The proceedings contain 45 papers. The topics discussed include: active learning for personalizing treatment;active exploration by searching for experiments that falsify the computed control policy;optimistic planning...
ISBN:
(纸本)9781424498888
The proceedings contain 45 papers. The topics discussed include: active learning for personalizing treatment;active exploration by searching for experiments that falsify the computed control policy;optimistic planning for sparsely stochastic systems;adaptive sample collection using active learning for kernel-based approximate policy iteration;tree-based variable selection for dimensionality reduction of large-scale control systems;high-order local dynamicprogramming;safe reinforcementlearning in high-risk tasks through policy improvement;agent self-assessment: determining policy quality without execution;reinforcementlearning algorithms for solving classification problems;reinforcementlearning in multidimensional continuous action spaces;grounding subgoals in information transitions;and directed exploration of policy space using support vector classifiers.
The proceedings contain 42 papers. The topics discussed include: approximate real-time optimal control based on sparse Gaussian process models;subspace identification for predictive state representation by nuclear nor...
ISBN:
(纸本)9781479945535
The proceedings contain 42 papers. The topics discussed include: approximate real-time optimal control based on sparse Gaussian process models;subspace identification for predictive state representation by nuclear norm minimization;active learning for classification: an optimistic approach;convergent reinforcementlearning control with neural networks and continuous action search;theoretical analysis of a reinforcementlearning based switching scheme;an analysis of optimistic, best-first search for minimax sequential decision making;information-theoretic stochastic optimal control via incremental sampling-based algorithms;policy gradient approaches for multi-objective sequential decision making: a comparison;and cognitive control in cognitive dynamic systems: a new way of thinking inspired by the brain.
The proceedings contain 28 papers. The topics discussed include: local stability analysis of high-order recurrent neural networks with multi-step piecewise linear activation functions;finite-horizon optimal control de...
ISBN:
(纸本)9781467359252
The proceedings contain 28 papers. The topics discussed include: local stability analysis of high-order recurrent neural networks with multi-step piecewise linear activation functions;finite-horizon optimal control design for uncertain linear discrete-time systems;adaptive optimal control for nonlinear discrete-time systems;optimal control for a class of nonlinear system with controller constraints based on finite-approximation-errors ADP algorithm;finite horizon stochastic optimal control of uncertain linear networked control system;real-time tracking on adaptive critic design with uniformly ultimately bounded condition;a novel approach for constructing basis functions in approximate dynamicprogramming for feedback control;and a combined hierarchical reinforcementlearning based approach for multi-robot cooperative target searching in complex unknown environments.
The proceedings contain 34 papers. The topics discussed include: a unified framework for temporal difference methods;efficient data reuse in value function approximation;constrained optimal control of affine nonlinear...
ISBN:
(纸本)9781424427611
The proceedings contain 34 papers. The topics discussed include: a unified framework for temporal difference methods;efficient data reuse in value function approximation;constrained optimal control of affine nonlinear discrete-time systems using GHJB method;algorithm and stability of ATC receding horizon control;online policy iteration based algorithms to solve the continuous-time infinite horizon optimal control problem;real-time motor control using recurrent neural networks;hierarchical optimal control of a 7-DOF Arm model;coupling perception and action using minimax optimal control;a convergent recursive least squares policy iteration algorithm for multi-dimensional Markov Decision Process with continuous state and action spaces;basis function adaptation methods for cost approximation in MDP;and executing concurrent actions with multiple Markov Decision Processes.
ADPRL 2011 is the third ieee International symposium on Approximate dynamicprogramming and reinforcementlearning. The area of approximate dynamicprogramming and reinforcementlearning is a fusion of a number of res...
ADPRL 2011 is the third ieee International symposium on Approximate dynamicprogramming and reinforcementlearning. The area of approximate dynamicprogramming and reinforcementlearning is a fusion of a number of research areas in engineering, mathematics, artificial intelligence, operations research, and systems and control theory. This symposium brings together researchers from different disciplines and will provide a remarkable opportunity for the academic and industrial community to address new challenges, share innovative yet practical solutions, and define promising future research directions.
