Solving coarse-grained DisCSPs with local search

Abstract

Distributed constraint satisfaction problems (DisCSPs) are often used to solve problems which are naturally distributed over a number of agents on different locations who cooperate to solve the overall problem. One common assumption in DisCSP resolution is that each agent is only responsible for one variable, i.e. problems are fine-grained. We consider the more realistic scenario where DisCSPs are coarse-grained. Thus, each agent holds multiple local variables, i.e. each agent is responsible for a complex local problem. In such cases, besides the constraints between variables held by different agents there are also local constraints between variables within an agent. Thus, agents have to find a balance between the emphasis they place on resolving their internal and external constraints. Placing slightly more emphasis on one group of constraints can compromise the collective ability of agents to reach agreement and solve problems. We introduce DisBO-wd, a stochastic local search algorithm based on DisBO (Distributed Breakout) which includes a weight decay mechanism and some randomisation. We also present Multi-DisPel, a penalty-based, distributed, local search algorithm which is able to solve coarse-grained DisCSPs efficiently. Multi-DisPeL uses penalties on values in order to escape local optima during problem solving rather than the popular weights on constraints. We compare Multi-DisPeL and DisBO-wd with other algorithms and show, empirically, that they are more efficient and at least as effective as state of the art algorithms in some problem classes.