LMTD planning algorithm from IPC 2011

Copied from original README

LMTD,IPC7 version Yanmei Hu, Dunbo Cai

Just like TFD, LMTD runs in three separate parts: translate(pddl->sas), preprocess(knowlege compilation), and search. Following is the steps to solve a planning task:

1. translate domain.pddl problem.pddl > output.sas (output.sas is for the inpput of preprocess) this step transforms the pddl tasks to sas tasks

2.preprocess < output.sas > output (do some pretreatments, for the search part)

3.search [1-7] t [1-1800] gpPn[1-3] plan_name < output

This part is for the search.The parameter [1-7] is for the order version(how to order the preconditions). The second and third parameters are for the time limit the running of search.The parameter g is for the gready search algorithm, p is for our heuristic choice, that is pc_heuristic P is for the choice of preferred operators and n is for the choice of no any heuristics.[1-3] here is for context state choice, every local problem has waiting nodes that store the unsatisfied preconditions, so we can choose any waiting nodes's context state as the final sate after the local problem is solved. plan_name store the plan for the successful running, while if unsuccessful, plan_name is empty. On the other hand, our planner is a anytime planner, so the plan for a planning task may be not just one. if our planner search several plans for a planning task, for example 3 plans( they are different), there will be 3 files, named plan_name1, plan_name2, plan_name3, for the output.

Note:for convenient, we simply the excutable of our planner, here is for how to run our planner combining three parts.

1. tempo-sat-LMTD/build

   This shell is to make our planner

2. tempo-sat-LMTD/plan domain.pddl problem.pddl plan_name

   This shell is to run our planner and get the plan for the input planning task.