The objective of this research is to develop robust and accurate tracking algorithms for various tracking applications. These tracking problems can be formulated as nonlinear filtering problems. The tracking algorithms will be developed based on an emerging promising nonlinear filter technique, known as sequential importance sampling (nick-name: particle filtering). This technique was introduced to the engineering community in the early years of 2000, and it has recently drawn significant attention from engineers and researchers in a wide range of areas. Despite the encouraging results reported in the current literature, there are still many open questions to be answered. For the first time, the major research effort will be focusing on making improvement to the particle filter based tracking algorithm in the following three aspects: (I) refining the particle filtering process by designing better proposal distributions (II) refining the dynamic model by using multiple-model method, (i.e. using switching dynamics and jump Markov process) and (III) refining system measurements by incorporating a data fusion stage for multiple measurement cues.