A Review of Kalman Filter with Artificial Intelligence Techniques
Sukkeun Kim, Ivan Petrunin, Hyo‐Sang Shin
Abstract
Kalman filter (KF) is a widely used estimation algorithm for many applications. However, in many cases, it is not easy to estimate the exact state of the system due to many reasons such as an imperfect mathematical model, dynamic environments, or inaccurate parameters of KF. Artificial intelligence (AI) techniques have been applied to many estimation algorithms thanks to the advantage of AI techniques that have the ability of mapping between the input and the output, the so-called "black box". In this paper, we found and reviewed 55 papers that proposed KF with AI techniques to improve its performance. Based on the review, we categorised papers into four groups according to the role of AI as follows: 1) Methods tuning parameters of KF, 2) Methods compensating errors in KF, 3) Methods updating state vector or measurements of KF, and 4) Methods estimating pseudo-measurements of KF. In the concluding section of this paper, we pointed out the directions for future research that suggestion to focus on more research for combining the categorised groups. In addition, we presented the suggestion of beneficial approaches for representative applications.