Optimizing the robustness of software against communication latencies in distributed reactive embedded systems This paper presents a formal approach of designing software robust against communication latencies that typically occur in distributed embedded systems. In this approach, the software’s data-flow is retimed and scheduled in order to achieve the maximum robustness against possible communication latencies. This robustness is derived individually for a given software and its distribution on a platform’s communication topology. Robustness is interpreted as the guaranteed amount of time, up to which the system does not change its externally observable behavior due to communication latencies. The software’s data-flow is given as a data-flow graph with nodes representing tasks and edges representing communication channels. A linear problem approach is employed that transforms elements of data-flow into variables of linear expressions. An implementation of the approach in the tool Cadmos together with the application on a case example from the automotive software engineering domain shows its practicability.