Robotic Control via Embodied Chain-of-Thought Reasoning

By Michał Zawalski, William Chen, Karl Pertsch, Oier Mees, Chelsea Finn, Sergey Levine. Conference on Robot Learning, 2024

358 citations recorded by the source corpus.

A key limitation of learned robot control policies is their inability to generalize outside their training data. Recent works on vision-language-action models (VLAs) have shown that the use of large, internet pre-trained vision-language models as the backbone of learned robot policies can substantially improve their robustness and generalization ability. Yet, one of the most exciting capabilities of large vision-language models in other domains is their ability to reason iteratively through complex problems. Can that same capability be brought into robotics to allow policies to improve performance by reasoning about a given task before acting? Naive use of"chain-of-thought"(CoT) style prompting is significantly less effective with standard VLAs because of the relatively simple training examples that are available to them. Additionally, purely semantic reasoning about sub-tasks, as is common in regular CoT, is insufficient for robot policies that need to ground their reasoning in sensory observations and the robot state. To this end, we introduce Embodied Chain-of-Thought Reasoning (ECoT) for VLAs, in which we train VLAs to perform multiple steps of reasoning about plans, sub-tasks, motions, and visually grounded features like object bounding boxes and end effector positions, before predicting the robot action. We design a scalable pipeline for generating synthetic training data for ECoT on large robot datasets. We demonstrate, that ECoT increases the absolute success rate of OpenVLA, the current strongest open-source VLA policy, by 28% across challenging generalization tasks, without any additional robot training data. Additionally, ECoT makes it easier for humans to interpret a policy's failures and correct its behavior using natural language.

Subfields: Artificial Intelligence, Cognitive Neuroscience

View Paper: https://www.semanticscholar.org/paper/56b545002ebb897de167a2de1603f06ab8f6bb0a

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 Cognitive Neuroscience

Citations in the corpus

CoT-VLA: Visual Chain-of-Thought Reasoning for Vision-Language-Action Models (2025, neutral): “…plans for multi-stage execution [39, 40], point trajectories [57], label bounding boxes of objects and gripper positions as additional observations [39], generate future image trajectories for open-loop following [32, 42], and generate fine-grained reward guidance for reinforcement learning [70].”
π0.5: a Vision-Language-Action Model with Open-World Generalization (2025, positive): “…both high-level and low-level inference, in a recipe that more closely resembles chain-of-thought [82] or test-time compute [39] methods, though unlike embodied chain-of-thought methods [88, 46, 61], the high-level inference process still runs at a lower frequency than low-level action inference.”
A Survey on Vision--Language--Action Models for Embodied AI (2024, neutral): “ECoT [93] re-trains OpenVLA [50] using reasoning chain datasets, enabling embodied CoT reasoning that further enhances its performance.”
TinyVLA: Toward Fast, Data-Efficient Vision-Language-Action Models for Robotic Manipulation (2024, neutral): “These models run efﬁciently, Vision-language models for robot learning: Robot learning [9], [18], [19], [20], [21] is an crucial topic in the robotics.”