Comparing predictive validity of four ballistic swing phase models of human walking

It is unclear to what extent ballistic walking models can be used to qualit atively predict the swing phase at comfortable walking speed. Different stu dy findings regarding the accuracy of the predictions of the swing phase ki nematics may have been caused by differences in (1) kinematic input, (2) mo del characteristics (e.g. the number of segments), and (3) evaluation crite ria. In the present study, the predictive validity of four ballistic swing phase models was evaluated and compared, that is, (1) the ballistic walking model as originally introduced by Mochon and McMahon, (2) an extended vers ion of this model in which heel-off of the stance leg is added, (3) a doubl e pendulum model, consisting of a two-segment swing leg with a prescribed h ip trajectory, and (4) a shank pendulum model consisting of a shank and rig idly attached foot with a prescribed knee trajectory. The predictive validi ty was evaluated by comparing the outcome of the model simulations with exp erimentally derived swing phase kinematics of six healthy subjects. In all models, statistically significant differences were found between model outp ut and experimental data. All models underestimated swing time and step len gth. In addition, statistically significant differences were found between the output of the different models. The present study shows that although q ualitative similarities exist between the ballistic models and normal gait at comfortable walking speed, these models cannot adequately predict swing phase kinematics. (C) 2001 Elsevier Science Ltd. All rights reserved.