Current Issue

15 March 2023, Volume 5 Issue 1

Previous Issue    Next Issue

Review Intelligence

The DAOs to AI for Science by DeSci: the state of the art and perspective

Fei-Yue WANG, Qinghai MIAO, Junping ZHANG, Wenbo ZHENG, Wenwen DING

2023, 5(1):  1-6.  doi:10.11959/j.issn.2096-6652.202310

Asbtract ( 833 )   HTML ( 254)   PDF (788KB) ( 1181 )   Knowledge map   

References | Related Articles | Metrics

The new wave of artificial intelligence technology represented by ChatGPT is promoting the comprehensive transformation of human society, the transformation of scientific research paradigm is accelerating, and an artificial intelligence-driven scientific research (AI for Science, AI4S) revolution is coming.The basic concepts and characteristics of AI4S were analyzed, and the development status of AI4S were briefly summarized from the perspectives of mathematics, physics, biology, and materials.Vigorously developing AI4S is of great significance to improving national competitiveness, developing social economy, and strengthening technical reserves.In order to promote the development of AI4S better, the following two points are essential: one is to change the contemporary teaching and education, and advocate AI for Education (AI4E) and Education for AI (E4AI); the other is to establish and adapt to the ＂new scientific research paradigm＂ with ＂new organization mode＂ in a ＂new research ecology＂, based on DAOs and DeSci, for open, fair and just sustainable support for AI4S.

Surveys and Prospectives

A survey of 3D object detection algorithms

Zhe HUANG, Yongcai WANG, Deying LI

2023, 5(1):  7-31.  doi:10.11959/j.issn.2096-6652.202312

Asbtract ( 1076 )   HTML ( 163)   PDF (6364KB) ( 978 )   Knowledge map   

Figures and Tables | References | Related Articles | Metrics

3D object detection is a fundamental problem in autonomous driving，virtual reality，robotics，and other applications.Its goal is to extract the most accurate 3D box characterizing interested targets from the disordered point clouds，such as the closest 3D box surrounding the pedestrians or vehicles.The target 3D box's location，size，and orientation are also output.Currently，there are two primary approaches for 3D object detection: (1) pure point cloud based 3D object detection，in which the point clouds are created by binocular vision，RGB-D camera，and lidar; (2) fusion-based 3D object detection based on the fusion of image and point cloud.The various representations of 3D point clouds were introduced.Then representative methods were introduced from three aspects: traditional machine learning techniques; non-fusion deep learning based algorithms; and multimodal fusion-based deep learning algorithms in progressive relation.The algorithms within and across each category were examined and compared，and the differences and connections between the various methods were analyzed thoroughly.Finally，remaining challenges of 3D object detection were discussed and explored.And the primary datasets and metrics used in 3D object detection studies were summarized.

Parallel transportation systems in era of metaverse

Qinghai MIAO, Yisheng LYU

2023, 5(1):  32-40.  doi:10.11959/j.issn.2096-6652.202302

Asbtract ( 235 )   HTML ( 37)   PDF (1253KB) ( 391 )   Knowledge map   

Figures and Tables | References | Related Articles | Metrics

As the metaverse receives more and more attention, how the metaverse will affect the development of the transportation system has also become a hot topic.The origin and current status of the metaverse were introduced, the incubation period, development period and maturity period of the future development of the metaverse were prospected, and the potential problems brought about by the metaverse were also discussed.Combined with the need theory of Maslow, it was analyzed and pointed out that the metaverse would gradually affect urban travel demands, and would bring about profound changes in traffic after entering the mature period.The parallel intelligence methods would play an essential role in guiding the healthy development of the metaverse and effectively serving social travel.The implementation of the parallel transportation system would play a key role in traffic planning, precise control, fine service, emergency management, etc.

Papers and Reports

Systems agriculture: modeling and control based on social and economic attributes of agriculture

Mengzhen KANG, Hequan SUN, Xiujuan WANG, Fei-Yue WANG

2023, 5(1):  41-50.  doi:10.11959/j.issn.2096-6652.202306

Asbtract ( 195 )   HTML ( 33)   PDF (2535KB) ( 368 )   Knowledge map   

Figures and Tables | References | Related Articles | Metrics

With the arrival and development of the fifth industrial revolution, biotechnology, information technology and artificial intelligence are deeply integrated.This provides strong support for the informatization and intelligentization of agriculture.Due to the social and economic characteristics of agriculture, it has become a consensus to build an agricultural physical and social information system toward both the planting system with biophysical properties and the management system with socio-economic properties.Inspired by systems biology, systems agriculture was proposed, which combined two other dimensions of villages and farmers information using the parallel agriculture framework to build and study the agricultural system.Supported by biotechnology, information technology and artificial intelligence technology, systems agriculture had involved monitoring and integrating multi-scale, multi-dimensional and multimodal information, and carried out systematic research covering agriculture, villages and farmers using systems theory, in order to serve the development of future villages.The intelligent technology and the specific cases involved in systems agriculture were summarized and analyzed, and the perspectives of systems agriculture were presented.

