Decision Transformer as a Foundation Model for Partially Observable Continuous Control

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Zhang, Xiangyuan, Mao, Weichao, Qiu, Haoran, Başar, Tamer
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916191038078976
author Zhang, Xiangyuan
Mao, Weichao
Qiu, Haoran
Başar, Tamer
author_facet Zhang, Xiangyuan
Mao, Weichao
Qiu, Haoran
Başar, Tamer
contents Closed-loop control of nonlinear dynamical systems with partial-state observability demands expert knowledge of a diverse, less standardized set of theoretical tools. Moreover, it requires a delicate integration of controller and estimator designs to achieve the desired system behavior. To establish a general controller synthesis framework, we explore the Decision Transformer (DT) architecture. Specifically, we first frame the control task as predicting the current optimal action based on past observations, actions, and rewards, eliminating the need for a separate estimator design. Then, we leverage the pre-trained language models, i.e., the Generative Pre-trained Transformer (GPT) series, to initialize DT and subsequently train it for control tasks using low-rank adaptation (LoRA). Our comprehensive experiments across five distinct control tasks, ranging from maneuvering aerospace systems to controlling partial differential equations (PDEs), demonstrate DT's capability to capture the parameter-agnostic structures intrinsic to control tasks. DT exhibits remarkable zero-shot generalization abilities for completely new tasks and rapidly surpasses expert performance levels with a minimal amount of demonstration data. These findings highlight the potential of DT as a foundational controller for general control applications.
format Preprint
id arxiv_https___arxiv_org_abs_2404_02407
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Decision Transformer as a Foundation Model for Partially Observable Continuous Control
Zhang, Xiangyuan
Mao, Weichao
Qiu, Haoran
Başar, Tamer
Systems and Control
Artificial Intelligence
Machine Learning
Robotics
Closed-loop control of nonlinear dynamical systems with partial-state observability demands expert knowledge of a diverse, less standardized set of theoretical tools. Moreover, it requires a delicate integration of controller and estimator designs to achieve the desired system behavior. To establish a general controller synthesis framework, we explore the Decision Transformer (DT) architecture. Specifically, we first frame the control task as predicting the current optimal action based on past observations, actions, and rewards, eliminating the need for a separate estimator design. Then, we leverage the pre-trained language models, i.e., the Generative Pre-trained Transformer (GPT) series, to initialize DT and subsequently train it for control tasks using low-rank adaptation (LoRA). Our comprehensive experiments across five distinct control tasks, ranging from maneuvering aerospace systems to controlling partial differential equations (PDEs), demonstrate DT's capability to capture the parameter-agnostic structures intrinsic to control tasks. DT exhibits remarkable zero-shot generalization abilities for completely new tasks and rapidly surpasses expert performance levels with a minimal amount of demonstration data. These findings highlight the potential of DT as a foundational controller for general control applications.
title Decision Transformer as a Foundation Model for Partially Observable Continuous Control
topic Systems and Control
Artificial Intelligence
Machine Learning
Robotics
url https://arxiv.org/abs/2404.02407