My Research at AIST



	Search Dictionary Traffic Map Weather


\| Theme \| Demo \| Publication \| Etc \| Japanese \|

| Theme of My Research

Evacuation simulator based on pedestrian flow measurement

Recently, the simulation of pedestrians in a crowd environment can be carried out by modeling the pedestrian flows by computer, because of the improvement of computer performance and algorithm. Such simulation technology has come to be used for the prediction of pedestrian flow in a large-scale event, safe verification of a building, and so on. In this research, evacuation action is modeled by measuring the trajectories of about 600 persons in a fire drill, and the simulator reproducing evacuation action is developed. Furthermore, the method of supporting evacuation guidance using the developed evacuation simulator is clarified.

Integration of medicine, engineering and sociology in emergency care

Since the patient who requires lifesaving treatment is losing consciousness in many cases, the usual need for explanation and consent has been disregarded. It considers that medical treatment is a process for the conviction between the parties based on information sharing, and this medical treatment is expressed as the sharing of medical treatment. Achievement of this medical treatment to share is needed in emergency care. In this research, by integration of the image analysis by information engineering, and the communication analysis by sociology, we aim to present a new fundamental model for explanation and consent at the lifesaving spot.

Pedestrian Trajectories Analysis in a Commercial Establishment using a Human Tracking

Seven Ubiquitous Stereo Vision(USV)s are installed in actual commercial establishments, pedestrians are tracked using the USVs, and the pedestrian trajectories are recorded during business hours. In this research, we proposed some methods for analyzing pedestrian trajectories. In detail, we proposed the two methods; 1) Comparing method intelligibly about how pedestrian trajectories changed when there are advertisements and events on the outskirts, 2) Visualization method for the pedestrian trajectories over a long period of time. These results from our methods will surely provide a new service about marketing for the administrators of institutions.

Human Tracking in Natural Environment using Stereo Camera

We are developing the Ubiquitous Stereo Vision (USV) which can calculate a disparity image also in a dark environment like the evening or night. We are especially doing research which extracts humans using this USV. The actual proof experiments are conducted in not only the limited environment such as a conventional laboratory but also more near actual environment such as a crossing, and a office complex, and so on. Since three-dimensional information is restored from the disparity image, not only the position of the human but also the height and the physique are measurable from this USV. Furthermore, a man-or-woman distinction, a family group, and so on may be able to be estimated. Probably, the USV is offered as a powerful sensor for human tracking.