Application of Fit CutMix data augmentation algorithm based on saliency information in medical images

Xinhuan LUO, Yixuan WANG, Wei LI, Xi CHEN

2023, 5(1):  58-68.  doi:10.11959/j.issn.2096-6652.202307

Asbtract ( 209 )   HTML ( 27)   PDF (1857KB) ( 233 )   Knowledge map   

Figures and Tables | References | Related Articles | Metrics

Deep convolutional neural network is one of the mainstream algorithms in the field of image classification, but its training requires a large number of labeled data, which leads to over fitting on small datasets such as Alzheimer's medical images.Data augmentation can increase the amount of training data, and CutMix data augmentation algorithm has been widely used recently.However, the augmented images generated by the CutMix series methods often ignore the significant area of the original image, and the design of the label of the augmented image takes only single factor into consideration.In order to solve these problems, the Fit CutMix data augmentation algorithm was proposed.Firstly, the region replacement strategy based on the transfer of saliency extreme value was used to generate augmented samples, so as to concentrate the regions with high saliency value in the source samples and target samples.Secondly, the area and saliency information of the source samples and the target samples were combined to assign the augmented sample label, which provided effective supervision information for the convolutional neural network.The experimental results showed that when Fit CutMix was used in ResNet50 to diagnose Alzheimer's disease, the accuracy was 96.6%, which was about 7% higher than that of directly using ResNet50, and at least 3% higher than that of applying existing methods.Therefore, the Fit CutMix data augmentation algorithm can effectively improve the recognition accuracy of deep convolutional neural network for medical images.

Parallel reasoning: a virtual-real interactive knowledge collaboration framework based on ACP approach

Xiao WANG, Linyao YANG, Bin HU, Jiachen HOU

2023, 5(1):  69-82.  doi:10.11959/j.issn.2096-6652.202254

Asbtract ( 205 )   HTML ( 17)   PDF (1141KB) ( 452 )   Knowledge map   

Figures and Tables | References | Related Articles | Metrics

Knowledge graph represents empirical knowledge based on structured triples, which can effectively describe the semantic relationships between real-world entities.Knowledge graph has become a critical standard technology of the new generation of artificial intelligence.The development, typical applications of multi-source knowledge graphs, and problems of knowledge collaboration were summarized.A multi-source knowledge graph collaboration framework based on ACP approach, i.e., parallel reasoning, was proposed, which realized the extraction, fusion, completion, and unbiased application of multi-source heterogeneous knowledge based on artificial systems, computational experiments, and parallel execution.In the end, simulation experiments were conducted on power grid dispatching to evaluate the effectiveness of parallel learning for solving the management and control problems of complex systems.

Sidelink resource allocation algorithm of C-V2X based on deep Q learning

Hui XU

2023, 5(1):  83-91.  doi:10.11959/j.issn.2096-6652.202309

Asbtract ( 153 )   HTML ( 22)   PDF (895KB) ( 234 )   Knowledge map   

Figures and Tables | References | Related Articles | Metrics

For the sidelink resource autonomous selection scheme of different priority services in celluar-vehicle to everything (C-V2X) system, the procedure of autonomous selection algorithm based on reference signal energy was analyzed, and the energy threshold equation was designed.For the energy equation parameter estimation problem, energy-based autonomous selection algorithm was combined with deep Q learning algorithm, and the optimal parameter value of the energy threshold equation was obtained by iteration of the finite-degree algorithm.Simulation results showed that the side resource allocation algorithm based on deep Q learning could ensure the side resource requirements of V2X services with different priorities, and improve the packet reception ratio performance of the system.

