Autoregressive Action Sequence Learning for Robotic Manipulation
- URL: http://arxiv.org/abs/2410.03132v3
- Date: Mon, 18 Nov 2024 02:06:46 GMT
- Title: Autoregressive Action Sequence Learning for Robotic Manipulation
- Authors: Xinyu Zhang, Yuhan Liu, Haonan Chang, Liam Schramm, Abdeslam Boularias,
- Abstract summary: Existing autoregressive architectures generate end-effector waypoints sequentially as word tokens in language modeling.
We extend causal transformers' single-token prediction to support predicting a variable number of tokens in a single step.
We propose the Autoregressive Policy architecture, which solves manipulation tasks by generating hybrid action sequences.
- Score: 32.9580007141312
- License:
- Abstract: Designing a universal policy architecture that performs well across diverse robots and task configurations remains a key challenge. In this work, we address this by representing robot actions as sequential data and generating actions through autoregressive sequence modeling. Existing autoregressive architectures generate end-effector waypoints sequentially as word tokens in language modeling, which are limited to low-frequency control tasks. Unlike language, robot actions are heterogeneous and often include continuous values -- such as joint positions, 2D pixel coordinates, and end-effector poses -- which are not easily suited for language-based modeling. Based on this insight, we introduce a straightforward enhancement: we extend causal transformers' single-token prediction to support predicting a variable number of tokens in a single step through our Chunking Causal Transformer (CCT). This enhancement enables robust performance across diverse tasks of various control frequencies, greater efficiency by having fewer autoregression steps, and lead to a hybrid action sequence design by mixing different types of actions and using a different chunk size for each action type. Based on CCT, we propose the Autoregressive Policy (ARP) architecture, which solves manipulation tasks by generating hybrid action sequences. We evaluate ARP across diverse robotic manipulation environments, including Push-T, ALOHA, and RLBench, and show that ARP, as a universal architecture, outperforms the environment-specific state-of-the-art in all tested benchmarks, while being more efficient in computation and parameter sizes. Videos of our real robot demonstrations, all source code and the pretrained models of ARP can be found at http://github.com/mlzxy/arp.
Related papers
- FAST: Efficient Action Tokenization for Vision-Language-Action Models [98.15494168962563]
We propose a new compression-based tokenization scheme for robot actions, based on the discrete cosine transform.
Based on FAST, we release FAST+, a universal robot action tokenizer, trained on 1M real robot action trajectories.
arXiv Detail & Related papers (2025-01-16T18:57:04Z) - Coarse-to-fine Q-Network with Action Sequence for Data-Efficient Robot Learning [62.3886343725955]
We introduce Coarse-to-fine Q-Network with Action Sequence (CQN-AS), a novel value-based reinforcement learning algorithm.
We study our algorithm on 53 robotic tasks with sparse and dense rewards, as well as with and without demonstrations.
arXiv Detail & Related papers (2024-11-19T01:23:52Z) - PIVOT-R: Primitive-Driven Waypoint-Aware World Model for Robotic Manipulation [68.17081518640934]
We propose a PrIrmitive-driVen waypOinT-aware world model for Robotic manipulation (PIVOT-R)
PIVOT-R consists of a Waypoint-aware World Model (WAWM) and a lightweight action prediction module.
Our PIVOT-R outperforms state-of-the-art open-source models on the SeaWave benchmark, achieving an average relative improvement of 19.45% across four levels of instruction tasks.
arXiv Detail & Related papers (2024-10-14T11:30:18Z) - ReKep: Spatio-Temporal Reasoning of Relational Keypoint Constraints for Robotic Manipulation [31.211870350260703]
Keypoint Constraints (ReKep) is a visually-grounded representation for constraints in robotic manipulation.
ReKep is expressed as Python functions mapping a set of 3D keypoints the environment to a numerical cost.
We present system implementations on a wheeled single-arm platform and a stationary dual-arm platform.
arXiv Detail & Related papers (2024-09-03T06:45:22Z) - SAM-E: Leveraging Visual Foundation Model with Sequence Imitation for Embodied Manipulation [62.58480650443393]
Segment Anything (SAM) is a vision-foundation model for generalizable scene understanding and sequence imitation.
We develop a novel multi-channel heatmap that enables the prediction of the action sequence in a single pass.
arXiv Detail & Related papers (2024-05-30T00:32:51Z) - RoboScript: Code Generation for Free-Form Manipulation Tasks across Real
and Simulation [77.41969287400977]
This paper presents textbfRobotScript, a platform for a deployable robot manipulation pipeline powered by code generation.
We also present a benchmark for a code generation benchmark for robot manipulation tasks in free-form natural language.
We demonstrate the adaptability of our code generation framework across multiple robot embodiments, including the Franka and UR5 robot arms.
arXiv Detail & Related papers (2024-02-22T15:12:00Z) - PACT: Perception-Action Causal Transformer for Autoregressive Robotics
Pre-Training [25.50131893785007]
This work introduces a paradigm for pre-training a general purpose representation that can serve as a starting point for multiple tasks on a given robot.
We present the Perception-Action Causal Transformer (PACT), a generative transformer-based architecture that aims to build representations directly from robot data in a self-supervised fashion.
We show that finetuning small task-specific networks on top of the larger pretrained model results in significantly better performance compared to training a single model from scratch for all tasks simultaneously.
arXiv Detail & Related papers (2022-09-22T16:20:17Z) - Deep Imitation Learning for Bimanual Robotic Manipulation [70.56142804957187]
We present a deep imitation learning framework for robotic bimanual manipulation.
A core challenge is to generalize the manipulation skills to objects in different locations.
We propose to (i) decompose the multi-modal dynamics into elemental movement primitives, (ii) parameterize each primitive using a recurrent graph neural network to capture interactions, and (iii) integrate a high-level planner that composes primitives sequentially and a low-level controller to combine primitive dynamics and inverse kinematics control.
arXiv Detail & Related papers (2020-10-11T01:40:03Z)
This list is automatically generated from the titles and abstracts of the papers in this site.
This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.