State Estimation Using Neural networks

Abstract: In underwater scenario, tracking i.e. computing motion parameters of targets is a challenging task. In this field Sonars act as sensors which provide directional or approximate positional measurements of static/moving targets in Polar coordinate system which makes the system parameters nonlinearly related with measurements. The prevalent solutions available in literature like EKF, UKF and PF approximate the relation through linearization etc. Artificial Neural Networks (ANN) is famous as nonlinear model free data driven estimators. They were applied with striking success in the field of control systems (System Identification), Speech processing (Noise cancellation) etc. The present work is a novel effort to apply ANN to handle nonlinearity and to improve tracking accuracy. Performance of proposed algorithm is compared with existing tracking algorithms EKF.

Keywords: Target Tracking, State Estimation, Neural networks, Maneuver.