Abstract:
To state most navigation problems, it is necessary to correctly formulate the problem of filtering the state of a stochastic dynamic system by indirect observations. If navigation is performed in a noncooperative scenario where the observed object and the measuring system do not interact, a priori information about the motion model is either limited to simple descriptions of the target and environmental conditions, or is absent altogether. For such cases, the article suggests several universal models that require minimum information about the parameters of a moving object. The models are based on simple motion at a constant speed. The lack of information about this speed complements navigation with the task of identifying it. The direction of movement is set to the average speed, which can vary from trajectory to trajectory, including depending on the coordinates of the object at the time of detection. The uncertainty of motion is modeled by several variants of additive (correlated and uncorrelated) disturbances that simulate chaotic motion while maintaining the general direction. A more complex model assumes a periodic change in the average speed of motion, which is analogous to motion with a piecewise constant velocity. Such changes in the motion model are modeled by the Poisson flow of events. It is noted that even in the presence of a priori information about this flow, the task of navigation is accompanied by an extremely difficult task of identification, since it is required to estimate the changed average speed over a short observation interval. Typical examples of calculated trajectories are given.
Fulltext:
PDF file (294 kB)