CUrrent RESEARCH PROJECTS (Ph.d. research project)
Learning-Based Accurate Cycle-Time Prediction for CNC Machines
This project aims to present a new strategy to accurately predict cycle-time for CNC machines. Based on Neural Networks' learning strategy, the scheme identifies kinematic profiles from the given geometric information.
past research Projects (M.E. research project)
Terrain-dependent Slip Risk Prediction for Planetary Exploration Rovers
Planetary exploration rovers play essential roles to investigate celestial surfaces carefully. Wheel slip due to loose soil poses mission failure, thus the rovers have to avoid high-slip areas. In order to identify such areas, this project aims to develop visual-information-based wheel-slip prediction for the rovers using machine learning.
The proposed approach mainly consists of two procedures: slope estimation from 3D information and terrain classification from image information to predict wheel slip corresponding to estimated slope angle and classified terrain. In order to validate the proposed approach, we develop a four-wheeled rover testbed with a rocked-bogie mechanism.
Experimental results show that the risk of wheel slip can be predicted considering the uncertainty of the slope estimation.
Traveling State Estimation of a Wheeled Robot for Lunar/Planetary Exploration
This project aims to estimate a state of a wheeled robot such as sinkage, slippage, and driving force to adopt challenging environments where the robot explore.
In order to estimate the traveling state of the robot, we implemented a ToF sensor and a 6-axis Force and Torque sensor on each wheel. In particular, wheel sinkage is estimated by 3D point clouds, slippage is estimated by calculating optical flow obtained from 2D amplitude images from a ToF sensor. In addition, the driving force is estimated by the acquired reaction force.
Proposed "Intelligent Wheel System" successfully estimated traveling state of the robot during indoor/outdoor field experiments.
Development and Evaluation of a Noise Reduction Method for a Time-of-Flight Camera
As a member of a team HAKUTO, which was a member to participate in Google Lunar XPRIZE, we developed a lunar exploration micro-rover. ToF cameras are promising technologies for robot awareness systems, which provide 3D point clouds for robots even in feature-less, unstructured lunar environments. In order to reduce the impact of lunar surface illumination conditions, we developed a software filtering method that fuses the 3D point cloud frames obtained at different integration times.
In a lunar analog environment at the Japan Aerospace Exploration Agency (JAXA), a ToF camera with the proposed method could capture accurate 3D point cloud.
Construction of a GPS simulator for a Satellite Ground Evaluation Environment
This project aims to develop GPS simulator for evaluating a GPS receiver, which is attached on a micro satellite and verifying a Kalman-filter-based micro-satellite trajectory estimation algorithms using the simulator.
As a result, we could verify that the GPS receiver obtained satellite position information on orbit generated from the trajectory estimation algorithms within 15 m of error.