AAAI 2022
Anthony Z. Liu(University of Michigan), Sungryull Sohn, Mahdi Qazwini(University of Michigan), Honglak Lee
Abstract
Real world tasks are compositional and hierarchical. Tasks can be composed of multiple subtasks (or sub-goals) that are dependent on each other. Real world tasks can also be parameterized, where subtasks can be composed of multiple interactions and objects, which we call entities. For example, a household kitchen environment may contain multiple objects (e.g. apple, pot) and multiple ways of interacting with said objects (e.g. place apple in pot, cook apple). To solve these tasks efficiently, an agent must infer subtask dependencies (e.g. an agent must execute pickup apple before place apple in pot), and generalize the inferred dependencies to new subtasks (e.g. place apple in pot is similar to place apple in pan). Moreover, an agent may also need to solve unseen tasks, which can involve unseen entities. To this end, we formulate parameterized subtask graph inference (PSGI), a method for modeling subtask dependencies using firstorder logic with subtask entities. To facilitate this, we learn entity attributes in a zero-shot manner, which are used as quantifiers (e.g. is_pickable(X)) for the parameterized subtask graph. We show this approach accurately learns the latent structure on hierarchical and compositional tasks more efficiently than prior work, and show PSGI can generalize by modelling structure on subtasks unseen during adaptation.