no code implementations • 25 Sep 2024 • Carlos E. Luis, Alessandro G. Bottero, Julia Vinogradska, Felix Berkenkamp, Jan Peters
Optimal decision-making under partial observability requires reasoning about the uncertainty of the environment's hidden state.
no code implementations • 23 Feb 2024 • Alessandro G. Bottero, Carlos E. Luis, Julia Vinogradska, Felix Berkenkamp, Jan Peters
In this paper, we propose an information-theoretic safe exploration criterion that directly exploits the GP posterior to identify the most informative safe parameters to evaluate.
no code implementations • 29 Dec 2023 • Melrose Roderick, Felix Berkenkamp, Fatemeh Sheikholeslami, Zico Kolter
In many real-world problems, there is a limited set of training data, but an abundance of unlabeled data.
no code implementations • 7 Dec 2023 • Carlos E. Luis, Alessandro G. Bottero, Julia Vinogradska, Felix Berkenkamp, Jan Peters
We propose a new UBE whose solution converges to the true posterior variance over values and leads to lower regret in tabular exploration problems.
no code implementations • 1 Dec 2023 • Petru Tighineanu, Lukas Grossberger, Paul Baireuther, Kathrin Skubch, Stefan Falkner, Julia Vinogradska, Felix Berkenkamp
Meta-learning is a powerful approach that exploits historical data to quickly solve new tasks from the same distribution.
no code implementations • 25 Nov 2023 • Melrose Roderick, Gaurav Manek, Felix Berkenkamp, J. Zico Kolter
A key problem in off-policy Reinforcement Learning (RL) is the mismatch, or distribution shift, between the dataset and the distribution over states and actions visited by the learned policy.
1 code implementation • 12 Aug 2023 • Carlos E. Luis, Alessandro G. Bottero, Julia Vinogradska, Felix Berkenkamp, Jan Peters
Quantifying uncertainty about a policy's long-term performance is important to solve sequential decision-making tasks.
1 code implementation • 7 Jul 2023 • Jiarong Pan, Stefan Falkner, Felix Berkenkamp, Joaquin Vanschoren
Bayesian optimization (BO) is a popular method to optimize costly black-box functions.
1 code implementation • 24 Feb 2023 • Carlos E. Luis, Alessandro G. Bottero, Julia Vinogradska, Felix Berkenkamp, Jan Peters
We consider the problem of quantifying uncertainty over expected cumulative rewards in model-based reinforcement learning.
1 code implementation • 9 Dec 2022 • Alessandro G. Bottero, Carlos E. Luis, Julia Vinogradska, Felix Berkenkamp, Jan Peters
We consider a sequential decision making task where we are not allowed to evaluate parameters that violate an a priori unknown (safety) constraint.
1 code implementation • 22 Nov 2021 • Petru Tighineanu, Kathrin Skubch, Paul Baireuther, Attila Reiss, Felix Berkenkamp, Julia Vinogradska
Bayesian optimization is a powerful paradigm to optimize black-box functions based on scarce and noisy data.
no code implementations • ICLR 2022 • Lukas P. Fröhlich, Maksym Lefarov, Melanie N. Zeilinger, Felix Berkenkamp
In contrast, model-based methods can use the learned model to generate new data, but model errors and bias can render learning unstable or suboptimal.
no code implementations • 1 Jan 2021 • Felix Berkenkamp, Anna Eivazi, Lukas Grossberger, Kathrin Skubch, Jonathan Spitz, Christian Daniel, Stefan Falkner
Transfer and meta-learning algorithms leverage evaluations on related tasks in order to significantly speed up learning or optimization on a new problem.
1 code implementation • NeurIPS 2020 • Sebastian Curi, Felix Berkenkamp, Andreas Krause
Based on this theoretical foundation, we show how optimistic exploration can be easily combined with state-of-the-art reinforcement learning algorithms and different probabilistic models.
