A parameter adaptive simulated annealing algorithm for frequency estimation

A simulated annealing scheme based on a parameter adaptive coding schedule is proposed. In the existing annealing schemes, the temperature parameter is predetermined for every iteration step and is independent of the unknown parameter values. In the proposed scheme, the cooling temperature is made proportional to the deviation of each individual parameter at the earlier iteration step. The other key difference in the proposed scheme is that it never accepts a higher energy level and remains at the present lower energy position. Instead, the Boltzmann distribution is used to accept a larger cooling temperature, i.e. a broader parameter search space. The algorithm is then applied to the well-known nonlinear optimization problem of frequency/angles of arrival estimation of multiple sources. Simulation results indicate that the proposed scheme converges to the minimum energy level in fewer iteration steps when compared to an existing fast annealing algorithm.<>