Research

The well-known mean squared error (MSE) based algorithms are sensitive to outliers or non-Gaussian noise. To handle this issue, we propose a novel metric which is called multi-kernel correntropy and derive some robust estimators. Relevant works:

• Multi-kernel Maximum Correntropy Kalman Filter

• Generalized Multi-kernel Maximum Correntropy Kalman Filter for Disturbance Estimation

Inertial measurement units are widely used in the field of gait assessment, human-robot interaction, motion animation, and virtual and augmented reality. However, their performances are greatly affected by external acceleration and magnetic disturbance. To cope with this issue, we employ the multi-kernel correntropy to replace the mean squared error (MSE) based cost function, and derive some robust orientation estimation estimators for IMUs. Relevant works:

• Compact Maximum Correntropy-Based Error State Kalman Filter for Exoskeleton Orientation Estimation

• Multi-kernel Maximum Correntropy Kalman Filter for Orientation Estimation

To provide the individualized assistance profile for walking in a community, we provide a preference-based learning scheme for walking assistance in a community. Relevant work:

• Preference-based Assistance Map Learning with Robust Adaptive Oscillators

We propose a mixed-control scheme for strength augmentation using exoskeletons. The exoskeleton can carry a heavy load and follow the user's different walking modes, e.g., level walking, stair ascend, and stair descent. Relevant work:

• Gait Planning And Control Method Of Lower Extremity Exoskeletal Robot