Reconstructing the flight kinematics of swarming and mating in wild mosquitoes: Supplementary Document

Particle filtering is a sequential Monte Carlo method [3] that uses importance sampling to draw samples from probability distributions. In a particle filter the target state is represented by a point mass particle set that is propagated and updated using conditional probability representations of the motion model and measurement model. Methods that improve the sampling efficiency include [3] resampling, Rao-Blackwellization, and the unscented transformation. Table 1 describes the sampling importance resampling (SIR) particle filter algorithm that we implement as part of our tracking system. The target estimate at each step is the maximum a posteriori (MAP) of the particle set defined as the particle having the maximum weight.