Reinforcement Learning and Feedback Control: Using Natural Decision Methods to Design Optimal Adaptive Controllers

By Frank L. Lewis, Draguna L. Vrabie, Kyriakos G. Vamvoudakis. IEEE Control Systems, 2012

987 citations recorded by the source corpus.

Subfields: Computational Theory and Mathematics, Control and Systems Engineering

View Paper: https://www.semanticscholar.org/paper/a66b0a309d9b27310442df2ac34e4f34a7f19a48

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computational Theory and Mathematics

1. Derong Liu
2. Ding Wang
3. Haibo He
4. Qinglai Wei
5. Zhong-Ping Jiang
6. Weinan Gao
7. Eduardo Alonso
8. Michael Fairbank

In Control and Systems Engineering

1. Derong Liu
2. Ding Wang
3. Haibo He
4. Qinglai Wei
5. Tim de Bruin
6. Zhong-Ping Jiang
7. Eduardo Alonso
8. Jens Kober

Citations in the corpus

• Safe Reinforcement Learning Using Robust MPC (2019, background): “Some attempts at combining RL and the linear quadratic regulator have been presented in [14], [15].”
• Value-gradient learning (2012, background): “Other ADP survey articles are available (Balakrishnan et al., 2008; Lendaris, 2009; Lewis and Vrabie, 2009; Lewis et al., 2012; Powell, 2009).”
• Adaptive Critic Nonlinear Robust Control: A Survey (2017, methodology): “From the view of convergence and optimality, the generalized value iteration and policy iteration for discrete-time systems [60], [63], [96] as well as the generalized policy iteration and value iteration for continuous-time systems [84], [88], [96] should be given special attention.”
• Reinforcement-Learning-Based Robust Controller Design for Continuous-Time Uncertain Nonlinear Systems Subject to Input Constraints (2015, methodology): “Traditionally, for utilizing RL approaches, a second NN called the action NN is introduced to approximate the control policy [23], [31]–[33], [39], [40].”
• Adaptive Dynamic Programming and Adaptive Optimal Output Regulation of Linear Systems (2016, methodology): “Departing from traditional model-based adaptive control methods [2], [17], [29], we invoke reinforcement learning theory [42] and approximate/adaptive dynamic programming (ADP) [5], [33], [38], [44] for non-model-based, data-driven adaptive optimal control design.”
• Data-Enabled Predictive Control: In the Shallows of the DeePC (2018, background): “ces that are representative for (and at the forefront of) different approaches in the literature. We begin with reinforcementlearningapproachesin which control actions are chosen to maximize a reward [16]–[19]. This requires either exploring the system via random control actions, or exploiting the k”
• Integral reinforcement learning and experience replay for adaptive optimal control of partially-unknown constrained-input continuous-time systems (2014, neutral): “A survey of RL-based feedback control designs is found in Lewis and Vrabie (2009), Lewis, Vrabie, and Vamvoudakis (2012).”
• Reinforcement learning for control: Performance, stability, and deep approximators (2018, methodology): “All the above approximation and stability considerations are even more di ffi cult when considering continuous-time dynamics and rewards as exemplified by integral RL (Lewis et al., 2012) and related approaches (Jiang and Jiang, 2012).”
• Policy Iteration Adaptive Dynamic Programming Algorithm for Discrete-Time Nonlinear Systems (2014, methodology): “Only in [36] and [37] a brief sketch of policy iteration algorithm for discrete-time systems was displayed, whereas the stability and convergence properties were not discussed.”