Research Topics
Now that ubiquitous society has come, computerized/intelligent control systems are achievable everywhere. Our laboratory is working on system control, machine learning, signal processing, and interdisciplinary study based on these topics. In what follows, after stating general overview, we show our current research topics. Please note that these are just examples and we always challenge a new topic.
— Intelligent System Control: What is our target —
As the term “control” is used in our daily life, this concept is familiar and basic. Control technology is widely used in industry, but at this moment only simple control laws are adopted in practice and higher performance is strongly desired. This can be achieved by modeling dynamics of the object and analyzing/designing the system in an intelligent way. It is also desirable to build a learning mechanism as a human shows improvement with his/her progress. It is thus our target to construct a highly intelligent control system. Because of its high level of abstraction, you might feel that it is not practical right now, but the knowledge you learn here will be surely useful when you enter the real world.
Robust Control / Adaptive control
We study advanced theories in post-modern robust/adaptive control and their applications. Various schemes of feedforward learning control(feedback error learning) are currently under investigation.
Independent Component Analysis
In this research, we study system identification, fault detection, and disturbance rejection via independent component analysis.
Probabilistic inference based optimal control / Reinforcement learning
We study stochastic optimal control and reinforcement learning.
Motor Skill Learning for Humanoid Robots
We are developing novel methods that enable robots to learn complex motor skills (e.g., biped walking, T-shirt wearing and clothing assistance) by optimal control and reinforcement learning.
Full-body Exoskeleton Robot Control for Walking Assistance
We propose an adaptive walking assistance strategy to control a full-body exoskeleton robot using an oscillator model.
(joint work with ATR CNS)
Active Motion Planning of Robots Based on Information Theoretic Criteria
We propose an active motion planning principle and its calculation algorithm based on information theoretic criteria such as the mutual information.
Intrinsically Motivation for Autonomous Robots
In order for the robot to truly autonomously work, it is necessary for the robot to determine the purpose by themselves rather than getting the purpose from designers. We are studying autonomous behavior acquisition of robot by designing reward based on the concept of intrinsic motivation.
Reservoir Computing for Multi-Objective Switchable Control
For the robot to acquire complex nonlinear dynamics such as human motion, it is effective to have nonlinear dynamics in the controller. We are studying motion control technology utilizing nonlinear dynamics of reservoir computing.
Marketings in Large-scale Online Social Networks
We are developing computationally efficient algorithms for locating the seeding nodes having the maximum influence on the outbreak of spreading processes over complex networks, toward achieving effective viral marketings in large-scale online social networks.
Network Centralities
We are developing large-scale algorithms for optimally tuning the network centralities (e.g., the PageRank) in complex networks under budget constraints.
Effectiveness and Optimality of the Bet-hedging Strategy
We are investigating the effectiveness and optimality of the bet-hedging strategy, an intrinsic mechanism of micro bio-populations for surviving through environmental fluctuations, by utilizing the analytical tools from the systems and control theory.
System Identification with Manifold Learning
We propose a nonlinear system identification scheme with input-output manifold learning. The scheme is based on a nonlinear dimensionality reduction method regularized by the latent dynamics structure.
Research Equipment
This is an anthropomorphic robot hand driven by pneumatic muscle, with the same DoF of human hand. On its fingertip, BioTac sensors by SynTouch are mounted, and tactile data such as pressure, vibration, temperature can be obtained. The collection of these data and the control of the hand are available via the Robot Operating System (ROS), and that allows us to apply this hand to various tasks. In this laboratory, we use this system for studies on the environment recognition of robot.
A wire-driven robot-arm with the same DOF as a human being and a robot hand with 3 fingers. The arm, without gears, provides the smooth and fast motion of a human arm. Torque control enables safe interaction with people. The mechanism that links movements of two fingers allows easy handling of various tasks. This equipment can be used in research in the development of human-like motor skills and object grasping.
This system is for measuring human’s movement. With 8 sensors, the motion of the upper half of the body could be measured with high frame rate. Since this is a wearing type, there is no blind spot and it’s available in many scenes. We utilize this system for the study on motor skill communication from human to robot.
The wireless CyberGlove II motion capture data glove is fully instrumented with 18 high-accuracy joint-angle measurements. It uses proprietary resistive bend-sensing technology to accurately transform hand and finger motions into real-time digital joint-angle data.
We utilize this system for various purposes such as robot hand manipulation.
This is a system for measuring, recording, and visualizing system of electromyogram (EMG) that may be generated when humans move muscles. Maximally 8ch information can be obtained with one machine, and we connect 2 this machines, accordingly, 16ch information can be obtained. In this laboratory, this system is utilized for the study on the human-robot interface, and on the motion modeling.
This is a system for measuring, recording, and visualizing system of electromyogram (EMG) that may be generated when humans move muscles (4ch, wireless). In this laboratory, this system is utilized for the study on the human-robot interface, and on the motion modeling.
Two-wheel drive type mobile robot. Remote control via Bluetooth is available, and this robot is utilized for validation experiments of control law of trajectory following and collision avoidance for nonholonomic systems.
Robot manipulators for the operation in space are required to be light and it requires large operation space. Since the arm part of such robot manipulators is generally made from flexible material, it is required to control the vibration of its tip caused by deflection of the link for the accurate position control. In this laboratory, we also try to obtain better response by adaptive control by learning.
Discs connected with flexible axis rotates with vibrating by motor input mounted the bottom of the system. The dynamics composition which relates to the actual system could be constructed by changing the number/position of discs and weights, fixed or non-fixed end. Also, disturbance generation system could be mounted externally, accordingly it allows us to do highly-reproducible experiments of disturbance removing.
This is well-known experiment system in control engineering. The pendulum will be controlled to be inverted by moving the cart. In this laboratory, these are utilized for experiment and validation for control theory.
This is an education material for programming developed by LEGO and MIT. Line tracing car and the inverted pendulum robot can easily constructed. In this laboratory, we take the advantage for demonstrations of control theory.
This is a computing machine for realtime robot control. This system consists of 24 cores CPU, 256GB memory, 1TB SSD, and GPU with 12GB memory, and enables us to process much data fast. We utilize this system to compute control laws and to learn models with big data.
This is the Baxter robot system, which has three compositions, Kinect, gripper and This system is built on ROS (Robot Operation System) network, where all compositions are managed by ROS core. Researches can conveniently control system with PC access to the ROS network. For our research, we use Baxter robot system to study robot handling deformable object.
With versatile arms and precise movements, the Nextage robot (15 DoF, 6 for arms × 2, 2 for neck, 1 for waist) is able to achieve and even surpass the manufacturing ability of most humans. It is suitable to be applied to Reinforcement Learning thanks to its highly functional body and scalable open software environment (based on C++ and Python with ROS). Our current project on Nextage is developing new Deep Reinforcement Learning Algorithm that learns control strategy from raw camera data.
