Filters








180,910 Hits in 5.6 sec

Learning-based Model Predictive Control for Safe Exploration [article]

Torsten Koller, Felix Berkenkamp, Matteo Turchetta, Andreas Krause
2018 arXiv   pre-print
In this paper, we present a learning-based model predictive control scheme that can provide provable high-probability safety guarantees.  ...  In our experiments, we show that the resulting algorithm can be used to safely and efficiently explore and learn about dynamic systems.  ...  Another area that has considered learning for control is model-based RL.  ... 
arXiv:1803.08287v3 fatcat:7v6bk4w7wzb67cv4ll4wsix32a

Learning-Based Model Predictive Control for Safe Exploration

Torsten Koller, Felix Berkenkamp, Matteo Turchetta, Andreas Krause
2018 2018 IEEE Conference on Decision and Control (CDC)  
We combine a provably safe learning-based MPC scheme that allows for input-dependent uncertainties with techniques from model-based RL to solve tasks with only limited prior knowledge.  ...  In this paper, we attempt to bridge the gap between learning-based techniques that are scalable and highly autonomous but often unsafe and robust control techniques, which have a solid theoretical foundation  ...  Reinforcement learning objective and MPC scheme We require an objective function that jointly encourages exploration and finding a good control strategy based on our current statistical model.  ... 
doi:10.1109/cdc.2018.8619572 dblp:conf/cdc/KollerBT018 fatcat:omofexgb6vbzrnuluspinjmnmu

Towards Safe Reinforcement-Learning in Industrial Grid-Warehousing

Per-Arne Andersen, Morten Goodwin, Ole-Christoffer Granmo
2020 Information Sciences  
On the other hand, model-based reinforcement learning tries to encode environment transition dynamics into a predictive model.  ...  The key contributions of this paper are summarized as follows: DVAE, a predictive model for n-state predictions, safety constraints using a constrained MDP scheme, safe exploration through risk-directed  ...  Safe predictive model In model-based reinforcement learning, the goal is to efficiently learn a predictive model that accurately learns the environment dynamics to predict future states given the current  ... 
doi:10.1016/j.ins.2020.06.010 fatcat:kzm4owdbzfhvvnwv653dvxm47u

Safe Interactive Model-Based Learning [article]

Marco Gallieri and Seyed Sina Mirrazavi Salehian and Nihat Engin Toklu and Alessio Quaglino and Jonathan Masci and Jan Koutník and Faustino Gomez
2019 arXiv   pre-print
The learned safe set and model can also be used for safe exploration, i.e., to collect data within the safe invariant set, for which a simple one-step MPC is proposed.  ...  This paper introduces Safe Interactive Model Based Learning (SiMBL), a framework to refine an existing controller and a system model while operating on the real environment.  ...  We are also grateful to Felix Berkenkamp for the support given while experimenting with their safe learning tools.  ... 
arXiv:1911.06556v2 fatcat:fx22zrnlhfdsxfqen66utzk3zi

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  ...  Previous review works focused on specific techniques-for example, learning-based model predictive control (MPC) (7) , iterative learning control (ILC) (8, 9) , model-based RL (10) , data-efficient policy  ... 
arXiv:2108.06266v2 fatcat:gbbe3qyatfgelgzhqzglecr5qm

LS3: Latent Space Safe Sets for Long-Horizon Visuomotor Control of Sparse Reward Iterative Tasks [article]

Albert Wilcox and Ashwin Balakrishna and Brijen Thananjeyan and Joseph E. Gonzalez and Ken Goldberg
2021 arXiv   pre-print
A promising strategy for learning in dynamically uncertain environments is requiring that the agent can robustly return to learned safe sets, where task success (and therefore safety) can be guaranteed  ...  We present a novel continuous representation for safe sets by framing it as a binary classification problem in a learned latent space, which flexibly scales to image observations.  ...  These methods learn predictive models over either images or a learned latent space, which are then used by model predictive control (MPC) to optimize image-based task costs.  ... 
arXiv:2107.04775v2 fatcat:kmhxtsvfu5hojdrwff2qk6x7he

Improving Safety in Reinforcement Learning Using Model-Based Architectures and Human Intervention [article]

Bharat Prakash, Mohit Khatwani, Nicholas Waytowich, Tinoosh Mohsenin
2019 arXiv   pre-print
We present a hybrid method for reducing the human intervention time by combining model-based approaches and training a supervised learner to improve sample efficiency while also ensuring safety.  ...  Currently, Safe RL systems use human oversight during training and exploration in order to make sure the RL agent does not go into a catastrophic state.  ...  Government is authorized to reproduce and distribute reprints for Government purposes notwithstanding any copyright notation herein.  ... 
arXiv:1903.09328v1 fatcat:b2zsyziyprc7rgiukw4durauzu

Risk Sensitive Model-Based Reinforcement Learning using Uncertainty Guided Planning [article]

Stefan Radic Webster, Peter Flach
2021 arXiv   pre-print
Identifying uncertainty and taking mitigating actions is crucial for safe and trustworthy reinforcement learning agents, especially when deployed in high-risk environments.  ...  In this paper, risk sensitivity is promoted in a model-based reinforcement learning algorithm by exploiting the ability of a bootstrap ensemble of dynamics models to estimate environment epistemic uncertainty  ...  Funding in direct support of this work: EPSRC Centre for Doctoral Training Studentship (EP/S022937/1). The authors declare no conflicts of interest.  ... 
arXiv:2111.04972v1 fatcat:4d2y4uqm5zcl7eelnddkrztx3m

