BC-Z: Zero-Shot Task Generalization with Robotic Imitation Learning

By Eric Jang, A. Irpan, Mohi Khansari, Daniel Kappler, Frederik D. Ebert, Corey Lynch, Sergey Levine, Chelsea Finn. 2022

813 citations recorded by the source corpus.

In this paper, we study the problem of enabling a vision-based robotic manipulation system to generalize to novel tasks, a long-standing challenge in robot learning. We approach the challenge from an imitation learning perspective, aiming to study how scaling and broadening the data collected can facilitate such generalization. To that end, we develop an interactive and flexible imitation learning system that can learn from both demonstrations and interventions and can be conditioned on different forms of information that convey the task, including pre-trained embeddings of natural language or videos of humans performing the task. When scaling data collection on a real robot to more than 100 distinct tasks, we find that this system can perform 24 unseen manipulation tasks with an average success rate of 44%, without any robot demonstrations for those tasks.

Subfields: Artificial Intelligence, Control and Systems Engineering, Reinforcement Learning

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Control and Systems Engineering

In Reinforcement Learning

Citations in the corpus

Do As I Can, Not As I Say: Grounding Language in Robotic Affordances (2022, methodology): “In our implementation, we train the individual skills either with image-based behavioral cloning, following the BC-Z method [13], or reinforcement learning, following MTOpt [14].”
Pre-Trained Language Models for Interactive Decision-Making (2022, background): “Several recent papers [36, 17, 15] propose to use LM for policy learning.”
PaLM-E: An Embodied Multimodal Language Model (2023, background): “Prior works focus on combining vision and language inputs in an embodied setting with the goal of direct action prediction (Guhur et al., 2022; Shridhar et al., 2022b;a; Zhang & Chai, 2021; Silva et al., 2021; Jang et al., 2022; Nair et al., 2022; Lynch et al., 2022; Brohan et al., 2022).”
Robotic Skill Acquisition via Instruction Augmentation with Vision-Language Models (2022, methodology): “…robotic control policies for following a wide range of textual instructions, often by integrating state-a of-the-art language embeddings and pretrained encoders with imitation learning on top of large, manually collected datasets of robotic demonstrations annotated with text commands [24].”
Unleashing Large-Scale Video Generative Pre-training for Visual Robot Manipulation (2023, neutral): “2022; Ahn et al., 2022; Driess et al., 2023; Du et al., 2023; Guhur et al., 2023; Jang et al., 2022; Lynch & Sermanet, 2020; Lynch et al., 2022; Mees et al., 2022a;c; Ha et al., 2023).”
What Matters in Language Conditioned Robotic Imitation Learning Over Unstructured Data (2022, methodology): “For example, the methods BC-Z and MIA [7], [8] use both behavior cloning, but different actions spaces and multi-modal alignment losses,”
R3M: A Universal Visual Representation for Robot Manipulation (2022, background): “Many of these works focus on collecting and learning from robot data itself [57, 58, 7, 8, 9, 10, 59].”
3D-VLA: A 3D Vision-Language-Action Generative World Model (2024, neutral): “The embodied datasets in use (Padalkar et al., 2023; Brohan et al., 2022; Jang et al., 2022) mainly consist of 2D images, deficient in 3D-related information.”
Evaluating Real-World Robot Manipulation Policies in Simulation (2024, neutral): “These advances are underpinned by large-scale datasets [11, 66] and expressive models [1, 6, 29].”
Interactive Language: Talking to Robots in Real Time (2022, methodology): “often requiring manual post-hoc success filtering [9], [10].”
Robotic Control via Embodied Chain-of-Thought Reasoning (2024, neutral): “Towards this goal, recent works have explored training “generalist robot policies” [10, 11, 12, 13, 14, 15, 16, 17] on diverse robot datasets [18, 19, 10, 20, 21, 22, 23, 24, 13, 14, 25, 15, 26, 6].”
RoboCasa: Large-Scale Simulation of Everyday Tasks for Generalist Robots (2024, methodology): “It has been used extensively in prior works [49, 31, 9, 2, 16, 6, 17].”
GELLO: A General, Low-Cost, and Intuitive Teleoperation Framework for Robot Manipulators (2023, neutral): “For manipulation, there have been great successes in leveraging imitation learning [1], [2], [3], [4] where system performance improves with larger datasets [5], [6], [7], [8].”
RT-2: Vision-Language-Action Models Transfer Web Knowledge to Robotic Control (2023, background): “Other works have incorporated pre-trained language models, often either as an instruction encoder [63, 64, 43, 41, 44, 1, 65] or for high-level planning [66, 19, 17, 67, 68, 69].”
RT-1: Robotics Transformer for Real-World Control at Scale (2022, methodology): “Z [23], as well as a larger version of BC-Z, which we refer to as BC-Z XL.”
Instruction-driven history-aware policies for robotic manipulations (2022, methodology): “Instruction-driven vision-based robotic manipulation has received growing attention for manipulations in 2D planar [57, 58] or recent 3D environments [8, 59, 60], and has been transferred to the real world [9, 10].”
Perceiver-Actor: A Multi-Task Transformer for Robotic Manipulation (2022, methodology): “This could be addressed in future works by using HG-DAgger style approaches to correct the agent [12].”