Rapider-YOLOX: lightweight object detection network with high precision

Zhouyu GU, Yuecheng YU, Tiantian Zhe

2023, 5(1):  92-103.  doi:10.11959/j.issn.2096-6652.202303

Asbtract ( 251 )   HTML ( 17)   PDF (3374KB) ( 305 )   Knowledge map   

Figures and Tables | References | Related Articles | Metrics

As a lightweight network structure, YOLOX-Nano has the advantage of fast running speed.However, the model still has the defects of weak feature extraction ability and insufficient detection accuracy in practical application.Therefore, an efficient object detection network Rapider-YOLOX which comprehensively balanced the detection speed and detection accuracy was proposed.Firstly, the highly efficient bottleneck module was designed to improve the feature extraction capability of depthwise convolutional blocks in the original YOLOX-Nano model.Secondly, the soft-SPP module was designed to avoid the loss of some important information in the original SPP module and improve the ability of multi-scale information fusion and information exchange between channels further.Finally, CIoU was introduced to improve the position accuracy of the prediction box by using the center distance and aspect ratio between the prediction box and the real box.The experimental results on PASCAL VOC2007 dataset showed that the mAP of Rapider-YOLOX model reached 77.92%, which was 3.79% higher than the original YOLOX-Nano.In addition, on GT1030 with only 384 CUDA cores, the FPS of the proposed method could reach 45.40.The FPS could also reach 23.94 on the CPU, which further improved detection accuracy and generalization performance of the network while ensuring the lightweight characteristics of the network.

Curriculum design for artificial intelligence and quantitative trading

Junhuan ZHANG, Zhengyi ZHU, Kewei CAI

2023, 5(1):  104-112.  doi:10.11959/j.issn.2096-6652.202311

Asbtract ( 392 )   HTML ( 56)   PDF (840KB) ( 484 )   Knowledge map   

Figures and Tables | References | Related Articles | Metrics

With the development of computer technology, especially the development of artificial intelligence, big data technology and blockchain technology, the transaction mode of traditional economic society and financial market has been changed.Quantitative trading is an important emerging trading mode in the contemporary financial market.To meet the social demand for quantitative trading talents, it is particularly important to explore a reasonable course system of AI and quantitative trading.Firstly, the recent development of quantitative trading and the application of AI in quantitative trading were analyzed.Then, the current situations and problems of the quantitative trading course were summarized.Finally, according to the problems, the suggestions on the course design for AI and quantitative trading were put forward in four aspects, which included teaching content system, teaching practice simulation platform, teacher training, and multi-channel practice platform.

Cardinalized balanced multi-Bernoulli filter SLAM method based on pose graph optimization

Zijing ZHANG, Fei ZHANG

2023, 5(1):  113-120.  doi:10.11959/j.issn.2096-6652.202305

Asbtract ( 147 )   HTML ( 13)   PDF (4761KB) ( 111 )   Knowledge map   

Figures and Tables | References | Related Articles | Metrics

In the complex indoor environment, the traditional SLAM method based on random finite set theory has the problems of low robot pose accuracy and large amount of calculation.To solve these problems, a cardinalized balanced multi-Bernoulli filter SLAM method based on pose graph optimization was proposed.First of all, the cardinalized balanced multi-Bernoulli filter was used to estimate the map features, which avoided data association.What is more, an adaptive information control method was proposed to enrich the prior information.Then, the pose graph optimization theory was combined with cardinalized balanced multi-Bernoulli filter SLAM through adaptive information control method to optimize the pose estimation of the robot.Finally, through experimental comparative analysis, the results show that this method have better SLAM accuracy and real-time performance than the RB-PHD-SLAM method.

Metaverses and parallel systems: the state of the art, comparisons and prospects

Yonglin TIAN, Yuanwen CHEN, Jing YANG, Yutong WANG, Xiao WANG, Qinghai MIAO, Ziran WANG, Fei-Yue WANG

2023, 5(1):  121-132.  doi:10.11959/j.issn.2096-6652.202313

Asbtract ( 267 )   HTML ( 42)   PDF (2038KB) ( 372 )   Knowledge map   

Figures and Tables | References | Related Articles | Metrics

With the development of technologies such as artificial intelligence and virtual reality, digital technology is continuously changing and enriching human experiences and production methods, and has become a powerful tool for controlling and managing complex systems.Metaverses and parallel systems provide feasible ways for the construction of digital systems and have gained much attention in scientific research and industrial applications.The development status of the metaverses and parallel systems were reviewed, the differences and connections between them were analyzed, and their future development was prospected, which was expected to provide reference and inspiration for the development of intelligent industries, intelligent economies, and intelligent societies.