Toshihiro Maki

Toshihiro Maki

Auto-generated profile, based on publicly available information

Toshihiro Maki

Institute of Industrial Science, The University of Tokyo, Tokyo, Japan

Research Focus & Metrics

Academic Career Length

21Years (2006-Present)

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

Representative Publications

View All Publications
19 July 2022 in IEEE Journal of Oceanic Engineering
Verified
Journal Article
Closed access
YWYang Weng
Yang Weng
JPJoni Pajarinen
Joni Pajarinen
RARiad Akrour
Riad Akrour
TMTakumi Matsuda
Takumi Matsuda
Jan PetersJan Peters
Jan Peters
TMToshihiro Maki
Toshihiro Maki
With the developments in underwater wireless optical communication (UWOC) technology, UWOC can be used in conjunction with autonomous underwater vehicles (AUVs) for high-speed data sharing among the vehicle formation during underwater exploration. A beam alignment problem arises during communication due to the transmission range, external disturbances and noise, and uncertainties in the AUV dynamic model. In this article, we propose an acoustic navigation method to guide the alignment process without requiring beam directors, light intensity sensors, and/or scanning algorithms as used in previous research. The AUVs need stably maintain a specific relative position and orientation for establishing an optical link. We model the alignment problem as a partially observable Markov decision process (POMDP) that takes manipulation, navigation, and energy consumption of underwater vehicles into account. However, finding an efficient policy for the POMDP under high partial observability and environmental variability is challenging. Therefore, for successful policy optimization, we utilize the soft actor–critic reinforcement learning algorithm together with AUV-specific belief updates and reward shaping based curriculum learning. Our approach outperformed baseline approaches in a simulation environment and successfully performed the beam alignment process from one AUV to another on the real AUV Tri-TON 2.

28 February 2020 in International Journal of Control, Automation, and Systems
Verified
Journal Article
Closed access
PDF file
TMToshihiro Maki
Toshihiro Maki
HHHiroumi Horimoto
Hiroumi Horimoto
TITakashi Ishihara
Takashi Ishihara
KKKazuya Kofuji
Kazuya Kofuji

1 June 2007
Verified
Conference Paper
Closed access
TMToshihiro Maki
Toshihiro Maki
HKHayato Kondo
Hayato Kondo
TUTamaki Ura
Tamaki Ura
TSTakashi Sakamaki
Takashi Sakamaki
Real-time accurate positioning is a key technology for autonomous underwater vehicles to perform close-range seafloor observations such as photo mosaicing. This paper proposes a real-time positioning method that realizes robust and drift-free positioning in a local area based on a passive acoustic landmarks set in the environment and a profiling sonar (profiler) mounted on the vehicle. The method stochastically updates the vehicle's position based on all the sensory data available. Although the position of landmarks is generally unknown, this paper assumes the position is known without error for simplification. As an observation model of the profiler we propose a radial-angular representation with a non-Gaussian distribution based on experimental data. The performance of the proposed method is verified through tank experiments using the AUV Tri-Dog 1. The effects of the number of landmarks and observation model on the positioning accuracy is discussed. Through comparison with ground truth and off-line estimations based on actual data obtained during the experiments, the positioning accuracy of the proposed method remains within 5 centimeters at all times during the 2 hour duration of the experiment with a traveled distance of 600 meters, with two landmarks at a distance of 10 meters.

9 February 2024 in IEEE Robotics and Automation Letters
Verified
Journal Article
Closed access
SCSehwa Chun
Sehwa Chun
CKChihaya Kawamura
Chihaya Kawamura
KOKenji Ohkuma
Kenji Ohkuma
TMToshihiro Maki
Toshihiro Maki
Biologging has been the prevalent method for studying the ecology and behavior of aquatic life. However, it has problems of operational inconvenience and stress on organisms. Recently, research towards remote tracking using Autonomous Underwater Vehicles (AUVs) has been gaining attention, particularly those equipped with multibeam imaging sonar (MBS) less affected by underwater light and turbidity. However, existing studies have focused mainly on 2D detection and tracking due to the inherent limitations of MBS. In this letter, we propose a novel method for 3D detection and tracking of a moving target for marine life surveys, utilizing the tilt control of MBS on AUVs. YOLOv3 is employed to identify the target's position in sonar imagery. Incorporation of the particle filter algorithm enhances position an estimation accuracy and enables a estimation of the target's state, velocity and orientation as well. 3D tracking method is introduced based on the estimated states. This method has been validated through tank experiments with a hovering-type AUV and a life-sized sea turtle replica, successfully achieving 3D tracking of the target moving horizontally and vertically. The approach demonstrated a position estimation with an RMSE of 0.23 m and velocity and orientation within the error of 0.022 m/s and 16.59 degrees, respectively.

4 March 2019 in IEEE Robotics and Automation Letters
Verified
Journal Article
Closed access
TMTakumi Matsuda
Takumi Matsuda
TMToshihiro Maki
Toshihiro Maki
TSTakashi Sakamaki
Takashi Sakamaki
This paper reports the sea experiments of a navigation method for multiple autonomous underwater vehicles (AUVs) that relies on a single high-performance AUV, which can estimate its self-state (position and pose) accurately without any support. In the method, there is a parent AUV and child AUVs. While the latter have minimal navigational sensors, such as a microelectromechanical gyro, a single-axis fiber optic gyro and cameras, the former can estimate its self-state without any support using high-grade navigational sensors, such as an inertial navigation system and a Doppler velocity log. As all the AUVs have acoustic positioning and communication devices, the child AUVs can also estimate their self-state as accurately as the parent AUV thanks to relative positioning with the parent AUV. We performed the sea experiments in a hydrothermal vent field using AUVs. Post-processing state estimation using sensor data from the experiments showed that the proposed method can realize longdistance navigation while maintaining an accuracy more than 5 times that for navigation independently performed by a child AUV.

1 September 2007
Verified
Conference Paper
Closed access
TMToshihiro Maki
Toshihiro Maki
HMHayato Mizushima
Hayato Mizushima
HKHayato Kondo
Hayato Kondo
TUTamaki Ura
Tamaki Ura
TSTakashi Sakamaki
Takashi Sakamaki
MYMasao Yanagisawa
Masao Yanagisawa
Although underwater vent fields are of great scientific interest, accurate visual mapping is difficult because of the presence of bubble plumes that degrade the accuracy of conventional acoustic positioning systems such as long base line (LBL) and super short base line (SSBL). The authors had proposed a visual mapping method of shallow vent fields with an autonomous underwater vehicle (AUV) equipped with a profiling sonar, where positioning is based on vertical rod-shaped acoustic reflectors and bubble plumes. Although performance was verified through a series of experiments, there remain two challenges as follows. Firstly, observation is terminated if the vehicle was surrounded by bubble plumes, since the vehicle tries to avoid collision with not only artificial reflectors but also collision-safe bubble plumes. Secondly, the observation area drifts by disturbance since the waypoints are defined relative to the vehicle's position after descending close to seafloor. This paper proposes a real-time path planning method of an AUV as a part of the proposed observation method. The path of the vehicle is defined based on the types of landmarks as well as the geometric relationship between the vehicle and the landmarks. The vehicle can distinguish landmark types using a sheet laser and a camera. The proposed method was implemented on the AUV Tri-Dog 1 and a series of experiments were carried out in order to verify its performance.