Performance of coordinating concurrent hierarchical planning agents using summary information

Summary information can find solutions at higher levels exponentially more quickly than at lower levels. We have identified heuristics and search techniques that can take advantage of summary information in finding coordinated plans. In addition, we have characterized a coordination algorithm that takes advantage of these search techniques and experimentally shown how it can make large improvements over an FAF ("fewest alternatives first") heuristic in finding optimal coordinated plans. More work is needed to show that these results translate to different domains, and future considerations include comparing this approach to other planning heuristics that capitalize on domain knowledge in order to better understand the relationship between the plan structure and the search performance. We expect the benefits of using summary information to also apply to hierarchical planning and wish to compare these techniques with current heuristics for concurrent hierarchical planning.