A Persistent Spatial Semantic Representation for High-level Natural Language Instruction Execution
A Persistent Spatial Semantic Representation for High-level Natural Language Instruction Execution
复制标题
DOI：
--
复制
发表时间：
2021-07
期刊：
影响因子：
--
通讯作者：
Valts Blukis;Chris Paxton;D. Fox;Animesh Garg;Yoav Artzi
中科院分区：
文献类型：
--
作者：
Valts Blukis;Chris Paxton;D. Fox;Animesh Garg;Yoav Artzi
自然语言提供了一个可访问的和表达的接口来指定机器人代理的长期任务。然而，非专家可能会指定这样的任务与高层次的指令，通过几个抽象层抽象特定的机器人动作。我们建议，在长期执行范围内弥合语言和机器人动作之间的差距的关键是持久性表示。我们提出了一个持久的空间语义表示方法，并展示了它如何使建立一个代理，执行分层推理，有效地执行长期任务。我们在ALFRED基准上评估了我们的方法，并获得了最先进的结果，尽管完全避免了常用的分步说明。
Natural language provides an accessible and expressive interface to specify long-term tasks for robotic agents. However, non-experts are likely to specify such tasks with high-level instructions, which abstract over specific robot actions through several layers of abstraction. We propose that key to bridging this gap between language and robot actions over long execution horizons are persistent representations. We propose a persistent spatial semantic representation method, and show how it enables building an agent that performs hierarchical reasoning to effectively execute long-term tasks. We evaluate our approach on the ALFRED benchmark and achieve state-of-the-art results, despite completely avoiding the commonly used step-by-step instructions.