Exploiting proper controller and bearing-only measurement to perform the target tracking with unmanned aerial vehicles (UAV) has become a popular trend due to the convenience of single bearing sensor and the flexibility of UAV. This article revisits the slowly moving target circumnavigation problem with UAV modelled as first-order integrator, presenting results in three aspects. First, we focus on the problem with discrete bearing data and control input, illustrating the principle parameter conditions for system convergence. Second, the thorough convergence analysis from a different perspective for a typical algorithm is presented, resulting the more accurate conclusion upon the target-circumnavigation system convergence. Third, based on the estimation property, an algorithm that exploits the rough information of target velocity is proposed, which significantly accelerates the estimation error convergence rate and improves the estimation performance. Simulation results are given that validates the conclusion that the estimation error and UAV-distance can be unbounded when the target velocity is specially designed under common assumptions in the literature, which corrects the conclusion in existing work, and also validates that the proposed algorithm possesses estimation performance obviously superior to that of existing algorithms especially for situation with constant-velocity target.

Conventional multi-target tracking (MTT) algorithms often struggle with scenarios governed by strong structural constraints, e.g.,exploiting integer programming solving the probabilistic objective is very computation expensive or intractable. This letter addresses the one-to-one association constraint, which is particularly relevant in high-fidelity tracking scenarios operating under a high detection-probability (PD) assumption, such as those with high signal-to-noise-ratio (SNR) sensor arrays. We propose a novel strong association constraint EM (SAC-EM) algorithm that embeds this hard constraint within a principled expectation-maximization (EM) framework. The E-step elegantly formulates the constrained data association as an entropic regularized optimal transport problem, which can be solved efficiently by the iterative proportional fitting (IPF) algorithm. The M-step then reduces to a weighted Kalman filter update. Comparative simulations demonstrate that the proposed algorithm achieves superior tracking accuracy over relevant baselines in the scenarios with constraints context, validating the EM framework as a powerful tool for integrating hard combinatorial constraints into probabilistic estimation.