Safe Zero-Shot Model-Based Learning and Control: A Wasserstein Distributionally Robust Approach [article]

Aaron Kandel, Scott J. Moura
2021 arXiv   pre-print
We identify and focus on scenarios of learning and controlling a system from scratch, starting with a randomly initialized model based on the strongest possible limitations on our prior knowledge of the  ...  This paper labels this scenario as a "zero-shot" control problem, based on popular zero-shot transfer problems in machine learning.  ...  Safe Zero-Shot Model-Based Learning and Control: A Wasserstein Distributionally Robust Approach Aaron Kandel and Scott J.  ... 
arXiv:2004.00759v3 fatcat:znxsbd35azdlld44japrbp5xuq

Safe Reinforcement Learning on Autonomous Vehicles [article]

David Isele, Alireza Nakhaei, Kikuo Fujimura
2019 arXiv   pre-print
We investigate how prediction provides a general and intuitive framework to constraint exploration, and show how it can be used to safely learn intersection handling behaviors on an autonomous vehicle.  ...  Recent work in safe reinforcement learning uses idealized models to achieve their guarantees, but these models do not easily accommodate the stochasticity or high-dimensionality of real world systems.  ...  PREDICTION We propose using prediction models to mask unsafe actions from the agent, and then allow the agent to freely explore the safe state space using traditional RL techniques.  ... 
arXiv:1910.00399v1 fatcat:esa66f5pxzfgngpbw5kd7aq3eq

Text/Conference Paper

Christoph Zimmer, Mona Meister, Duy Nguyen-Tuong
2019 Jahrestagung der Gesellschaft für Informatik  
The proposed approach generates data appropriate for time series model learning, i.e. input and output trajectories, by dynamically exploring the input space.  ...  Learning time-series models is useful for many applications, such as simulation and forecasting.  ...  In safe active learning, [20] proposes a method for safe exploration based on the GP variance for stationary, i.e. pointwise, measurements.  ... 
doi:10.18420/inf2019_44 dblp:conf/gi/ZimmerMN19 fatcat:sb5yhqxwlbcf3i7t3jhlbgovte

Safe Reinforcement Learning with Chance-constrained Model Predictive Control [article]

Samuel Pfrommer, Tanmay Gautam, Alec Zhou, Somayeh Sojoudi
2022 arXiv   pre-print
We address the challenge of safe RL by coupling a safety guide based on model predictive control (MPC) with a modified policy gradient framework in a linear setting with continuous actions.  ...  We show theoretically that this penalty allows for a provably safe optimal base policy and illustrate our method with a simulated linearized quadrotor experiment.  ...  Model Predictive Control Model predictive control is a purely optimization-based planning framework.  ... 
arXiv:2112.13941v2 fatcat:3rbp3kk7iveefkjpvtzs4thj7u

Reinforcement learning based algorithm with Safety Handling and Risk Perception

Suhas Shyamsundar, Tommaso Mannucci, Erik-Jan van Kampen
2016 2016 IEEE Symposium Series on Computational Intelligence (SSCI)  
This paper presents the setup and the results of a reinforcement learning problem utilizing Q-learning and a Safety Handling Exploration with Risk Perception Algorithm (SHERPA) for safe exploration in  ...  The agent has to explore its environment safely and must learn the optimal action for a given situation from the feedback received from the environment.  ...  [6] propose a Qlearning based navigation algorithm for autonomous systems that shows good performance and learning for safe navigation in an unknown environment.  ... 
doi:10.1109/ssci.2016.7849367 dblp:conf/ssci/ShyamsundarMK16 fatcat:4rewatc4hvchlggqlxggy4dq7e

Safe Reinforcement Learning for Legged Locomotion [article]

Tsung-Yen Yang, Tingnan Zhang, Linda Luu, Sehoon Ha, Jie Tan, Wenhao Yu
2022 arXiv   pre-print
Model-free reinforcement learning provides promising tools to tackle this challenge. However, a major bottleneck of applying model-free reinforcement learning in real world is safety.  ...  Designing control policies for legged locomotion is complex due to the under-actuated and non-continuous robot dynamics.  ...  Ramadge for the helpful discussion and members of Google Research for helpful discussion and support of the experiments.  ... 
arXiv:2203.02638v1 fatcat:er3ctxha4vcotjwxtgfpeyqpxi

A predictive safety filter for learning-based control of constrained nonlinear dynamical systems [article]

Kim P. Wabersich, Melanie N. Zeilinger
2021 arXiv   pre-print
Safety is thereby established by a continuously updated safety policy, which is based on a model predictive control formulation using a data-driven system model and considering state and input dependent  ...  The predictive safety filter receives the proposed control input and decides, based on the current system state, if it can be safely applied to the real system, or if it has to be modified otherwise.  ...  Learning-based model predictive control: Originating from concepts in robust model predictive control (MPC), extensions of MPC schemes to safe learning-based methods have been proposed, see e.g.  ... 
arXiv:1812.05506v4 fatcat:chelnjgrrjgpnkvfjsoxarygae
« Previous Showing results 1 — 15 out of 180,910 results