A trajectory-oriented estimation method based on polynomial time series model and its decoupled implementation

In general, the complex nature of the motion model is prone to resulting in computational difficulties in target tracking, which holds true particularly in the trajectory-oriented estimation (TOE) method. As TOE needs to take into account historical information of multiple data frames, the computational complexity further escalates, which exerts adverse influence on the tracking performance. On that account, this paper puts forward a target tracking approach on the basis of TOE. we model the entire target trajectory and its kinematic evolution as a polynomial time series. The polynomial coefficients are estimated in accordance with the criterion of the minimum sum of squared deviations, and an recursive solution is deduced in matrix form. To lower the computational complexity, we further developed a decoupled implementation of recursive polynomial coefficients solution to circumvent matrix operations. Aside from that, decoupled solution instances of different polynomial orders are established to estimate target trajectories with distinct motion characteristics. Simultaneously, the smoothing state introduced during decoupling can eliminate the fluctuations spawned from uncertainties such as unknown biases in the measurements data, which gives rise to superior accuracy of state estimation. As suggested by the simulation results, this TOE algorithm exhibits better tracking performance compared to traditional recursive filtering algorithms, which effectively heightens the operational speed while guaranteeing tracking accuracy.