On a probabilistic approach to synthesize control policies from example datasets

This paper is concerned with the design of control policies from example datasets. The case considered when just a black box description system to be controlled available and affected by actuation constraints. These constraints are not necessarily fulfilled (possibly, noisy) data under same as one which these collected. In this context, we introduce number methodological results compute policy datasets that: (i) makes behavior closed-loop similar illustrated in data; (ii) guarantees compliance We recast problem finite-horizon optimal give an explicit expression for its solution. Moreover, turn our findings into algorithmic procedure. procedure gives systematic tool policy. effectiveness approach via numerical example, where use real collected test drives synthesize merging car on highway.