Learning Distributional Demonstration Spaces for Task-Specific Cross-Pose Estimation

TL;DR

This paper introduces a novel method for learning multimodal relative placement tasks that can handle diverse object configurations with minimal demonstrations, improving flexibility over previous relational reasoning approaches.

Contribution

The proposed approach extends relational reasoning networks with properties to learn multimodal placement solutions while maintaining translation-invariance and relational features.

Findings

Learned precise placement with 10-20 demonstrations

Handled diverse objects without human annotations

Retained relational and translation-invariant properties

Abstract

Relative placement tasks are an important category of tasks in which one object needs to be placed in a desired pose relative to another object. Previous work has shown success in learning relative placement tasks from just a small number of demonstrations when using relational reasoning networks with geometric inductive biases. However, such methods cannot flexibly represent multimodal tasks, like a mug hanging on any of n racks. We propose a method that incorporates additional properties that enable learning multimodal relative placement solutions, while retaining the provably translation-invariant and relational properties of prior work. We show that our method is able to learn precise relative placement tasks with only 10-20 multimodal demonstrations with no human annotations across a diverse set of objects within a category.