In particular, it is easy to build a state tracker when the IS
system is flat i.e. when there exists a remarkable
output, called a flat output, which completely sums up the differential behavior of the IS system (or equivalently the zero dynamics for this output is trivial). There is then a one-to-one correspondence between trajectories of the IS system and trajectories of the flat output, and for that reason a controller tracking the flat output (simply obtained by IO linearization with respect to the flat output) is actually a state tracker. Note that existence of a flat output is indeed a property of x = f(x,u) alone, and has nothing to do with the tracking y=h(x) In a nutshell, the idea is simply to split the output tracking problem in two steps: 1. design, once for all, a state tracker using the flat output (this is a closed-loop problem) (2. design a trajectory generator to feed the state tracker. this is an open-loop problem). The role of the trajectory generator is to create a reference for the flat output (or equivalently a reference state trajectory) from the desired output trajectory. We thus indirectly control the tracking output through the flat output. An interest in studying flatness grows because many classes of systems, also nonlinear, important in applications are flat. The truth is that any flat system can be feedback linearized using dynamic feedback. Flatness is a system property, and it indicates that the nonlinear structure of the system is well characterized and one can exploit this structure in designing control algorithms for motion planning,
trajectory generation, or stabilization. Another advantage
of flatness comparing to dynamic feedback linearizations that it is a geometric property of a system, independent of a coordinate selection. Specifically, in a traditional control approach to systems, when one classifies a system as linear in a state-space, this is meaningless from the geometric point of view because the system is linear in the specific coordinate representation. In this way, flatness can be considered the proper geometric notion of linearity even though the system may be nonlinear in any selected coordinates. We say that the feedback u=k(x,z,w), z=a(x,z,w), is endogenous if the open-loop dynamics x= f (x,u) is equivalent to the closed-loop dynamics. x= f (x,k(x,z,w)), z=a(x,z,w), A dynamics is flat if and only if it is linearizable by endogenous feedback and a coordinate change. The controllability property is preserved by equivalence. The flatness property is helpful in trajectory tracking problems for nonlinear systems. Let us stress that flatness is not another way of performing feedback linearization, but it is a structural property of a system, and it helps to reveal the system properties necessary for an application of a feedback controller, e.g., feedback linearization, backstepping, or passivity-based control.
lf the value of the control at time t depends only on the
values, at the same instant of time, of the state x and of the external reference input, the control is said to be a Static (or Memoryless) State Feedback Control. Otherwise, if the control depends also on a set of additional state variables, i.e. if this control is itself the output of an appropriate dynamical system having its own internal state, driven by x and by the external reference input, we say that a Dynamic State Feedback Control is implemented. In a single-input single-output system, the most convenient structure for a Static State Feedback Control is the one in which the input variable u is set equal to u = (x) + (x)v (4.10) where v is the external reference input (see Fig. 4.2). In fact, the composition of this control with a system of the form x = f(x) + g(x)u y = h(x) yields a closed loop characterized by the similar structure x = f(x) + g(x)(x) + g(x)(x)v y = h(x)
The dynamics of correspond to the dynamics describing
the "internal" behavior of the system when input and initial conditions have been chosen in such a way as to constrain the output to remain identically zero. These dynamics, which are rather important in many of our developments, are called the zero dynamics of the system.