Sufficient conditions for establishing maximum likelihood estimates for nonlinear process parameters

This paper considers nonlinear systems modeled by x(t) = f[t, x(t), u(t), ??]. The discrete observation process is corrupted by a zero-mean white Gaussian sequence {v(ti)}; i.e., z(ti) = y(ti) + v(ti), where y(ti) = h[x(ti), u(ti), ??], 1, 2, ... k. Sufficient conditions for identifiability of parameters ?? using the maximum likelihood estimate are established for this stochastic problem by extending previous results for deterministic systems and noiseless observations.