Safety and optimality in learning-based control at low computational cost

2025/05/12 by Baumann, Dominik, Kowalczyk, Krzysztof, Rojas, Cristian R. +2 · 1 citation
#FOS: Computer and information sciences #FOS: Electrical engineering #Machine Learning (cs.LG) #Systems and Control (eess.SY) #electronic engineering #information engineering

paper · doi:10.48550/arxiv.2505.08026

Abstract

Applying machine learning methods to physical systems that are supposed to act in the real world requires providing safety guarantees. However, methods that include such guarantees often come at a high computational cost, making them inapplicable to large datasets and embedded devices with low computational power. In this paper, we propose CoLSafe, a computationally lightweight safe learning algorithm whose computational complexity grows sublinearly with the number of data points. We derive both safety and optimality guarantees and showcase the effectiveness of our algorithm on a seven-degrees-of-freedom robot arm.

Citations

• PACSBO: Probably approximately correct safe Bayesian optimization
• A computationally lightweight safe learning algorithm
• Information-Theoretic Safe Exploration with Gaussian Processes
• Safe Learning in Robotics: From Learning-Based Control to Safe Reinforcement Learning
• GoSafe: Globally Optimal Safe Robot Learning
• Safe and Efficient Model-free Adaptive Control via Bayesian Optimization
• Adaptive and Safe Bayesian Optimization in High Dimensions via One-Dimensional Subspaces
• When Gaussian Process Meets Big Data: A Review of Scalable GPs
• Stagewise Safe Bayesian Optimization with Gaussian Processes
• Benchmarking Deep Reinforcement Learning for Continuous Control
• Bayesian Optimization with Safety Constraints: Safe and Automatic\n Parameter Tuning in Robotics
• A General Framework for Constrained Bayesian Optimization using Information-based Search
• Bayesian Optimization with Unknown Constraints
• Online Least Squares Estimation with Self-Normalized Processes: An Application to Bandit Problems
• Optimization Under Unknown Constraints

Cited by

• Computationally lightweight classifiers with frequentist bounds on predictions

Related

• A computationally lightweight safe learning algorithm
• GoSafe: Globally Optimal Safe Robot Learning
• Safe Learning in Robotics: From Learning-Based Control to Safe Reinforcement Learning
• Safe Online Control-Informed Learning
• A Physics-Informed Machine Learning Framework for Safe and Optimal Control of Autonomous Systems