| Journal and Conference Paper

Masahiro Nomura, Shuhei Watanabe, Yoshihiko Ozaki, Masaki Onishi,
"Warm Starting Method for CMA-ES, "
Workshop on Meta-Learning (MetaLearn@NeurIPS2019), Dec. 2019.
[pdf]
Shuhei Watanabe, Yoshihiko Ozaki, Yoshiaki Bando, Masaki Onishi,
"Speeding up of the Nelder-Mead Method by Data-driven Speculative Execution, "
Asian Conference on Pattern Recognition (ACPR), Nov. 2019.
Takumi Saito, Masaki Onishi, Ikushi Yoda, Satomi Kuroshima, Michie Kawashima,
Koutaro Uchida, Jun Oda, Shiro Mishima, Tetsuo Yukioka,
"Analysis of Team Medical Care Using Integrated Information from the Trajectories of and Conversations Among Medical Personnel,"
Precision Health and Medicine (W3PHAI2019), pp.159-168, 2019.
Yoshihiko Ozaki, Masaki Onishi,
"Practical Deep Neural Network Performance Prediction for Hyperparameter Optimization,"
KDD Workshop Automation in Machine Learning (AutoML), Aug. 2019.
Yoshihiko Ozaki, Shuhei Watanabe, Masaki Onishi,
"Accelerating the Nelder - Mead Method with Predictive Parallel Evaluation,"
ICML Workshop on Automated Machine Learning (AutoML), June 2019.
[pdf]
Shusuke Shigenaka, Shunki Takami, Yoshihiko Ozaki, Masaki Onishi, Tomohisa Yamashita, Itsuki Noda,
"Evaluation of Optimization for Pedestrian Route Guidance in Real-world Crowded Scene,"
International Conference on Autonomous Agents and MultiAgent Systems(AAMAS2019), pp.2192-2194, May 2019.
[pdf]
Shusuke Shigenaka, Shunki Takami, Masaki Onishi, Tomohisa Yamashita, Itsuki Noda,
"Estimating Pedestrian Flow in Crowded Situations with Data Assimilation,"
10th International Workshop on Optimization in Multiagent Systems (OptMAS2019), May 2019.
[pdf]
Ryo Nishida, Masaki Onishi, Koichi Hashimoto,
"Construction of a Route Choice Model for Application to a Pedestrian Flow Simulation,"
International Workshop on Behavior analysis and Recognition for knowledge Discovery (BiRD2019), March 2019.
in conjunction with the IEEE International Conference on Pervasive Computing and Communications (PerCom2019)
Shunki Takami, Masaki Onishi, Kazunori Iwata, Nobuhiro Ito, Yohsuke Murase, Takeshi Uchitane,
"An Environment for Combinatorial Experiments in a Multi-agent Simulation for Disaster Response,"
Principles and Practice of Multi-Agent Systems (PRIMA2018), pp.646-654, Nov. 2018.
Takuya Okano, Masaki Onishi, Itsuki Noda,
"A Parameter Investigation of the ?-greedy Exploration Ratio Adaptation Method in Multi-agent Reinforcement Learning,"
European Workshop on Reinforcement Learning, Oct. 2018.
Ryo Nishida, Masaki Onishi, Koichi Hashimoto,
"Multiple Destinations Pedestrian Model using an Improved Social Force Model, "
10th International Workshop on Agents in Traffic and Transportation (ATT2018), July 2018.
(held in conjunction with ECAI/IJCAI, AAMAS and ICML conferences).
Qiong Chang, Masaki Onishi, Tsutomu Maruyama,
"Fast Convolution Kernels on Pascal GPU with High Memory Efficiency,"
High Performance Computing Symposium (HPC), SpringSim, pp.375-386, April 2018.
Best Paper Award
Yoshihiko Ozaki, Masaki Yano, Masaki Onishi,
"Effective hyperparameter optimization using Nelder-Mead method in deep learning, "
IPSJ Transactions on Computer Vision and Applications, (2017) 9:20.
[Link]
Takao Arai, Shoichi Ohta, Masaki Onishi, Miyu Taniguchi, Junya Tsurukiri, Kenichiro Kumasaka, Katsuhiro Nagata, Kensuke Suzuki, Ken Harigae, Tetsuo Yukioka,
"Empirical research on the utility of a preparation manual for a disaster medical response drill,"
Open Access Emergency Medicine, vol. 2017:9, pp.81-88, 2017.
[Link]
Yuri Miyagi, Masaki Onishi, Chiemi Watanabe, Takayuki Itoh, Masahiro Takatsuka,
"Classification and Visualization for Symbolic People Flow Data, "
Journal of Visual Languages and Computing, vol.43, pp.91-102, Dec. 2017.
[Link]
Masaki Onishi, Shinnosuke Nakashima,
"Mutual Interaction Model between the Number of People in Real Space and the Number of Tweets in Virtual Space"
