Latticed-LTL synthesis in the presence of noisy inputs

In the classical synthesis problem, we are given a specification ψ over sets of input and output signals, and we synthesize a finite-state transducer that realizes ψ: with every sequence of input signals, the transducer associates a sequence of output signals so that the generated computation satisfies ψ. In recent years, researchers consider extensions of the classical Boolean setting to a multi-valued one. We study a multi-valued setting in which the truth values of the input and output signals are taken from a finite lattice, and so is the satisfaction value of specifications. We consider specifications in latticed linear temporal logic (LLTL). In LLTL, conjunctions and disjunctions correspond to the meet and join operators of the lattice, respectively, and the satisfaction values of formulas are taken from the lattice too. The lattice setting arises in practice, for example in specifications involving priorities or in systems with inconsistent viewpoints. We solve the LLTL synthesis problem, where the goal is to synthesize a transducer that realizes the given specification in a desired satisfaction value. For the classical synthesis problem, researchers have studied a setting with incomplete information, where the truth values of some of the input signals are hidden and the transducer should nevertheless realize ψ. For the multi-valued setting, we introduce and study a new type of incomplete information, where the truth values of some of the input signals may be noisy, and the transducer should still realize ψ in the desired satisfaction value. We study the problem of noisy LLTL synthesis, as well as the theoretical aspects of the setting, like the amount of noise a transducer may tolerate, or the effect of perturbing input signals on the satisfaction value of a specification. We prove that the noisy-synthesis problem for LLTL is 2EXPTIME-complete, as is traditional LTL synthesis.