Session: 07-10-02: Medical Robotics, Rehabilitation, and Surgery

Paper Number: 191158

Adaptive Gait Generation and Optimal Design of Rehabilitation Mechanisms Based on Individual Features 

In response to the inter-individual variation in trajectory requirements during rehabilitation training for patients with lower-limb motor dysfunction, conventional rehabilitation devices with fixed parameters often fail to adequately accommodate differences in height, lower-limb dimensions, and gait patterns among patients, thereby limiting training adaptability and rehabilitation effectiveness. To achieve personalized rehabilitation based on individual characteristics, this study constructs an integrated framework for gait trajectory prediction and rehabilitation linkage optimization driven by anthropometric parameters. Taking patient anthropometric parameters as inputs, the framework first generates individualized target gait trajectories, and then uses the predicted target trajectories to optimize parameters and compare the performance of candidate rehabilitation linkages, thus providing a systematic approach for the design of personalized rehabilitation schemes.

For gait trajectory prediction, a Random Forest algorithm is employed to establish the mapping between anthropometric parameters and gait trajectories, thereby generating adaptive target gait trajectories for individuals with different body characteristics. The results show that the predicted trajectories remain highly consistent with the measured trajectories in terms of overall contour, swing trend, and major geometric features. For two randomly selected subjects from the test set, the maximum error in the error curves is 3.13 cm, which is approximately 4.17% of the normal human stride length (75 cm) and falls within a clinically acceptable range. This indicates that the established gait prediction model can effectively achieve adaptive trajectory generation for individual differences.

For rehabilitation linkage optimization, the planar four-bar linkage, the Watt-I six-bar linkage, and the Stephenson-III six-bar linkage are selected as candidate configurations. Under unified constraints and a unified error evaluation criterion, the Covariance Matrix Adaptation Evolution Strategy (CMA-ES) is adopted to optimize the linkage parameters. The optimization aims to minimize the contour error between the end-point trajectory of the linkage and the target gait trajectory, and the trajectory fitting capability of the three linkage types is systematically compared. The results show that all three linkages can fit the target trajectory, but with significant differences in fitting accuracy. The planar four-bar linkage exhibits the poorest fitting performance, whereas the two six-bar linkages achieve substantially higher fitting accuracy. Among them, the Stephenson-III six-bar linkage performs best, with a mean pointwise error of only 0.3613 cm, demonstrating the strongest ability to approximate the target trajectory. These results indicate that, compared with the four-bar linkage, the six-bar linkages with stronger structural expressiveness offer more pronounced advantages in generating complex rehabilitation gait trajectories. Based on the optimized parameters of the Stephenson-III six-bar linkage, a lower-limb rehabilitation device is further designed. Compared with some existing lower-limb rehabilitation devices, the proposed device features a simpler mechanism composition, simpler control, and lower implementation cost, demonstrating the engineering feasibility of the proposed framework.

This study not only provides an effective solution to the difficulty of obtaining target gait trajectories in clinical rehabilitation training, but also offers a theoretical basis and methodological reference for rehabilitation robot linkage selection and parameter customization. The proposed personalized rehabilitation framework can improve the matching between rehabilitation training schemes and patient physiological characteristics, and thus has positive significance for enhancing the specificity and effectiveness of lower-limb rehabilitation training.

Presenting Author: Yang Wang Hefei University of Technology

Presenting Author Biography: Wang Yang
A PhD student from Hefei University of Technology, specializing in the fields of mechanisms and rehabilitation robots.

Authors:

Tingting Xie Hefei University of Technology
Ping Zhao Hefei University of Technology
Yuan Li Hefei University of Technology
Yang Wang Hefei University of Technology
Xu Han Hefei University of Technology