| In target tracking, a discrete time estimation problem with Gaussian noise is encountered. If the observation model contains interference as well as an observation noise, neither the Kalman nor the extended Kalman filter algorithm can, in general, be used to estimate the states. This dissertation presents some new algorithms which can be used to estimate the states of both linear and nonlinear models with or without interference.; First, linear models with interference are considered, and then the classical estimation algorithms are applied to estimate the states. Next, three completely new smoothing algorithms were developed for both linear and nonlinear discrete models with or without random interference. For the disturbance noise, observation noise, and interference; only independency is required; Gaussian assumption is not necessary. Moreover, the motion and observation models are not restricted to be linear functions of the disturbance noise and interference. The models' functions are not limited to continuous functions, i.e., they could be noncontinuous functions. These smoothing algorithms estimate the states by first quantizing them and then using Decoding Techniques of Information Theory. Simulation results, which are also presented, have shown that for either linear or nonlinear models with interference, the new smoothing algorithms perform very well. Furthermore, in the absence of interference, they perform better than the extended Kalman filter algorithm for some nonlinear discrete models with Gaussian disturbance and observation noises. |