Model-based Reinforcement Learning reinforcement-learning +2
no code implementations • L4DC 2020 • Sebastian Curi, Silvan Melchior, Felix Berkenkamp, Andreas Krause
Finally, we show experimentally that our learning algorithm performs well in stable and unstable real systems with hidden states.
no code implementations • ICLR 2020 • Noah Siegel, Jost Tobias Springenberg, Felix Berkenkamp, Abbas Abdolmaleki, Michael Neunert, Thomas Lampe, Roland Hafner, Nicolas Heess, Martin Riedmiller
In practice, however, standard off-policy algorithms fail in the batch setting for continuous control.
no code implementations • 19 Feb 2020 • Noah Y. Siegel, Jost Tobias Springenberg, Felix Berkenkamp, Abbas Abdolmaleki, Michael Neunert, Thomas Lampe, Roland Hafner, Nicolas Heess, Martin Riedmiller
In practice, however, standard off-policy algorithms fail in the batch setting for continuous control.
no code implementations • NeurIPS 2019 • Matteo Turchetta, Felix Berkenkamp, Andreas Krause
Existing algorithms for this problem learn about the safety of all decisions to ensure convergence.
1 code implementation • 16 Jul 2019 • Silvan Melchior, Sebastian Curi, Felix Berkenkamp, Andreas Krause
Finally, we show experimentally that our learning algorithm performs well in stable and unstable real systems with hidden states.
1 code implementation • 27 Jun 2019 • Torsten Koller, Felix Berkenkamp, Matteo Turchetta, Joschka Boedecker, Andreas Krause
We evaluate the resulting algorithm to safely explore the dynamics of an inverted pendulum and to solve a reinforcement learning task on a cart-pole system with safety constraints.
no code implementations • 10 Jan 2019 • Felix Berkenkamp, Angela P. Schoellig, Andreas Krause
In this paper, we present the first BO algorithm that is provably no-regret and converges to the optimum without knowledge of the hyperparameters.
1 code implementation • ICLR 2019 • Nikolay Nikolov, Johannes Kirschner, Felix Berkenkamp, Andreas Krause
Efficient exploration remains a major challenge for reinforcement learning.
no code implementations • 13 Nov 2018 • Robin Spiess, Felix Berkenkamp, Jan Poland, Andreas Krause
In this paper, we present a deep learning approach that uses images of the sky to compensate power fluctuations predictively and reduces battery stress.
no code implementations • 27 Sep 2018 • Felix Berkenkamp, Debadeepta Dey, Ashish Kapoor
Deep reinforcement learning has enabled robots to complete complex tasks in simulation.
1 code implementation • 2 Aug 2018 • Spencer M. Richards, Felix Berkenkamp, Andreas Krause
We demonstrate our method by learning the safe region of attraction for a simulated inverted pendulum.
1 code implementation • 22 Mar 2018 • Torsten Koller, Felix Berkenkamp, Matteo Turchetta, Andreas Krause
However, these methods typically do not provide any safety guarantees, which prevents their use in safety-critical, real-world applications.
1 code implementation • 23 Feb 2018 • Shromona Ghosh, Felix Berkenkamp, Gireeja Ranade, Shaz Qadeer, Ashish Kapoor
We specify safety constraints using logic and exploit structure in the problem in order to test the system for adversarial counter examples that violate the safety specifications.
1 code implementation • NeurIPS 2017 • Felix Berkenkamp, Matteo Turchetta, Angela P. Schoellig, Andreas Krause
Reinforcement learning is a powerful paradigm for learning optimal policies from experimental data.
Model-based Reinforcement Learning reinforcement-learning +2
no code implementations • 3 Mar 2017 • Alonso Marco, Felix Berkenkamp, Philipp Hennig, Angela P. Schoellig, Andreas Krause, Stefan Schaal, Sebastian Trimpe
In practice, the parameters of control policies are often tuned manually.
1 code implementation • NeurIPS 2016 • Matteo Turchetta, Felix Berkenkamp, Andreas Krause
We define safety in terms of an, a priori unknown, safety constraint that depends on states and actions.
2 code implementations • 15 Mar 2016 • Felix Berkenkamp, Riccardo Moriconi, Angela P. Schoellig, Andreas Krause
The ROA is typically estimated based on a model of the system.
Systems and Control
3 code implementations • 14 Feb 2016 • Felix Berkenkamp, Andreas Krause, Angela P. Schoellig
While an initial guess for the parameters may be obtained from dynamic models of the robot, parameters are usually tuned manually on the real system to achieve the best performance.
3 code implementations • 3 Sep 2015 • Felix Berkenkamp, Angela P. Schoellig, Andreas Krause
One of the most fundamental problems when designing controllers for dynamic systems is the tuning of the controller parameters.
Robotics