Effect of Knowledge Representation on Model Based Planning: Experiments Using Logic Programming Encodings

Answer Set Programming: Towards Efficient and Scalable Knowledge Representation and Reasoning

Volume

Papers from the 2001 AAAI Spring Symposium

Contents

In this paper we implement planning using answer set programming. We consider the action language A and its extensions. We show that when the domain is described using richer features such as qualification, ramification and conditional effects not only the encoding is smaller, but also it takes less time to find a plan. We also show that encoding of Bacchus and Kabanza’s style temporal constraints is fairly straightforward in answer set planning. Finally, unlike other model enumeration planning encodings, in our encoding we can just give an upper bound of the length of the plan, instead of the exact length. We illustrate the above features using the blocks world example from the AIPS planning contests.