Proceedings of International Conference on Pattern Recognition(ICPR), TuPT4.6, pp.2074-2079, Dec. 2016.
Yuri Miyagi, Masaki Onishi, Chiemi Watanabe, Takayuki Itoh, Masahiro Takatsuka,
"Feature Extraction and Visualization for Symbolic People Flow Data, "
International Conference Information Visualisation (IV2016), July 2016.
The Best Paper Award
Shoichi Ohta, Munekazu Takeda, Ryo Sasaki, Hirotaka Uesugi, Hironobu Kamagata, Kentaro Kawai, Satomi Kuroshima, Michie Kawashima, Masaki Onishi, and Ikushi Yoda,
"Sustainable Training-Model Development Based on Analysis of Disaster Medicine Training,"
Journal of Disaster Research, vol.10, no.5, pp.900-918, Oct. 2015.
Masaki Onishi,
"[Invited Paper] Analysis and Visualization of Large-Scale Pedestrian Flow in Normal and Disaster Situations,"
ITE Transactions on Media Technology and Applications, vol. 3, no. 3, pp.170-183, July 2015.
[pdf]
Ryosuke Ichikari, Masaki Onishi, Takeshi Kurata,
"MR fitting simulation with virtual camera motions for wheelchair users, "
International Conference on Artificial Reality and Telexistence (ICAT2013), Dec. 2013.
Hirokatsu Kataoka, Yoshimitsu Aoki, Kenji Iwata, Yutaka Satoh, Ikushi Yoda, Masaki Onishi,
"Big Trajectory Data Analysis for Clustering and Anomaly Detection,"
IAPR International Conference on Machine Vision Applications (MVA2013), pp.451-454, May 2013.
Yosuke NONAKA, Masaki ONISHI, Tomohisa YAMASHITA, Takashi OKADA, Atsushi SHIMADA and Rin-ichiro TANIGUCHI,
"Effective Walking Velocity Modeling for Pedestrian Simulator,"
International Conference on Quality Control by Artificial Vision (QCAV2013), pp.131-135, May 2013.
Fukute Aya, Onishi Masaki, Itoh Takayuki,
"A Linked Visualization of Trajectory and Flow Quantity to Support Analysis of People Flow, "
17th International Conference on Information Visualization (iV2013), pp.561-567, July 2013.
Y. Kobayashi, T. Okamoto and M. Onishi,
"Generation of Obstacle Avoidance Based on Image Features and Embodiment, "
International Journal of Robotics and Automation, vol.27, no 4, pp.364-376, 2012.
Yuichi Kobayashi, Masaki Onishi, Shigeyuki Hosoe and Zhiwei Luo,
"Multi-tasking Arbitration and Behaviour Design for Human Interactive Robots,"
International Journal of Systems Science, pp.1-17, Jan. 2012.
Tomohisa Yamashita, Shunsuke Soeda, Masaki Onishi, Itsuki Noda,
"Exhaustive Testing of Evacuation Plan with High-Speed Evacuation Simulator,"
International Scientific and Technical Conference Emergency Evacuation of People from Buildings, pp.357-364, March 2011.
Takafumi Yamamoto, Yoji Yamada, Masaki Onishi, Yoshihiro Nakabo,
"A 2D Safety Vision System for Human-Robot Collaborative Work Environments Based upon The Safety Preservation Design Policy,"
IEEE International Conference on Robotics and Biomimetics (IEEE-ROBIO 2011), pp.2049-2054, Dec. 2011.
Masaki Onishi, Ikushi Yoda,
"Visualization of Customer Flow in an Office Complex over a Long Period,"
Proceedings of International Conference on Pattern Recognition (ICPR), pp.1747-1750, Aug. 2010.
Yoji Yamada, Yoshiki Takami, Yoshihiro Nakabo, Masaki Onishi,
"A Safety Vision System Preserving Safety of Human-Robot Collaborative Work Environments by Continuous Detection of Passive Markings,"
Proceedings of International Conference Safety of Industrial Automated Systems (SIAS), Session A5, June 2010.
Taichi Okamoto, Yuichi Kobayashi and Masaki Onishi,
"Acquisition of Body and Object Representation Based on Motion Learning and Planning Framework,"
Proceedings of the 9th International Conference on Intelligent Systems Design and Applications, pp.737-742, 2009.
Keiko Tsuchiya, Michie Kawashima, Tetsuo Yukioka, Shoichi Ota, Shiro Mishima, Jun Oda, Chikako Kawahara, Masaki Onishi, Ikushi Yoda,
"Information Sharing in a Pre-hospital Setting - A time related corpus based approach,"
Proceedings of the 14th Conference of Pan-Pacific Association of Applied Linguistics (PAAL), pp.269-274, 2009.

Top Page, Research at RIKEN, BMC, Research at Osaka Pref. Univ.