Martinez-Villalpando, E. "Estimation of Ground Reaction Force and Zero Moment Point on a Powered Ankle-Foot Prothesis"

Abstract

Commercially available ankle-foot prostheses are passive when in contact with the ground surface, and thus, their mechanical properties remain fixed across different terrains and walking speeds. The passive nature of these prostheses causes many problems for lower extremity amputees, such as a lack of adequate balance control during standing and walking.

The ground reaction force (GRF) and the zero moment point (ZMP) are known to be basic parameters in bipedal balance control. This thesis focuses on the estimation of these parameters using two prostheses, a powered ankle-foot prototype and an instrumented, mechanically-passive prosthesis worn by a transtibial amputee. The main goal of this research is to determine the feasibility of estimating the GRF and ZMP primarily using sensory information from a force/torque transducer positioned proximal to the ankle joint. The location of this sensor is ideal because it allows the use of a compliant artificial foot to be in contact with the ground, in contrast to rigid foot structures employed by walking robots. Both, the active and passive, instrumented prostheses were monitored with a wearable computing system designed to serve as a portable control unit for the active prototype and as an ambulatory gait analysis tool. A set of experiments were conducted at MIT's gait laboratory whereby a below-knee amputee subject, using the prosthetic devices, was asked to perform single-leg standing tests and slow-walking trials. For each experiment, the GRF and ZMP were computed by combining the kinetic and kinematic information recorded from a force platform and a 3D motion capture system. These values were statistically compared to the GRF and ZMP estimated from the data collected by the embedded prosthetic sensory system and portable computing unit. The average RMS error and correlation factor were calculated for all experimental sessions.

Using a static analysis procedure, the estimation of the vertical component of GRF had an averaged correlation coefficient higher than 0.96. The estimated ZMP location had a distance error of less than 1 cm, equal to 4% of the anterior-posterior foot length or 12% of the medio-lateral foot width. These results suggest that it is possible to estimate the GRF between the ground and a compliant artificial prosthesis with a sensor positioned between the knee and the ankle joint. Moreover, this sensory information is sufficient to closely estimate the ZMP location during the single support phase of slow walking and while standing on one leg.

This research contributes to the development of fully integrated artificial extremities that mimic the behavior of the human ankle-foot complex, especially to help improve the postural stability of lower extremity amputees.