Learning control of quantum-mechanical systems by laboratory identification of effective input-output maps

This paper addresses a new algorithm for laboratory learning control of quantum mechanical systems. The learning control is achieved through the laboratory identification of a successive number of effective input-output maps for the quantum mechanical system. The input to these maps consists of a linearly independent set of control fields, and the output consists of the resultant observations in the laboratory. In this paper, the observations are taken as the temporal evolution of the target expectation value. From this information, an effective input-output map is produced, and the best control to meet a desired objective can then be identified within the domain of the map. The process may be repeated as desired, if higher quality control is necessary. The basic logic behind this laboratory learning based approach is presented, along with a simulated illustration of its behavior for a simple few-state control problem.