Composable Planning with Attributes

TL;DR

This paper introduces a composable planning framework that leverages environment attributes to enable agents to generalize to complex tasks by planning in attribute space and executing learned policies.

Contribution

It proposes a method to learn attribute transition policies and build a graph for planning, allowing flexible task solving without task-specific training.

Findings

Successfully generalizes to longer, more complex tasks.

Demonstrates effectiveness in 3D block stacking, grid-world, and StarCraft.

Enables planning through attribute space for task generalization.

Abstract

The tasks that an agent will need to solve often are not known during training. However, if the agent knows which properties of the environment are important then, after learning how its actions affect those properties, it may be able to use this knowledge to solve complex tasks without training specifically for them. Towards this end, we consider a setup in which an environment is augmented with a set of user defined attributes that parameterize the features of interest. We propose a method that learns a policy for transitioning between "nearby" sets of attributes, and maintains a graph of possible transitions. Given a task at test time that can be expressed in terms of a target set of attributes, and a current state, our model infers the attributes of the current state and searches over paths through attribute space to get a high level plan, and then uses its low level policy to execute the plan. We show in 3D block stacking, grid-world games, and StarCraft that our model is able to generalize to longer, more complex tasks at test time by composing simpler learned policies.