A Markovian model for predicting the impact of observation conditions on the reliability of sensory systems

To be reliable a sensory system must be able to determine the likelihood of success of the different sensing strategies under it disposal. For this reliability estimation to be meaningful, it has to be dynamic so as to respond to any changes in the observation conditions. It is therefore desirable to develop assessment methods that can continuously evaluate the reliability of potential sensory strategies taking in consideration changes in the observation conditions. This paper proposes a novel approach for capturing the impact of observation conditions on the reliability of sensory systems. The approach revolves around a new measure of information variation measure which provides the sensory system with a quantitative measure of the quality of the information gathered by it sensors. Since it can predict a more realistic estimate of the success rate based on the state of the observation conditions, the model provides valuable information that can be used to perform more effective sensor planning based on reliability considerations, and hence has the potential to enhance the capability of sensory systems in dealing with unstructured applications.