Model-based cancellation of biodynamic feedthrough with a motorized manual control interface.

Because of vehicle ride motion, an operator on board a moving vehicle will tend to impose unintended forces on a manual control interface. This phenomenon is called Biodynamic feedthrough. Biodynamic feedthrough occurs in two scenarios, which may he distinguished by whether the machine under control is an auxiliary machine or the vehicle itself. In auxiliary machine control, Biodynamic feedthrough simply degrades manual control performance. In vehicle control, Biodynamic feedthrough forces close a loop around the manual control interface and vehicle, possibly leading to oscillations that may jeopardize safe vehicle operation. This dissertation presents a model of the human-machine system comprising the biomechanics and volitional manual control activity of the human operator and the dynamics of the vehicle and manual control interface. In the construction of the biomechanics model a distinction is made between the transmittance from vehicle acceleration to Biodynamic force and the driving point impedance at the control interface. The biomechanics model leads to a system identification experiment that identifies the transmittance using measurements of vehicle acceleration and interaction force between the operator's hand and a temporarily immobilized control interface. The estimated transmittance becomes the foundation for a model-based feedthrough cancellation controller. System identification results revealed similar but sufficiently distinct estimates for the various subjects, which prompted the construction of individualized cancellation controllers for each subject. The efficacy of the individualized cancellation controllers was tested in two experiments, each with twelve subjects. The first experiment involved the auxiliary machine control scenario and the second the vehicle control scenario. In auxiliary machine control experiments, the controller improved human performance in a pursuit tracking task as indicated by a 27% reduction in root mean square average tracking error and 32% increase in time on target (p p < 0.05).Ph.D.Applied SciencesElectrical engineeringIndustrial engineeringMechanical engineeringUniversity of Michigan, Horace H. Rackham School of Graduate Studieshttp://deepblue.lib.umich.edu/bitstream/2027.42/125506/2/3192783.pd