%0 Journal Article %1 kamthe2017dataefficient %A Kamthe, Sanket %A Deisenroth, Marc Peter %D 2017 %K bayesian generative-models probability reinforcement-learning %T Data-Efficient Reinforcement Learning with Probabilistic Model Predictive Control %U http://arxiv.org/abs/1706.06491 %X Trial-and-error based reinforcement learning (RL) has seen rapid advancements in recent times, especially with the advent of deep neural networks. However, the majority of autonomous RL algorithms require a large number of interactions with the environment. A large number of interactions may be impractical in many real-world applications, such as robotics, and many practical systems have to obey limitations in the form of state space or control constraints. To reduce the number of system interactions while simultaneously handling constraints, we propose a model-based RL framework based on probabilistic Model Predictive Control (MPC). In particular, we propose to learn a probabilistic transition model using Gaussian Processes (GPs) to incorporate model uncertainty into long-term predictions, thereby, reducing the impact of model errors. We then use MPC to find a control sequence that minimises the expected long-term cost. We provide theoretical guarantees for first-order optimality in the GP-based transition models with deterministic approximate inference for long-term planning. We demonstrate that our approach does not only achieve state-of-the-art data efficiency, but also is a principled way for RL in constrained environments.

@article{kamthe2017dataefficient, abstract = {Trial-and-error based reinforcement learning (RL) has seen rapid advancements in recent times, especially with the advent of deep neural networks. However, the majority of autonomous RL algorithms require a large number of interactions with the environment. A large number of interactions may be impractical in many real-world applications, such as robotics, and many practical systems have to obey limitations in the form of state space or control constraints. To reduce the number of system interactions while simultaneously handling constraints, we propose a model-based RL framework based on probabilistic Model Predictive Control (MPC). In particular, we propose to learn a probabilistic transition model using Gaussian Processes (GPs) to incorporate model uncertainty into long-term predictions, thereby, reducing the impact of model errors. We then use MPC to find a control sequence that minimises the expected long-term cost. We provide theoretical guarantees for first-order optimality in the GP-based transition models with deterministic approximate inference for long-term planning. We demonstrate that our approach does not only achieve state-of-the-art data efficiency, but also is a principled way for RL in constrained environments.}, added-at = {2019-10-09T00:13:52.000+0200}, author = {Kamthe, Sanket and Deisenroth, Marc Peter}, biburl = {https://www.bibsonomy.org/bibtex/25e76aa1e6fd131d17d3dae7eb75ce4d9/kirk86}, description = {[1706.06491] Data-Efficient Reinforcement Learning with Probabilistic Model Predictive Control}, interhash = {7eb48af81f594c1526b1fef7dae530a0}, intrahash = {5e76aa1e6fd131d17d3dae7eb75ce4d9}, keywords = {bayesian generative-models probability reinforcement-learning}, note = {cite arxiv:1706.06491Comment: Accepted at AISTATS 2018,}, timestamp = {2019-10-09T00:13:52.000+0200}, title = {Data-Efficient Reinforcement Learning with Probabilistic Model Predictive Control}, url = {http://arxiv.org/abs/1706.06491}, year = 2017 }