Computing Convex Coverage Sets for Faster Multi−Objective Coordination
Diederik Roijers‚ Shimon Whiteson and Frans Oliehoek
In this article, we propose new algorithms for multi-objective coordination graphs (MO-CoGs). Key to the efficiency of these algorithms is that they compute the convex coverage set (CCS) instead of the Pareto coverage set (PCS), i.e., the Pareto front. Not only is the CCS a sufficient solution set for a large class of problems, it also has important characteristics that facilitate more efficient solutions. We propose two main algorithms for computing the CCS in MO-CoGs. Convex multi-objective variable elimination (CMOVE) computes the CCS by performing a series of agent eliminations, which can be seen as solving a series of local multi-objective subproblems. Variable elimination linear support (VELS) iteratively identifies the single weight vector \bf w that can lead to the maximal possible improvement on a partial CCS and calls variable elimination to solve a scalarized instance of the problem for \bf w. VELS is faster than CMOVE for small and medium numbers of objectives and can compute an \varepsilon-approximate CCS in a fraction of the runtime. In addition, we propose variants of these methods that employ AND/OR tree search instead of variable elimination to achieve memory efficiency. We analyze the runtime and space complexities of these methods, prove their correctness, and compare them empirically against a naive baseline and an existing PCS method, both in terms of memory-usage and runtime. Our results show that, by focusing on the CCS, these methods achieve much better scalability in the number of agents than the current state of the art.