The Third Competition on Knowledge Engineering for Planning and Scheduling
暂无分享,去创建一个
We report on the staging of the third competition on knowledge engineering for AI planning and scheduling systems, held during ICAPS-09 at Thessaloniki, Greece in September 2009.. We give an overview of how the competition has developed since its first run in 2005, and its relationship with the AI planning field. This run of the competition focused on translators that when input with some formal description in an application-area-specific language, output solverready domain models. Despite a fairly narrow focus within knowledge engineering, seven teams took part in what turned out to be a very interesting and successful competition. Background and Rationale Knowledge engineering for planning and scheduling (P&S) systems is the process that deals with the acquisition, formulation, validation and maintenance of application knowledge, and the fusion of this knowledge with appropriate solver machinery to create a working system. The International Competition on Knowledge Engineering for Planning and Scheduling has been running since 2005 as a bi-annual event promoting the development and importance of the use of knowledge engineering methods and techniques within P & S. Past events include ICKEPS05 held during ICAPS at Monterey, California in June 2005, and ICKEPS-07 held during ICAPS at Providence, Rhode Island in September 2007. We report here on the third running of the competition, ICKEPS-09, held during ICAPS at Thessaloniki, Greece in September 2009. Clearly, the main focus in P & S is centered around solver engines which accept a domain and task model as input, and output solutions to P&S problems. This focus needs to be complemented with research on the construction, validation, and optimisation of the domain models and the domain model languages. The ICKEPS competition series was founded in order to encourage complementary research into the knowledge engineering aspects of P&S. ICKEPS has promoted the development and sharing of tools and platforms that promise more rapid, accessible, and effective ways to construct reliable and efficient P&S systems. This includes domain modelling, heuristic acquisition, planner-domain matching, domain knowledge validation and so forth. ICKEPS promotes the knowledge-intensive aspects of P&S by evaluating knowledge engineering tools within a competitive forum. The first two competitions focused on the more general aspects of knowledge engineering for planning, spanning knowledge acquisition, validation and refinement. For the third competition, we decided to focus in on a particular aspect of KE, as follows. It is important for the field of domain independent P&S that general solver engines can be accessed and used by non-AI experts, much in the way that constraint programming technology has been packaged and is available to the wider community. Considerable advances have been made in the last decade on the generality and efficiency of planning engines: we were concerned that knowledge engineering issues would limit their use outside of the community. One way to increase access is to consider application areas where a planning function would be potentially useful, and where there is already use of formal description languages. Experts in that area may be familiar with their own description languages, but not with P&S description languages such as PDDL. The task would then be to create a translator from the application language to a P&S solver input language, so that P&S solvers could be embedded into tool support in the application without the need for a planning expert. Another translator might also be needed – one that translates output from the solver (plans) back to the application. While being of obvious potential benefit to the application domain, this also promotes the visibility, usability, and exploitation of current P&S solvers, leading to further development of the technology as their use in new applications uncover new directions and challenges. Hence for the 3rd competition we focused on tools, translators and techniques that when input with a model described in an application-area-specific language, output solver-ready domain models. We were targeting application areas such as Web Services, Workflow, Business Modeling, E-Learning, Games, Narrative Generation etc, as there had already been some progress in using embedded P&S engines in these areas. As well as being useful tools in their own right, we postulate that the study of the translation process may highlight fundamental research problems in P & S, particularly in the use of domain-independent solvers. Many users in application areas of P & S would be tempted to implement their own solver, and embed specific heuristics. Rather, with ICKEPS-09 we sought to promote the use of existing domainindependent solvers, and highlight the research challenges, such as the expressiveness of their input/output languages.