A Standing Support Arm Design for Robotic Wheelchairs Using PPO-Based RL Strategy

Abstract

The aging of the population poses significant challenges, particularly in the mobility of the elderly. Robotic wheelchairs have emerged as a promising solution. However, the natural decline in the physical strength of the elderly makes independent sitting and standing movements challenging for wheelchair users. This study addresses the challenges of independent sit-to-stand movements for elderly wheelchair users by employing proximal policy optimization (PPO), a reinforcement learning (RL) method, to design a standing support arm. The research explores the potential and effectiveness of the standing support arm in aiding sit-to-stand motions. Through RL training in a simulated environment, the study demonstrates the support arm's capability to assist elderly individuals. Moreover, the comparison results show that the human model cannot complete independent standing without our standing support arm under the same training method. This further demonstrates that our approach provides a promising solution to the mobility challenges faced by an aging population, showing its potential in enhancing independence and reducing caregiver workload.