Aerospace Instrument-Making Annotation << Back Algorithm for Tracking Maneuvering Targets Based on a Transformer Wang Zhong, Neusypin K.A., Du Kehao In the field of maneuvering target tracking, traditional algorithms typically use several predefined mathematical models to approximate target motion. However, the inherent randomness and unpredictability of maneuvering targets pose a great challenge for accurate motion modeling. To address this problem, a multi-model estimation algorithm is developed that ensures the selection of the best-fi t model at a given time and performs dynamic weight assignment using Transformer. The effectiveness of the transformer-based multi-model estimation algorithm is demonstrated on the example of radar tracking of maneuvering targets, which significantly improves the accuracy and speed of tracking. Keywords: deep learning, multi-model estimation, Transformer, normalization, radar tracking, maneuvering target. DOI: 10.25791/aviakosmos.3.2025.1468 Pp. 48-59.