Let's say i have the following dynamics $$G_d= \frac{e^{-3s}}{1+s}$$ $$G_p=\frac{e^{-4s}}{(2+s)(3+s)}$$
now on dynamic analysis we get $$G_{ff}=-\frac{G_d}{G_p}$$ we can see $G_{ff}$ is physically not realizable. what i want to do is design a controller that can effectively work if not completely. so this is what we have $$G_{ff}=e^{s}*\frac{s^2+5s+6}{1+s}$$ i converted this to $$G_{ff}=(1+s+\frac{s^2}{2})*\frac{(s^2+5s+6)}{(1+s)*(1+\lambda*s)^3}$$ so basically I expanded exponential part into a taylor series, and i divided the whole function by $(1+\lambda*s)^3$ in order to get a physically realizable function, and i have taken $\lambda$ to be very small, $\lambda=.0001$. Unfortunately the response on simulink to this system is showing huge oscillations and the controller is not working. What can i do to improve this model and am i going in the right direction.