34254 research outputs found
Sort by
Correction of Nonlinearity and Hybrid Waveforms for High Sensitivity Frequency Modulated Continuous Wave 4D LiDAR
No full textThis paper employs a 4D LiDAR system coupled with Frequency Modulated Continuous Wave (FMCW) technology to measure the distance and velocity of target objects. This is achieved by comparing the frequency differences between transmission and return signals while also utilizing the rate of frequency change to calculate the target object's speed and distance.
The entire 4D LiDAR system is optimized in this paper to address issues such as low SNR in the beat frequency spectrum, noise, insufficient resolution, and sensitivity problems. These optimizations include correcting non-linearity, generating hybrid waveforms, using self-injection locked oscillators, and avalanche photodiodes.
As DFB LASER may introduce non-linearity during modulation, the paper proposes a method to correct these issues using self-heterodyne and heterodyne techniques to measure the frequency of the modulated laser. It also controls the generation of predistorted aveforms through iterative learning to correct non-linearity.
This paper further improves upon traditional Pure FMCW technology, which is limited by inadequate resolution for measuring low-speed velocity images. The system is made capable of simultaneously measuring low speeds and distances through the use of hybrid waveforms. Additionally, self-injection locked oscillators and avalanche photodiodes are introduced to optimize the entire system. Imaging experiments are conducted using a mechanical optical scanner, scanning 40x40 pixels to measure the distance and velocity of objects.
The analysis results indicate that introducing APD (Avalanche Photodiode) and selfinjection oscillators can reduce the average error rate by up to 50% when compared to traditional PIN-type photodiodes, even at speeds as low as 0.005 mm/s. To simplify the system for productization, the injection system is removed, and hybrid waveforms are used along with the Chirp LASER as the light source. While this results in an increase in speed error of approximately 20%, it is better suited for practical product applications
Research on Patent Application and Utilization in Higher Education Institutions: Taking National Sun Yat-sen University for example
No full textUniversities possess significant research and development capabilities, drawing attention from various sectors. Converting these capabilities into tangible industries to further foster national scientific research and development is a crucial contemporary challenge. In the past, universities were perceived as ivory towers detached from real-life applications. However, with the changing times and advancements in information technology, their role has gradually evolved from purely academic to research-oriented, becoming essential hubs for scientific and innovative endeavors.
In light of this trend, we passed the "Science and Technology Fundamental Law" in 1999, aiming to elevate scientific and technological standards, drive economic development, and enhance national competitiveness. Over two decades have passed since the enactment of this law, prompting the present study to explore its efficacy. Focusing on National Sun Yat-sen University as a case study, this empirical research investigates whether the Science and Technology Fundamental Law has indeed encouraged faculty members to apply for patents and how the university has employed the management theory of "open innovation" to effectively govern these patents, fostering value creation.
I begin by conducting a literature review on how universities apply open innovation and the significance of patents. Subsequently, they employ descriptive statistical analysis to examine the research funding, patent applications, and maintenance status of 1,013 faculty members at National Sun Yat-sen University. To further explore details not apparent from descriptive statistics, the researchers conduct interviews with three professors. Through this research process, the effectiveness of the overall system and related legislation is evaluated, and corresponding management mechanisms and strategic recommendations are provided to facilitate the application of open innovation in both academic and industrial domains
Design and Implementation of Latency-Reduced Reconfigurable Hardware Chip Architecture of Successive Cancellation List (SCL) Decoder
No full textPolar Code is currently the only coding method that can achieve Shannon's Limit. In addition to this, Polar Code has been identified as the coding method for the control channels of Enhanced Mobile Broadband (eMBB) in 5G, highlighting its critical role in communication technology development. To adapt to different channel environments, utilizing a single mode and a single code length (N) with a fixed number of survival path (L) cannot meet the requirements. Therefore, we utilize a reconfigurable and double-column architecture to design decoder that support two types of survival path (L) for Successive Cancellation List (SCL) decoding. In this system, with a code length of 1024 (N=1024), support is provided for 2 and 4 survival paths (L=2,4). Based on the characteristics of channel polarization, this approach significantly reduces the latency of the Successive Cancellation List (SCL) decoder.
