Non-linear MAP estimation using a continuous-state Viterbi algorithm

The Viterbi algorithm (VA) is a forward dynamic programming technique that computes a manufacturing automation protocol (MAP) state sequence estimate for a Markov process from noise corrupted measurements. A continuous state Viterbi algorithm (CSVA), in which the metric computation is performed using piecewise linear approximations and Bellman's maximum transform (R. Bellman, W. Karush, 1962), is presented. Simulation results that show that the CSVA performs faster and with a smaller error variance than a corresponding VA are also presented. >