h‐1), far beyond those reported counterparts. Experimental results theoretical calculations reveal that...
Inspired by the success of DeepSeek-R1, we explore potential rule-based reinforcement learning (RL) in large reasoning models. To analyze dynamics, use synthetic logic puzzles as training data due to their controllable complexity and straightforward answer verification. We make some key technical contributions that lead effective stable RL training: a system prompt emphasizes thinking answering process, stringent format reward function penalizes outputs for taking shortcuts, recipe achieves...