The system developed three main technologies: Multi-bit Decoding Leaf Nodes (MDLN) based on the Path Selecting Strategy (PSS), and High-speed Sorter (HSS), and optimized high-performance Partial-sum Units (PSU). The chip, implemented using TSMC-40nm CMOS (Multi-Vt) process, occupies 0.6592mm^2 core area, achieving an operating frequency of 476.19MHz. The strategy of reducing splitting occurrences in MDLN design, along with a highly flexible hardware architecture, significantly reduced clock cycles, presenting a robust design approach for academia and industry
A Multi-Agent Reinforcement Learning Architecture based on Proxy Rewards Distributed by a Pivot Observer
No full textThis paper proposes a architecture, referred to as "Multi-Agent Reinforcement Learning based on Pivot Observer (MAPO)." The pivot observer learns to assign the importance of each agent and facilitates independent learning of individual strategies for multiple agents. Finally, the strategies of all agents are combined to form a final strategy. This paper also introduces the concept of the pivot observer assigning proxy rewards, which are defined based on cosine similarity of state transitions and can be used to address sparse reward tasks.
In terms of model training, the paper presents a mechanism that separates the training of the pivot observer and agents, with distinct loss functions and update frequencies. Within both of the pivot observer and agents, attention mechanisms and recurrent neural networks are combined to allow the model to make decisions based on past experiences while focusing on regions that are particularly important for decision-making. Additionally, a parameter mask based on the Variational Information Bottleneck theory is introduced in the pivot observer, to prune its policy network's parameters as well as reduce parameters and computational complexity of the model.
Experimental evaluations of this paper are conducted on the Atari game tasks, comparing the results with a classic hierarchical reinforcement learning framework commonly used for these tasks. The experimental results ultimately demonstrate superior performance of the model proposed in this paper over the hierarchical reinforcement learning framework
A Hierarchical Reinforcement Learning Architecture with Distilled High-Level Observation
No full textThis paper is based on the concept of hierarchical reinforcement learning, which effectively decomposes a challenging task into high and low levels for collaborative processing. The high level primarily focuses on acquiring goals and directions over longer time intervals, while the low level interacts with the environment based on the goals provided by the high level. These two levels observe the environment from different perspectives, aiming to derive optimal strategies through their distinct viewpoints.
This paper extends research on hierarchical reinforcement learning by introducing a Recurrent Neural Network (RNN) architecture with an attention mechanism, referred to as the Recurrent Attention Mechanism (RAM). This architecture enables the network to concentrate on the most critical aspects while considering both past and future context. To ensure the preservation of key information during extended message transmission, the paper proposes the Distilled Recurrent Attention Mechanism (DRAM). DRAM incorporates the bottleneck theory within the high-level framework, allowing the filtration of information to retain essential elements. These filtered, vital feature vectors are then passed to the RAM, facilitating the retention of crucial information during state transitions over extended time intervals.
After integrating DRAM into the hierarchical reinforcement learning framework, the paper introduces the Hierarchical Reinforcement Learning Architecture with Distilled High-Level Observation (HDO). Finally, the proposed HDO architecture is validated using Atari games and compared to other well-known hierarchical reinforcement learning approaches, demonstrating that the HDO framework introduced in this paper can better handle challenging tasks
Examining the Practice of University Social Responsibility from a Public-Private Collaboration Perspective\uef\ubcA Case Study of the Revitalization of Qianzhen Second Public Market
No full textWith the changing times and economic environment, traditional markets have gradually lost their prosperity and vitality. A variety of factors threaten the survival of traditional markets, including changes in consumer shopping habits, the rise of large shopping centers, competition from online shopping, and the aging and inadequate maintenance of market facilities. The decline of traditional markets not only impacts the local economy but also the community. Integral to people's daily lives, the decline of traditional markets has the potential to weaken community cohesion, possibly causing local residents to lose their original social spaces and putting the local culture at risk of fading into obscurity or being forgotten. How to promote the revitalization of traditional markets to achieve sustainable development has become an important issue for the public sector and local community.
