Informational maneuvering in dynamic environment

By allowing sensors to continuously and autonomously correlate the outcomes of their previous observations and use them to plan for next observations, more effective sensory activities can be achieved. This necessitates that parameters which influence the performance of the sensory task, such as the spatial position and the control parameters of the sensor, possess some sort of a controllable dynamical behavior. In this paper, the authors propose a mathematical formulation which ties together, the state of uncertainty of the sensor and the parameters that control its sensing activities, The proposed model is based on the work of Malyshev et al. (1989) on “Observation Process Optimization”. First a model which mimics the uncertainty behavior of the sensor as a function of its control parameters is constructed using a set of first order differential equations which the authors call the differential-uncertainty control model. The active sensing problem is then defined as optimizing an objective function which is constrained by the differential-uncertainty control model as well as other resource constraints which deemed to be significant to the sensory task. A Kalman filter is used, recursively, to update the sensor´s estimate of the state of the environment. The authors demonstrate how the sensor continuously optimizes the value of its control parameters, using the formulation, so as to respond to changes in its working environment. An example of a moving vision sensor tracking a moving object is provided to explain the proposed formulation