In recent efforts to improve the theory of human control behaviour, different system identification techniques have been developed to estimate human linear time-invariant feedback- and feedforward control behavior. This thesis focuses on a newly developed subsystem identification
...
In recent efforts to improve the theory of human control behaviour, different system identification techniques have been developed to estimate human linear time-invariant feedback- and feedforward control behavior. This thesis focuses on a newly developed subsystem identification (SSID) technique, that estimates feedback- and feedforward behaviour from the closed-loop system data. To analyse SSID, the method is first applied to a versatile preview model. Consequently, the performance of SSID is compared to other system identification techniques used in similar applications, differing from SSID in the required number of forcing functions, the cost function data and the optimization algorithm. It is concluded that the SSID technique does not perform optimally, both in terms of estimation accuracy (average parameter estimation error of 30%) and computational time. Best results (average parameter estimation error of 10%) are obtained with a method that uses two forcing functions, the control output signal as cost function data and a Nelder-Mead simplex optimization algorithm.