This study examines the case of the Second Public Market in Qianzhen, applying the collaborative governance model of Ansell and Gash and the perspective of University Social Responsibility. The research delves into the establishment of trust, challenges, and difficulties in collaborative cooperation. Through qualitative research methods, Participant Observation, and In-Depth Interviews, 14 participants from Qiancao team of National Sun Yat-sen University, the public sector, and community organizations were invited to explore how Public-Private collaborative governance promotes market regeneration and the practical implementation of University Social Responsibility. The study found that the significance of the Qiancao team in the operation and interaction of Public-Private collaboration, highlighting its importance in governance and market revitalization. Recommendations are provided for challenges encountered in the collaborative governance process, emphasizing communication, feedback among stakeholders, information sharing, and transparency. These insights aim to serve as valuable references for future research and practical policy
Population dynamics and genetic diversity of black sea cucumber (Holothuria atra) in Duzaiping, SiaoLiouciou, southern Taiwan
No full textFrom August 2021 to December 2022, I investigated the size, distribution, density, fission, and regeneration of the Holothuria atra in the intertidal zone of Duzaiping, SiaoLiouciou, Pingtung County, Taiwan. I also analyzed its genetic diversity and compared it with another fission population of Haikou, Pingtung County, Taiwan, by using the sequences of mitochondrial DNA COI genes. H. atra in Duzaiping was mainly composed of small individuals less than 80 grams, accounting for 99.1% (n=1200) of the population, the densest and smallest H. atra population in the literature. The unusually high density and small size of H. atra in Duzaiping are due to fission and an enclosed, limited habitat and food resource of the intertidal zone. The average size was smaller (16.88-20.14 g) in late fall and winter (November - January) and larger (39.25-45.91 g) in summer (June - August); the average weight in winter was half of that in summer. Newly fission individuals were found monthly, suggesting that fission could occur throughout the year but peaks in late fall to early winter (November to January). No small individual of sexual recruits was found throughout the year, suggesting that the population increased mainly by fission. In July and August 2021, after 15 days of heavy rainfall, mass mortality was recorded. The density decreased from 5.6 \uc2\ub1 1.42 m-2 (May 2020) to 2.66 \uc2\ub1 1.34 m-2 (September 2021). The mortality is 52.5%. Density recovers to 5.12 \uc2\ub1 1.34 m-2 after ten months. Reduced salinity (15 \ue2\ub0) due to heavy rainfall and high temperatures (37-39 \uc2\ub0C) during the ebb tides were the leading causes. The size of H. atra was larger in the high-tide zone than in the low-tide zone, which may be due to the higher tolerance of the large individuals to the challenging environment in the high-tide zone. The density of H. atra was higher in the middle and low tide zones, probably because the environment was more stable than that in the high tide zone. Regeneration in the laboratory is that the anterior ends can feed in 18th-20th days, but the initial food is organic matter. On the 35th-41th day, the anterior ends excrete fine sand. The posterior ends regenerate small tentacles and feed on the 45th - 47th day. Sixty individuals from two populations were collected and analyzed for 60 COI sequences (646 bp). The number of haplotypes was 1, the haplotype diversity was 0, and the nucleotide diversity was 0 from the Duzaiping population. The number of haplotypes was 3, the haplotype diversity was 0.34253, and the nucleotide diversity was 0.00523 from the Haikou population. H. atra populations in the two sites were highly differentiated (\uce\ua6st =0.70027). We infer the low haplotype diversity and low nucleotide diversity in both sites resulted from fission (asexual reproduction) and post-settlement mortality
The Practical of Rukai Ancient Ballads in Ethnic Education\uef\ubcA Case Study of the Qingye Tribe
No full textThe music of the Rukai tribe is a significant cultural asset, with songs playing a crucial role in the lives of the Rukai people. Regardless of age, the Rukai people have a passion for singing, demonstrating skilled vocal techniques. Rukai songs cover a wide range of themes, including life rituals, narratives, games, labor, entertainment, and emotional expressions. In Rukai life, singing is not only an indispensable form of basic wisdom and knowledge but also an integral part of everyday life for everyone. Through singing, the nuances of life are transformed into the rhythm of music, adding charm to every moment.
In Rukai culture, singing holds a vital role in various occasions such as work, leisure, entertainment, sorrow, love, or war. The vocabulary of Rukai songs often uses metaphors from nature, animals, and plants, spreading like melodies through the mountains while recounting the history of the Rukai tribe. To this day, the Rukai people preserve the singing techniques of their ancestors, passing down ancient lyrics and expressions, creating the unique characteristics of Rukai songs.
Currently in Taiwan, many indigenous tribes, indigenous scholars, and educational institutions in indigenous schools actively encourage cultural revitalization efforts. I am involved in AUBA Elementary School in Pingtung County, which has implemented ethnic experimental education based on Rukai songs, extracting cultural elements from songs through field research and constructing teaching curricula.
