KARTAM@SUSHI.STANFORD.EDU (Nabil Kartam) (03/03/88)
I am investigating the utility of AI-techniques in the Project Planning domain. I have studied in depth most papers about AI-planners such as STRIPS, NOAH, NONLIN, and SIPE. During this year, I have been using SIPE to plan different construction projects. I would like to share some of my experience with concern researchers and hopefully get some answers for some questions. I dicover that SIPE is unable to provide the most parallel network and would like to know whether you think NOAH or NONLIN will be able to do so. e.g. suppose we have the following (B1 and B2 are specific beams ; C1, C2, and C3 are specific columns ) The preconditions of building B1 are building C1 and building C2 The preconditions of building B2 are building C2 and building C3 if our goal is to build B1 and B2, then the plan should be: ---- ---- | C1 |------------------------->| B1 | ---- / ---- / / ---- / | C2 |\ ---- \ \ \ ---- \ ---- | C3 |------------------------->| B2 | ---- ---- However, the plan obtained by SIPE looks like this: ---- ---- ---- ---- | C1 |---------->| B1 |------| C3 |--------->| B2 | ---- / ---- ---- ---- / / ---- / | C2 | ---- Although this plan is logically correct, it is not the most parallel plan. There is no reason why the column C3 should start after B1. In fact, C3 should be scheduled without any preconditions. I know that David Wilkins is working to solve this problem in his planner, but I would like to know whether you think this is a hard or simple problem and whether there is any planner that could produce the most parallel plan. Thanks, Nabil Kartam Construction Eng.&Mgt Stanford University kartam@sushi -------