Shunsuke Saruwatari

Shunsuke Saruwatari

Auto-generated profile, based on publicly available information

Shunsuke Saruwatari

Graduate School of Information Science and Technology, The University of Osaka, Suita, Osaka, Japan

Research Focus & Metrics

Academic Career Length

7Years (2020-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
17 March 2021 in IEEE Access
Verified
Journal Article
Open access
SKSorachi Kato
Sorachi Kato
TFTakeru Fukushima
Takeru Fukushima
TMTomoki Murakami
Tomoki Murakami
HAHirantha Abeysekera
Hirantha Abeysekera
YIYusuke Iwasaki
Yusuke Iwasaki
SSShunsuke Saruwatari
Shunsuke Saruwatari
This study aims to determine the upper limit of the wireless sensing capability of acquiring physical space information. This is a challenging objective because, at present, wireless sensing studies continue to succeed in acquiring novel phenomena. Thus, although we have still not obtained a complete answer, a step is taken toward it herein. To achieve this, CSI2Image, a novel channel state information (CSI)-to-image conversion method based on generative adversarial networks (GANs), is proposed. The type of physical information acquired using wireless sensing can be estimated by checking whether the reconstructed image captures the desired physical space information. We demonstrate three types of learning methods: generator-only learning, GAN-only learning, and hybrid learning. Evaluating the performance of CSI2Image is difficult because both the clarity of the image and the presence of the desired physical space information must be evaluated. To solve this problem, we propose a quantitative evaluation methodology using an image-based object detection system. CSI2Image was implemented using IEEE 802.11ac compressed CSI, and the evaluation results show that CSI2Image successfully reconstructs images. The results demonstrate that generator-only learning is sufficient for simple wireless sensing problems; however, in complex wireless sensing problems, GANs are essential for reconstructing generalized images with more accurate physical space information.

12 November 2023
Verified
Conference Paper
Open access
AAAditya Arun
Aditya Arun
SSShunsuke Saruwatari
Shunsuke Saruwatari
SSSureel Shah
Sureel Shah
DBDinesh Bharadia
Dinesh Bharadia
Understanding the location of ultra-wideband (UWB) tag-attached objects and people in the real world is vital to enabling a smooth cyber-physical transition. However, most UWB localization systems today require multiple anchors in the environment, which can be very cumbersome to set up. In this work, we develop XRLoc, providing an accuracy of a few centimeters in many real-world scenarios. This paper will delineate the key ideas that allow us to overcome the fundamental restrictions that plague a single anchor point from localization of a device to within an error of a few centimeters. We deploy a VR chess game using everyday objects as a demo and find that our system achieves 2.4 cm median accuracy and 5.3 cm 90th percentile accuracy in dynamic scenarios, performing at least 8× better than state-of-art localization systems. Additionally, we implement a MAC protocol to furnish these locations for over 10 tags at update rates of 100 Hz, with a localization latency of ~1 ms. We have additionally open-sourced our system's codebase at https://github.com/ucsdwcsng/xrloc.git

7 May 2025 in IEEE Transactions on Wireless Communications
Verified
Journal Article
Closed access
Frodouard MinaniFrodouard Minani
Frodouard Minani
MKMakoto Kobayashi
Makoto Kobayashi
TFTakuya Fujihashi
Takuya Fujihashi
AAAbdul Alim
Abdul Alim
SSShunsuke Saruwatari
Shunsuke Saruwatari
TWTakashi Watanabe
Takashi Watanabe
Non-terrestrial networks (NTNs) contribute significantly to offering connectivity services for the upcoming sixth generation (6G). However, many services demand real-time uplink data transmission. Owing to the dynamic topology, instability of NTN channel status, resource constraints, and tremendous mobile users on ground networks; resilient and effective resource management and channel accessibility pose substantial challenges. This work investigates the uplink sum-rate maximization and latency minimization problems with a channel allocation scheme for user equipments (UEs) to high-altitude platforms (HAPs) of a particular NTN when UEs simultaneously transmit data to HAPs and HAPs forward it to low-earth orbit satellite (LEO-sat) via multiple subcarriers. We formulate a fairness resource allocation optimization problem reflecting the system utility and load of the HAP-LEO-sat link for aerial-to-space uplinks. To tackle the high-dimensional non-convex optimization problems and short visibility of LEO-sat in the learning, we advocate a novel deep reward-based multi-agent reinforcement learning (DeepRMARL) with prioritized experience replay (PER) and priority sampling approach (PSA). Furthermore, we present a long-short-term memory (LSTM) model for future environment channel prediction due to time-varying channels between UEs and HAPs. The results explicitly unveil the feasibility of our method in optimizing total uplink rate, achieving a higher fairness index, minimizing latency by approx. 12.2%, and predicting future environment states with approx. 92.8% accuracy.

