A method for combinatorial explosion avoidance of AI planner and the application on test case generation

摘要

Combinatorial explosion is a key issue that leads to failures of planning for many planners. To avoid it, we modified the planner of IPP and divided its fact file into several small parts, and the method is called goal-decompounded. We also expanded the arithmetic of IPP. The modified planner we called MF-IPP able to handle multiple fact files, which avoided the combinatorial explosion. We applied the method on the GUI test case generation. The main idea was to produce the initial test case from planner firstly, and then propose a way of solution expanding to reinforce the generation. At last, we compared the performance of the two planners, and the result showed that MF-IPP can avoid the combinatorial explosion well.