Affiliations: Departamento de Informática, Universidad Carlos III de Madrid, Av. Universidad 30, Leganés, Madrid, Spain. E-mails: rfuentet@inf.uc3m.es, trosa@inf.uc3m.es
Correspondence:
[*]
Corresponding author: Raquel Fuentetaja, Departamento de Informática, Universidad Carlos III de Madrid, Av. Universidad 30, Leganés, Madrid, Spain. E-mail: rfuentet@inf.uc3m.es.
Abstract: One of the claims of Automated Planning is the advantage of representing planning tasks in a high-level language where there is a correspondence between the real environment and the problem being modelled. However, under certain circumstances some representations can be inefficient because they contain objects with irrelevant identities. Instead of forcing the designer to envision efficient alternatives, we propose an automatic compilation that recognises such problems and translates them into an alternative representation. In this representation, objects with irrelevant names are aggregated and their properties are encoded as counters. This class of problems is of special interest in environments where objects are created by domain actions and the names assigned to these new objects are not relevant for solving the task. The experimental results show that our compilation can reduce the size of the instantiation and the branching factor, resulting in an important improvement of the scalability of planners.
Keywords: Automated Planning (AP), Artificial Intelligence (AI), representation in AP
