Multiple Object Trajectography Using Particle Swarm Optimization Combined to Hungarian Method

The problem of simultaneous trajectography of several dynamical objects is formulated as an optimization problem. The available observations consist in a series of photographs showing undiscriminated objects. The goal is to find the object initial states so that the resulting trajectories match as well as possible the set of observations. An assignment problem is solved at each observation date by the Hungarian method, yielding a deviation cost between the simulated trajectories and the measurements. A fitness function summing the deviation costs is minimized by a particle swarm algorithm. The method is illustrated on a space orbitography application.