Rearrangement Planning

Rearrangement planning focuses on developing algorithms that enable robots to efficiently manipulate objects within a space to achieve a desired configuration, addressing challenges like limited perception, confined spaces, and complex object interactions. Current research emphasizes hierarchical planning, leveraging techniques like graph attention networks and transformers to model object dependencies and optimize action sequences, often incorporating deep reinforcement learning or integer linear programming for efficient solution finding. These advancements are crucial for improving robotic manipulation capabilities in various applications, from household robotics and warehouse automation to more complex assembly tasks.

Papers