Neural Constraint Satisfaction: Hierarchical Abstraction for Combinatorial Generalization in Object Rearrangement
This addresses the challenge of combinatorial generalization in object rearrangement for embodied agents, which is an incremental improvement over existing methods.
The paper tackled the problem of object rearrangement for embodied agents by developing a hierarchical abstraction approach to infer underlying entities from visual inputs and achieve combinatorial generalization, resulting in a method that outperforms current offline deep RL methods on simulated rearrangement tasks.
Object rearrangement is a challenge for embodied agents because solving these tasks requires generalizing across a combinatorially large set of configurations of entities and their locations. Worse, the representations of these entities are unknown and must be inferred from sensory percepts. We present a hierarchical abstraction approach to uncover these underlying entities and achieve combinatorial generalization from unstructured visual inputs. By constructing a factorized transition graph over clusters of entity representations inferred from pixels, we show how to learn a correspondence between intervening on states of entities in the agent's model and acting on objects in the environment. We use this correspondence to develop a method for control that generalizes to different numbers and configurations of objects, which outperforms current offline deep RL methods when evaluated on simulated rearrangement tasks.