Interacting multiple model filter with error monitoring and recovery technique of perception net

Abstract

The paper is concerned with the problem of tracking a manoeuvring target. The error monitoring and recovery technique of the perception net is utilised to improve the tracking performance for a highly manoeuvring target. The interacting multiple model (IMM) filter is well known as being the most cost-effective hybrid-state estimation scheme. Both the detection of a manoeuvring target and the compensation of the estimated state can be achieved by employing a properly implemented error monitoring and recovery technique. IMM and full IMMPDA filters which employ an error monitoring and recovery technique show a better tracking performance compared with the conventional IMM and IMMPDA filters for a highly manoeuvring target, even in a clutter environment. They also reduce the peak values of the estimation errors when manoeuvring starts and finishes. The effectiveness of the proposed method is validated through simulations by comparing it with a conventional IMM filter. A similar evaluation is performed for a conventional IMMPDA filter.