PROJECT TITLE :

Multiple Scan Data Association by Convex Variational Inference - 2018

ABSTRACT:

Data association, the reasoning over correspondence between targets and measurements, could be a drawback of fundamental importance in target tracking. Recently, belief propagation (BP) has emerged as a promising methodology for estimating the marginal chances of measurement to focus on association, providing quick, correct estimates. The excellent performance of BP in the actual formulation used may be attributed to the convexity of the underlying free energy, that it implicitly optimizes. This Project studies multiple scan information association issues, i.e., issues that reason over correspondence between targets and many sets of measurements, that might correspond to totally different sensors or different time steps. We tend to notice that the multiple scan extension of the only scan BP formulation is nonconvex and demonstrate the undesirable behavior that can result. A convex free energy is constructed using the recently proposed fractional free energy (FFE). A convergent, BP-like algorithm is provided for the single scan FFE, and used in optimizing the multiple scan free energy using primal-twin coordinate ascent. Finally, based on a variational interpretation of joint probabilistic information association (JPDA), we develop a sequential variant of the algorithm that's almost like JPDA, however retains consistency constraints from previous scans. The performance of the proposed methods is demonstrated on a bearings solely target localization problem.