Distributed Constraint Solving

Abstract

Combinatorial search algorithms of the type discussed in Chapter 6 are usually described in terms of a single agent manipulating or traversing a representation of the search space. As shown, however, when the constraint that the search algorithm is run sequentially is removed, they may be viewed as a collection of agents, organized in a tree-shaped network, which co-operate in finding an optimal solution, passing suggested solutions upwards and bounds to solutions downwards. There is no centralized representation of the search space or overall central planner.