Meta-Control: Automatic Model-based Control Synthesis for Heterogeneous Robot Skills

Tianhao Wei, Liqian Ma, Rui Chen, Weiye Zhao, Changliu Liu
Proceedings of The 8th Conference on Robot Learning, PMLR 270:2295-2346, 2025.

Abstract

The requirements for real-world manipulation tasks are diverse and often conflicting; some tasks require precise motion while others require force compliance; some tasks require avoidance of certain regions while others require convergence to certain states. Satisfying these varied requirements with a fixed state-action representation and control strategy is challenging, impeding the development of a universal robotic foundation model. In this work, we propose Meta-Control, the first LLM-enabled automatic control synthesis approach that creates customized state representations and control strategies tailored to specific tasks. Our core insight is that a meta-control system can be built to automate the thought process that human experts use to design control systems. Specifically, human experts heavily use a model-based, hierarchical (from abstract to concrete) thought model, then compose various dynamic models and controllers together to form a control system. Meta-Control mimics the thought model and harnesses LLM’s extensive control knowledge with Socrates’ “art of midwifery” to automate the thought process. Meta-Control stands out for its fully model-based nature, allowing rigorous analysis, generalizability, robustness, efficient parameter tuning, and reliable real-time execution.

Cite this Paper

BibTeX
@InProceedings{pmlr-v270-wei25a, title = {Meta-Control: Automatic Model-based Control Synthesis for Heterogeneous Robot Skills}, author = {Wei, Tianhao and Ma, Liqian and Chen, Rui and Zhao, Weiye and Liu, Changliu}, booktitle = {Proceedings of The 8th Conference on Robot Learning}, pages = {2295--2346}, year = {2025}, editor = {Agrawal, Pulkit and Kroemer, Oliver and Burgard, Wolfram}, volume = {270}, series = {Proceedings of Machine Learning Research}, month = {06--09 Nov}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v270/main/assets/wei25a/wei25a.pdf}, url = {https://proceedings.mlr.press/v270/wei25a.html}, abstract = {The requirements for real-world manipulation tasks are diverse and often conflicting; some tasks require precise motion while others require force compliance; some tasks require avoidance of certain regions while others require convergence to certain states. Satisfying these varied requirements with a fixed state-action representation and control strategy is challenging, impeding the development of a universal robotic foundation model. In this work, we propose Meta-Control, the first LLM-enabled automatic control synthesis approach that creates customized state representations and control strategies tailored to specific tasks. Our core insight is that a meta-control system can be built to automate the thought process that human experts use to design control systems. Specifically, human experts heavily use a model-based, hierarchical (from abstract to concrete) thought model, then compose various dynamic models and controllers together to form a control system. Meta-Control mimics the thought model and harnesses LLM’s extensive control knowledge with Socrates’ “art of midwifery” to automate the thought process. Meta-Control stands out for its fully model-based nature, allowing rigorous analysis, generalizability, robustness, efficient parameter tuning, and reliable real-time execution.} }
Endnote
%0 Conference Paper %T Meta-Control: Automatic Model-based Control Synthesis for Heterogeneous Robot Skills %A Tianhao Wei %A Liqian Ma %A Rui Chen %A Weiye Zhao %A Changliu Liu %B Proceedings of The 8th Conference on Robot Learning %C Proceedings of Machine Learning Research %D 2025 %E Pulkit Agrawal %E Oliver Kroemer %E Wolfram Burgard %F pmlr-v270-wei25a %I PMLR %P 2295--2346 %U https://proceedings.mlr.press/v270/wei25a.html %V 270 %X The requirements for real-world manipulation tasks are diverse and often conflicting; some tasks require precise motion while others require force compliance; some tasks require avoidance of certain regions while others require convergence to certain states. Satisfying these varied requirements with a fixed state-action representation and control strategy is challenging, impeding the development of a universal robotic foundation model. In this work, we propose Meta-Control, the first LLM-enabled automatic control synthesis approach that creates customized state representations and control strategies tailored to specific tasks. Our core insight is that a meta-control system can be built to automate the thought process that human experts use to design control systems. Specifically, human experts heavily use a model-based, hierarchical (from abstract to concrete) thought model, then compose various dynamic models and controllers together to form a control system. Meta-Control mimics the thought model and harnesses LLM’s extensive control knowledge with Socrates’ “art of midwifery” to automate the thought process. Meta-Control stands out for its fully model-based nature, allowing rigorous analysis, generalizability, robustness, efficient parameter tuning, and reliable real-time execution.
APA
Wei, T., Ma, L., Chen, R., Zhao, W. & Liu, C.. (2025). Meta-Control: Automatic Model-based Control Synthesis for Heterogeneous Robot Skills. Proceedings of The 8th Conference on Robot Learning, in Proceedings of Machine Learning Research 270:2295-2346 Available from https://proceedings.mlr.press/v270/wei25a.html.

Related Material