Resilient Distributed Constraint Optimization Problems

Dynamic distributed constraint optimization problems (Dynamic DCOPs) are useful in modeling various distributed combinatorial optimization problems that are dynamically changing over time. Previous attempts assume that it is possible for agents to take on a new solution each time the problem changes. However, in some applications, it is necessary to commit to a single solution at the start of the problem regardless of how the problem may change in the future. In this paper, we make the following contributions: (i) We propose a Resilient DCOP (R DCOP) model, which is a dynamic DCOP but the goal is to find a single solution that is resilient to future changes of the problem; (ii) We introduce a naive complete algorithm to solve this problem as well as several enhancement methods that can be used to speed it up; (iii) We provide theoretical analyses on the complexity of this problem and of our algorithm; and (iv) We empirically demonstrate the feasibility of our algorithm to solve R DCOPs.