Symmetric duality in continuous-time programming via minmax duality (Q2639336)

Consider the following pair of continuous-time nonlinear programming problems: \[ (P)\quad \inf \int^{T}_{0}[f(x,y)(t)-D_ yf(x,y)y(t)]dt\quad s.t.\quad \int^{T}_{0}D_ yf(x,y)y(t)dt\leq 0, \] \[ x(t)\geq 0,\quad y(t)\geq 0\text{ for all } t\in [0,T],\quad x\in W^ n[0,T],\quad y\in W^ m[0,T];\quad and \] \[ (D)\quad \sup \int^{T}_{0}[f(x,y)(t)-D_ xf(x,y)x(t)]dt\quad s.t.\quad \int^{T}_{0}D_ xf(x,y)x(t)dt\geq 0, \] \[ x(t)\geq 0,\quad y(t)\geq 0\text{ for all } t\in [0,T],\quad x\in W^ n[0,T],\quad y\in W^ m[0,T]. \] Using a method introduced in the static symmetric case by Stoer, duality results are established. A number of special cses, including one in which the functions are positively homogeneous, are considered. Finally, it is shown that if f is skew-symmetric, i.e., \(f(x,y)(t)=-f(y,x)(t)\), then (P) is self-dual.