Enhancing recognizability of robotics environments

We consider the problem of state recognizability in robotics environments modeled by partially observable Markov decision processes. To make the model of robot-environment interaction more reliable, in the usual state transition table, we add to the state transition probabilities an additional continuous metric via the mean and the variance of some significant sensor measurements suitable to be kept under a continuous form, such as odometric measurements. These information allow one to greatly enhance the state recognizability. Our approach is general, and can be applied to any robotics application that requires compensation of the uncertainties due to sensor errors and to the randomness of robot action effects on its environment. We have devised some possible applications to modeling the interaction between a manipulator and its world, but in this paper, only a specific application to the navigation problem for a mobile robot is illustrated to show the feasibility of our approach.