The proceedings contain 49 papers. The topics discussed include: fitted Q iteration with CMACs;reinforcement-learning-based magneto-hydrodynamic control hypersonic flows;a novel fuzzy reinforcementlearning approach i...
详细信息
ISBN:
(纸本)1424407060
The proceedings contain 49 papers. The topics discussed include: fitted Q iteration with CMACs;reinforcement-learning-based magneto-hydrodynamic control hypersonic flows;a novel fuzzy reinforcementlearning approach in two-level intelligent control of 3-DOF robot manipulators;knowledge transfer using local features;particle swarm optimization adaptivedynamicprogramming;discrete-time nonlinear HJB solution using approximation dynamicprogramming: convergence proof;dual representations for dynamicprogramming and reinforcementlearning;an optimal ADP algorithm for a high-dimensional stochastic control problem;convergence of model-based temporal difference learning for control;the effect of bootstrapping in multi-automata reinforcementlearning;and a theoretical analysis of cooperative behavior in multi-agent Q-learning.
This paper investigates optimal tracking control of linear stochastic systems with multiplicative state-dependent and input-dependent noise via a novel model-free integral reinforcementlearning algorithm. We have con...
详细信息
This paper investigates optimal tracking control of linear stochastic systems with multiplicative state-dependent and input-dependent noise via a novel model-free integral reinforcementlearning algorithm. We have conquered two major obstacles in this work. Firstly, the model-free tracking control on linear stochastic systems has scarcely been pivoted academically and it demands innovative methods and mathematical strategies to progress. An augmented stochastic differential equation of It & ocirc;'s type has been constructed while the control objective has been equated to minimization of the expected quadratic cost function. A model-dependent algorithm is shown to solve the stochastic algebraic Riccati equation. Second, control of linear stochastic systems essentially values for its sophisticated formation and sustained application. Inspired by optimal stationary control for linear stochastic systems, an integral reinforcementlearning algorithm utilizing the adaptivedynamicprogramming method has been developed to scrap the reliance on the complete knowledge of system dynamics. The convergence of the core matrix and the sequence of control policies and the system stability has been rigorously studied following. Finally, numerical simulations are performed to demonstrate the efficacy of the proposed integral reinforcementlearning methodology. Note to Practitioners-This paper focuses on the problem of optimal tracking control on linear stochastic systems, which is motivated by optimal stationary control and robust adaptivedynamicprogramming. Stochastic multiplicative noises are pervasive in modern control systems and engineering fields, such as power systems, aerospace systems, and industrial processes, thus important in modeling the random perturbation in system parameters and coefficients. However, research on tracking control on linear stochastic systems are challenging and languishing. First, we need to eradicate the dependence on complete knowledge system dy
Continuous-time reinforcementlearning (CT-RL) methods hold great promise in real-world applications. adaptivedynamicprogramming (ADP)-based CT-RL algorithms, especially their theoretical developments, have achieved...
详细信息
Continuous-time reinforcementlearning (CT-RL) methods hold great promise in real-world applications. adaptivedynamicprogramming (ADP)-based CT-RL algorithms, especially their theoretical developments, have achieved great successes. However, these methods have not been demonstrated for solving realistic or meaningful learning control problems. Thus, the goal of this work is to introduce a suite of new excitable integral reinforcementlearning (EIRL) algorithms for control of CT affine nonlinear systems. This work develops a new excitation framework to improve persistence of excitation (PE) and numerical performance via input/output insights from classical control. Furthermore, when the system dynamics afford a physically-motivated partition into distinct dynamical loops, the proposed methods break the control problem into smaller subproblems, resulting in reduced complexity. By leveraging the known affine nonlinear dynamics, the methods achieve well-behaved system responses and considerable data efficiency. The work provides convergence, solution optimality, and closed-loop stability guarantees of the proposed methods, and it demonstrates these guarantees on a significant application problem of controlling an unstable, nonminimum phase hypersonic vehicle (HSV).
暂无评论