Particle Filtering For Target Tracking

Particle filtering is a sequential Monte Carlo technique that recursively computes the posterior probability density function using the concept of “Importance Sampling”. This paper considers the application of particle filtering technique to a target tracking application, in which a radar sends a signal towards a target and estimates the state (position and velocity) of the target using the observations (time delay and Doppler shift) from the reflected signal. State model and measurement model have been derived for the proposed target tracking problem. Effectiveness of particle filtering technique has been demonstrated by comparing the results with those obtained with Kalman filtering technique. The prediction error obtained by using particle filtering technique is found to be significantly less than that error obtained from Kalman filtering technique. Keywords— Kalman filtering, likelihood function, observation model, Particle filtering, posterior density, state model.