Dear IPC-5 competitor, The second test domain (openstacks) is now available on olimpia from /home/ipc5/domains/openstacks. I attach a short description of the domain and its variants. Moreover, please find below some *IMPORTANT* information and clarifications about the output files containing the plans computed by your competing planners and the ADL-to-STRIPS compiled versions of the test domains. Please read this carefully. Finally, all important information files and emails that you receive about IPC-5 are now also in the competition website (follow the link: "Information Files for the Competitors" of http://ipc5.ing.unibs.it). Regards, Alfonso ======================================================================= The following contains important information about - Use of Validator and tolerance value - The "PlanningTechnique" field - The ADL-to-STRIPS compiled versions Use of Validator and tolerance value ------------------------------------ Please remember to validate your output files (plans) using the latest available version of the Validator (VAL): http://www.cis.strath.ac.uk/~rh/val.html For all non-temporal plans only integer time points should be used. For (metric-)temporal plans, in every plan that you generate, when you compute the action start times, AS TOLERANCE VALUE PLEASE USE *0.001*. Very briefly, the tolerance is a necessary parameter for VAL to correctly manage the definition of mutex actions during the validation of a PDDL2.2 plan. The command line of VAL is: validate -t example: validate -t 0.001 domain.pddl p01.pddl p01.soln For more information about the tolerance parameter of VAL, please see the documentation that is available from the VAL website indicated above. For any problem concerning the use of VAL, please contact Derek Long (Derek.Long@cis.strath.ac.uk) and send a copy of your mail to me. The "PlanningTechnique" field ----------------------------- As specified in a previous message this field in the output file generated by your planner should be considered only by those planners that use multiple, alternative techniques or subplanners. Of course, it must be the case that your planner *AUTOMATICALLY* chooses the technique(s) that it uses to solve a problem. In other words, the command line to run your planner must be the same for every problem tested. If you have any doubt about this field in the output file, please let me know. ADL-to-STRIPS compiled versions ------------------------------- If a domain variant contains ADL constructs which can be compiled away using existing compilers, we will provide the compiled versions. If a planner can handle both the original and the compiled version, then the team of the planner can CHOOSE one of the two versions. If the compilation time that we observed when we generate the compiled files is not significant, then we will not consider the CPU time required to compile the domain. However, if the compilation time IS significant, AND it will be necessary to compare a planner that used the ADL version and a planner that used the compiled STRIPS version, then we will take this compilation time into account. The compilation time will be indicated in the compiled files. The compiled files are in the SUB-DIRECTORY "Strips" of the original version. Note that there is one compiled domain for each test problem. Propositional test domain variants are generated using the compiler ADL2STRIPS based on FF's code (from the web page of IPC-4); Time and MetricTime variants are generated using the ADL2STRIPS compiler based on LPG's code.