Proactive Dynamic Distributed Constraint Optimization Problems

The Distributed Constraint Optimization Problem (DCOP) formulation is a powerful tool for modeling multi-agent coordination problems. To solve DCOPs in dynamic environment, Dynamic (D-DCOPs) have been proposed to model the inherent dynamism present many D-DCOPs sequence of static problems by reacting changes environment as agents observe them. Such reactive approaches ignore knowledge about future problem. overcome this limitation, we introduce Proactive (PD-DCOPs), novel formalism presence exogenous uncertainty. In contrast approaches, PD-DCOPs are able explicitly possible problem and take such information into account when solving dynamically changing proactive manner. additional expressivity allows it wider variety distributed optimization Our work presents both theoretical practical contributions that advance current DCOP models: (i) We which how will change over time; (ii) develop exact heuristic algorithms manner; (iii) provide results complexity new class DCOPs; (iv) empirically evaluate determine trade-offs between two classes. final contribution important our first identify characteristics classes excel in.