The introduction of algebraic operations on the set of trajectories of a nonlinear control system (Q912039)

For a linear control system \(\dot x=Ax+Bu\), if \(t\to x(t,u^ 1)\) and \(t\to x(t,u^ 2)\) denote solutions (each for some initial data) corresponding to controls \(u^ 1,u^ 2\), respectively, then \(t\to x(t,u^ 1+u^ 2)=x(t,u^ 1)+x(t,u^ 2)\) and for real \(\lambda\), \(t\to x(t,\lambda u^ 1)=\lambda x(t,u^ 1)\). In other words, the sum of solutions corresponding to controls \(u^ 1\) and \(u^ 2\) is a solution corresponding to control \(u^ 1+u^ 2\), and \(\lambda\) times a solution corresponding to control \(u^ 1\) is a solution corresponding to control \(\lambda u^ 1.\) Next, consider an n-dimensional affine (control linear) system \[ (1)\quad \dot x=X(x)+uY(x), \] X, Y smooth vector fields on \(R^ n\). The problem considered is to define a new notion of addition (denoted \(\oplus)\) and scalar multiplication (denoted \(\odot)\) so that if \(t\to x(t,u^ 1)\), \(t\to x(t,u^ 2)\) denote solutions of (1) for controls \(u^ 1\), \(u^ 2\), respectively, then \(t\to x(t,u^ 1)\oplus x(t,u^ 2)\) is the solution corresponding to control \(u^ 1\oplus u^ 2\) and \(t\to \lambda \odot x(t,u^ 1)\) is the solution corresponding to control \(\lambda \odot u^ 1\). In essence, the system (1) is ``linear'' relative to the new operations \(\oplus\), \(\odot\). Under conditions too complicated to state here, the existence of these new operations is shown to be equivalent to the existence of solutions of certain partial differential equations related to (1). This is, in essence, another attempt to linearize system (1). The results are not as conclusive as the necessary and sufficient conditions for linearization via the feedback group (state feedback and a local coordinate change) known as the Hunt-Su-Meyer theorem.