9 September 2021 in IEEE Access
Verified
Journal Article
Open access
TMTakato Motoo
Takato Motoo
JKJiei Kawasaki
Jiei Kawasaki
TFTakuya Fujihashi
Takuya Fujihashi
SSShunsuke Saruwatari
Shunsuke Saruwatari
TWTakashi Watanabe
Takashi Watanabe
In online multiplayer shooting games, a long network delay can adversely impact player performance because it leads to large synchronization errors in the game information, such as location errors, between players. To mitigate synchronization errors between players, we propose a novel client-side delay compensation system for online multiplayer shooting games. The proposed delay compensation system comprises regression-based and reinforcement-based methods, which accurately predict present game information based on past game information from within each player's device. Specifically, the regression-based method utilizes linear regression to accurately predict a player's present location based on past location information, whereas the reinforcement-based method utilizes deep reinforcement learning to predict a player's present commands corresponding to the player's current location, based on past snapshots of the game fields. The advantage of the proposed system over the existing server-side delay compensation methods, e.g., Lag Compensation, is that they display accurate game information in each player's device as if the player devices were synchronized with each other; thus, any unfair advantages to the players are prevented. We developed a shooting game simulator to experimentally compare the proposed system with existing ones under constant and fluctuating network delays in loss-free and loss-prone networks. The movement models of different players were considered in the experiments. The results demonstrate that the proposed system simultaneously achieves a lower synchronization error between players and higher shooting accuracy, compared to Lag Compensation. For example, the average decrease in player shooting hit rate under a constant network delay of 100 ms is 0.55, 0.076, and 0.0039 without delay compensation and with the proposed regression-based and reinforcement-based methods, respectively. Under the same network delay, the proposed regression-based and reinforcement-based methods show the improvement of 16 pixels and 14 pixels in terms of the synchronization error compared to Lag Compensation, respectively.

26 April 2021 in IEEE Open Journal of the Computer Society
Verified
Journal Article
Open access
SYShunpei Yamaguchi
Shunpei Yamaguchi
RORitsuko Oshima
Ritsuko Oshima
JOJun Oshima
Jun Oshima
RSRyota Shiina
Ryota Shiina
TFTakuya Fujihashi
Takuya Fujihashi
TWTakashi Watanabe
Takashi Watanabe
Human collaboration has a great impact on the performance of multi-person activities. The analysis of speaker information and speech timing can be used to extract human collaboration data in detail. Some studies have extracted human collaboration data by identifying a speaker with business-card-type sensors. However, it is difficult to realize speaker identification for business-card-type sensors at low cost and high accuracy because of spikes in the measured sound pressure data, ambient noise in the non-speaker sensor, and synchronization errors across each sensor. This study proposes a novel sound pressure sensor and speaker identification algorithm to realize speaker identification for business-card-type sensors. The sensor extracts the user's speech at low cost and high accuracy by employing a peak hold circuit and time synchronization module for spike mitigation and precise time synchronization. The algorithm identifies a speaker with high accuracy by removing ambient noise. The evaluations show that the algorithm accurately identifies a speaker in a multi-person activity considering varying numbers of users, environmental noises, and reverberation conditions as well as long or short utterances. In addition, the peak hold circuit enables accurate extraction of speech and the synchronization error between the sensors is always within ±30 μs, that is, negligible error.

16 November 2022 in IEEE Access
Verified
Journal Article
Open access
SYShunpei Yamaguchi
Shunpei Yamaguchi
MNMotoki Nagano
Motoki Nagano
SOShunpei Ohira
Shunpei Ohira
RORitsuko Oshima
Ritsuko Oshima
JOJun Oshima
Jun Oshima
TWTakashi Watanabe
Takashi Watanabe
Collaborative learning is an educational approach to teaching and learning that involves groups of learners collaborating to solve a problem, complete a task, or create a product. To enhance the performance of collaborative learning, the studies in [1]–[4] develop an IoT system and quantitatively extract collaboration between learners. The studies acquire sensor data from IoT badges on learners and analyze learning activities with the acquired sensor data on a computer. However, existing studies are not user-friendly for learning analysts who are unfamiliar with information technology owing to complex software installation and command line interface (CLI) operation. Such drawbacks hinder the wide expansion of technology and the exploration of new learning patterns in learning science. Considering high usability for analysts, this paper proposes novel web services named Sensor-based Regulation Profiler Web Services (SRP Web Services) for collaboration analysis with IoT badges. The proposed web application consists of front-end on Next.js and back-end on FastAPI, SQLite, and Python and extracts key points in learning activities for the analysts from the acquired sensor data on a web browser. Experimental evaluations showed that the proposed web services support learning analysts in quantitative analysis of learning activities with high usability. In addition, SRP Web Services are scalable with hundreds of users.