Namiko Saito, Joao Moura, Tetsuya Ogata, Marina Y. Aoyama, Shingo Murata, Shigeki Sugano, and Sethu Vijayakumar (The University of Edinburgh, UK)
ICRA 2023
Abstract
Nextage open cuts and arranges a pizza. The task consists in a robot picking a pickle and fixing on the cutting board, cutting a pickle with a vegetable knife, followed by placing a piece of it on a pizza dough and returning the remaining to the storage box. Then picking the box with pizza and moving it to the cutting board, cutting the pizza with a pizza cutter, and finally placing the box and knives in their initial positions.
One effective strategy for handling a long-horizon task is to break it down into short-horizon subgoals, that the robot can execute sequentially. The long-horizon task requires the robot to consider: (1) the order of the subtasks, (2) multiple subtask selection, (3) coordination of dual-arm, and (4) variations within a subtask. In this paper, we propose extending a predictive learning model using deep neural networks with a Subgoal Proposal Module (SPM), with the goal of making such tasks realizable.
Presentation Video
