MOBILE AGENT SEARCH BY MIMICKING THE EYE MOVEMENT

We address the problem of search for a hidden object by a mobile agent. It is assumed that the agent is equipped by all appropriate sensors and actuators, so the problem is in the definition of the trajectory of the search agent such that the search successfully terminates in minimal time. Formally, such problem can be represented in the terms of optimisation problem, but, unfortunately, the complete solution can be obtained only for extremely simplified and practically useless formulations.

Recent studies of this problem demonstrate that the best practical solution of the search problem is provided by the intermittent trajectory that includes ballistic and diffusive parts. Fortunately, it was also observed that the animals and birds follow the same trajectories while searching for food. Usually, such foraging process is considered as practically optimal, so the problem is reduced to specification of the appropriate intermittent process.

In the report, we present our results regarding the eyes motion while the human searchers for objects at the picture. The formal stochastic process that models such motion is used for the control of the mobile agent searching for a hidden object. We demonstrate that the eyes movement is represented by certain type of stochastic process similar to the widely used Levy flight, but, surprisingly, not exactly Levy flight. In addition, we also clarify several issues in the research of eyes movement, in particular, the formal dependence of eyes movement on the structure of the picture.

The indicated results are illustrated by the eyes trajectories obtained in the experiments and the simulated search trajectories by intermittent strategies and by the Levy flights