Filters








4 Hits in 1.1 sec

GoSafe: Globally Optimal Safe Robot Learning [article]

Dominik Baumann and Alonso Marco and Matteo Turchetta and Sebastian Trimpe
2021 arXiv   pre-print
We derive conditions for guaranteed convergence to the global optimum and validate GoSafe in hardware experiments.  ...  SafeOpt is an efficient Bayesian optimization (BO) algorithm that can learn policies while guaranteeing safety with high probability.  ...  To guarantee optimality, we thus need to assume that the trajectory for the globally optimal parameters a lies within the largest safe area in state space that can be learned without risking a failure.  ... 
arXiv:2105.13281v1 fatcat:kas62vsd7bdnpacwymhu2imln4

GoSafeOpt: Scalable Safe Exploration for Global Optimization of Dynamical Systems [article]

Bhavya Sukhija, Matteo Turchetta, David Lindner, Andreas Krause, Sebastian Trimpe, Dominik Baumann
2022 arXiv   pre-print
We demonstrate the superiority of GoSafeOpt over competing model-free safe learning methods on a robot arm that would be prohibitive for GoSafe.  ...  This work proposes GoSafeOpt as the first algorithm that can safely discover globally optimal policies for high-dimensional systems while giving safety and optimality guarantees.  ...  GoSafeOpt can handle more complex and realistic dynamical systems compared to existing model-free learning methods for safe global exploration, such as GoSafe.  ... 
arXiv:2201.09562v3 fatcat:hks533e6ynfnbppobvuikgzmqm

Safe Learning in Robotics: From Learning-Based Control to Safe Reinforcement Learning [article]

Lukas Brunke, Melissa Greeff, Adam W. Hall, Zhaocong Yuan, Siqi Zhou, Jacopo Panerati, Angela P. Schoellig
2021 arXiv   pre-print
The last half-decade has seen a steep rise in the number of contributions on safe learning methods for real-world robotic deployments from both the control and reinforcement learning communities.  ...  Our review includes: learning-based control approaches that safely improve performance by learning the uncertain dynamics, reinforcement learning approaches that encourage safety or robustness, and methods  ...  implementation; and GoSafe (81), to explore beyond the initial safe region.  ... 
arXiv:2108.06266v2 fatcat:gbbe3qyatfgelgzhqzglecr5qm

Scalable Safe Exploration for Global Optimization of Dynamical Systems [article]

Bhavya Sukhija, Matteo Turchetta, David Lindner, Andreas Krause, Sebastian Trimpe, Dominik Baumann
2022
Our experiments on a robot arm that would be prohibitive for GoSafe demonstrate that GoSafeOpt safely finds remarkably better policies than competing safe learning methods for high-dimensional domains.  ...  This work proposes GoSafeOpt as the first algorithm that can safely discover globally optimal policies for complex systems while giving safety and optimality guarantees.  ...  GoSafeOpt In this section, we present our novel algorithm, GoSafeOpt, which retains the sample efficiency of SafeOpt and the global exploration of GoSafe to safely discover globally optimal policies for  ... 
doi:10.48550/arxiv.2201.09562 fatcat:paulnuq5rfajhkainfippwyska