Better time constrained search via randomization and postprocessing

Most of the satisficing planners which are based on heuristic search iteratively improve their solution quality through an anytime approach. Typically, the lowest-cost solution found so far is used to constrain the search. This avoids areas of the state space which cannot directly lead to lower cost so-lutions. However, in conjunction with a post-processing plan improvement system such as ARAS, this bounding approach can harm a planner’s performance. The new anytime search framework of Diverse Any-Time Search avoids this behaviour by taking advantage of the fact that post-processing can often improve a lower quality input plan to a superior final plan. The framework encourages di-versity of “raw ” plans by restarting and using randomization, and does not use previous solutions for bounding. This gives a post-processing system a more diverse set of plans to work on, which improves performance. When adding both Diverse Any-Time Search and the ARAS post-processor to LAMA-2011, the winner of the most recent IPC planning competi-tion, and AEES, the Anytime Explicit Estimation Algorithm, the performance on the 550 IPC 2008 and IPC 2011 problems is improved by almost 60 points according to the IPC metric, from 511 to over 570 on LAMA-2011, and 73 points from 440 to over 513 on AEES