Simulating Human Lifting Motions Using Fuzzy-Logic Control

Human motion simulation is an ill-posed problem. In order to predict unique lifting motion trajectories, a motion simulation model based on fuzzy-logic control is presented. The human body was represented by a 2-D five-segment model, and the neural controller was specified by fuzzy logic. Fuzzy rules were defined with their antecedent part describing the fuzzy variables of scaled positional error and velocity, and with their consequent part describing scaled angular velocity. These rules were generated according to certain trends in the fuzzy variable trajectories observed from actual lifting motions. An optimization procedure was performed to specify the parameters of the membership functions by minimizing the differences between the simulated and actual final postures. Simulations were obtained for 14 novel lifting motions from seven participants. Overall, results indicated that the presented model simulated lifting motions with an accuracy that was at least comparable to some previous human motion simulation models. The accuracy of the simulations differed between joints and was highest near the beginning and end of the motions. Strengths and limitations of the modeling approach are discussed. The use of fuzzy-logic control appears to be a fruitful basis for future simulations of lifting and other human tasks.