Evaluating Pelvic External Forces and Moments in Assisted Walking: Implications for the Development of Robotic-Assisted Walkers

Securing sufficient walking volume is crucial for maintaining and improving gait function after stroke. Robot-assisted gait training devices play a significant role in enhancing gait ability; however, challenges such as their large size, high cost, and limited adaptability exist. The long-term goal of this research is to develop a robot-assisted wheeled walker for rehabilitation usable without constraints of location and time. This study aimed to quantify the manual assistance techniques employed by physical therapists (PTs) in controlling the pelvis during gait training and utilize these findings for robot control strategies. Two healthy adult volunteers simulating individuals with limited right lower-limb load (approximately 70% of body weight) walked with a wheeled walker while eight PTs provided manual assistance by applying external forces and moments to the participants' pelvis via a sensor-equipped pelvic belt. Data analysis revealed recurring patterns and notable periodicity (Average Correlation Coefficient ≥ 0.6) in several key assistive actions. Specifically, we identified peaks for: forward force by the right hand (0.175 N/kg at 19% of gait cycle), upward force by the left hand (0.475 N/kg at 30%), pelvic rotational moment by the left hand (0.013 Nm/kg at 44%), and upward force by the right hand (0.212 N/kg at 79%). These assistive techniques effectively increased weight-bearing on the right limb, prolonged the right stance phase, and extended the left swing phase, resulting in improved gait symmetry. The magnitude and timing of these assistive actions were consistent across PTs, underscoring that PTs employ consistent and periodic manual techniques to control the pelvis during gait assistance for individuals with unilateral weight-bearing limitations. The findings not only shed light on how PTs manually compensate for gait asymmetry but also offer a framework for designing control algorithms in robot-assisted wheeled walkers aimed at achieving more natural gait patterns.