Tracing and implementation of IMM Kalman filtering feed-forward compensation technology based on neural network

UAV has great varieties, large controllable velocity and angular velocity, which makes high requirements on automatically identifying capability and tracking accuracy of ground search and tracking system. It happens frequently that servo feed-forward compensation technology is added in the search and tracking system to improve the tracking accuracy. However, accurate estimations of target velocity and acceleration becomes the difficult point of controlling feed-forward compensation technology. This paper proposes to adopt IMM Kalman filter technology based on neural network to estimate velocity and acceleration of the moving targets, which served as input variable of the servo feed-forward compensation system to eliminate the error of the missing distance caused by velocity and acceleration. Neural network can identify target types and makes self-adaptive adjustment on IMM Kalman filter parameters, which is helpful to improve estimation accuracy. Experimental results show that IMM Kalman filter feed-forward compensation technology based on neural network in the search and tracking system can improve the tracking accuracy of the system by more than 3 times than the conventional Kalman filter compensation, and the model verification is effective.