Personalized gait trajectory generation for lower-limb rehabilitation robots using anthropometric features

Area of Science:

• Biomechanics
• Robotics
• Machine Learning

Background:

• Conventional gait generation methods lack adaptability for variable speeds and rehabilitation needs.
• Accurate gait analysis is crucial for developing effective assistive and rehabilitative technologies.

Purpose of the Study:

• To develop an improved Gaussian process regression (IGPR) model for adaptable gait generation.
• To provide quantitative gait references for wearable devices in rehabilitation.

Main Methods:

• Constructed an IGPR model using hybrid kernel functions.
• Incorporated 12 anthropometric indicators, walking speed, and training modes.
• Validated predictions with hip and knee joint angle root mean square errors (RMSE) of 3.31° and 4.37°, respectively.

Main Results:

• The IGPR model demonstrated accurate gait trajectory prediction.
• Offline gait generation and robot tracking verification confirmed stable trajectory following in tests with healthy volunteers.

Conclusions:

• The developed IGPR model offers a robust solution for variable gait generation.
• This approach provides valuable quantitative gait data for wearable devices, with potential applications in stroke patient rehabilitation.