The goal of AUBA Elementary School is to incorporate "Rukai songs" into the experimental education curriculum. Each Rukai song learned by the children will be dissected, exploring the cultural significance of ancient lyrics. Teachers will then compile these insights into a curriculum, gradually reconnecting with the lost cultural depth of modern Rukai society.
The design of this experimental curriculum helps students to deeply understand and experience Rukai culture. Through the emotions, values, and lifestyles embedded in the songs, children can connect with their cultural roots, building confidence and self-esteem in the learning process. The outcomes of this project will not only impact school education but also have a profound effect on the inheritance and revival of Rukai culture.
By studying my father as a research subject, I aim to document some traditional Rukai songs. This effort is not only to preserve my father's unique voice and singing style but also to trace the footsteps of past lives through the lyrics. I conducted a series of activities, including collection, organization, and recording, ultimately producing a complete album.
This album serves as a tribute to my father and a precious preservation of traditional Rukai songs. I hope this album becomes a valuable resource for scholars engaged in indigenous music research or education, as well as for Rukai elementary schools in Pingtung and Taitung counties, Rukai churches, and indigenous artists (singers).
Through this initiative, I hope to share and promote Rukai music and culture. The album will convey our traditional values and way of life, allowing more people to understand and appreciate the cultural heritage of the Rukai tribe. This is also our commitment to the future, to continue passing on our culture, allowing it to continue flourishing in modern society
Multi-Keyword Searchable Identity-Based Proxy Re-Encryption from Lattices
No full textWith the advancement of technology, more and more people are sharing data through cloud servers. To protect data privacy, encryption before uploading has become a popular solution. While encryption ensures privacy, searching for target data between ciphertexts takes effort. Therefore, searchable encryption has become an important research topic. On the other hand, since the advancement of quantum computers will lead to the crisis of cracking traditional encryption algorithms, it is necessary to design encryption schemes that can resist quantum attacks. In view of this, to improve the security and performance of cloud servers, we propose a multi-keyword searchable identity-based proxy re-encryption scheme from lattices. The proposed scheme applies the following key cryptographic technologies to enhance encryption performance. First and most important, the proposed scheme provides multi-keyword searches. It is the first lattice-based encryption scheme that can support both AND and OR operations, improving search flexibility. Second, proxy re-encryption is applied to allow proxy servers to reduce clients' computation and transmission costs. Third, the proposed scheme employs identity encryption to reduce the computation and transmission costs caused by certificates. Moreover, the access structure of the proposed scheme is not based on a Linear Secret Sharing Scheme (LSSS), thereby avoiding errors in decryption or search results caused by the structure of LSSS. Finally, we make a formal security proof of the proposed scheme under the D-LWE assumption
Social media as a communication tool for grassroots lobbying in Austria
No full textOver the last decade, we have witnessed an increased professionalization of social media practices by political parties and organizations in Austria and around the world. Advocacy groups and political actors increasingly use social media to achieve their political goals. One of those methods used, is grassroots lobbying.
This paper aims to summarize the current scientific state of affaire and putting it into an Austrian perspective, by examining existing literature and quantitively analyzing the Facebook, X (Twitter) and Instagram accounts of all 183 Austrian National Council Members.
While social media is ideal for mobilizing the public in Austria (and around the world), due to its interactive features, fostering feedback loops and real-time communication ability, it is less useful for communication with Austrian politicians. Even though Austrian parliament members attribute great importance to social media in their political work, they are not using its interactivity to their fullest extent. The biggest motives for using social media for Austrian politicians are self-presentation, fast and uncomplicated information sharing and communication, and reaching out to (new) target audiences.
Despite the widespread adoption of Facebook and Instagram by Austrian parliamentarians, their utilization for interaction and political engagement remains limited. In contrast, X (Twitter) is actively employed for interaction, yet its prevalence among members of the Austrian National Council is not substantial. Furthermore, younger Members of Parliament do not demonstrate a significantly heightened level of interactive engagement on social media.
Generally, individuals from smaller, newer, and left-leaning parties engage more actively on social media. On the contrary, members of the right leaning FP\uc3, despite being the first party in Austria to harness the potential of social media and maintaining popularity on various platforms (excluding Twitter), exhibit minimal interaction with their constituents