In which we try to give a basic intuitive sense of what reinforcement learning is and how it differs and relates to other fields, e.g., supervised learning and neural networks, genetic algorithms and artificial life, control theory. Intuitively, RL is trial and error (variation and selection, search) plus learning (association, memory). We argue that RL is the only field that seriously addresses the special features of the problem of learning from interaction to achieve long-term goals.

In this paper, we propose and analyze a class of actor-critic algorithms. These are two-time-scale algorithms in which the critic uses temporal difference (TD) learning with a linearly parameterized approximation architecture, and the actor is updated in an approximate gradient direction based on information provided by the critic. We show that the features for the critic should ideally span a subspace prescribed by the choice of parameterization of the actor. We study actorcritic algorithms for Markov decision processes with general state and action spaces. We state and prove two results regarding their convergence.

A preliminary unedited version of this paper was incorrectly published as part of Volume

The curse of dimensionality gives rise to prohibitive computational requirements that render infeasible the exact solution of large-scale stochastic control problems. We study an efficient method based on linear programming for approximating solutions to such problems. The approach “fits ” a linear combination of pre-selected basis functions to the dynamic programming cost-to-go function. We develop error bounds that offer performance guarantees and also guide the selection of both basis functions and “state-relevance weights ” that influence quality of the approximation. Experimental results in the domain of queueing network control provide empirical support for the methodology. (Dynamic programming/optimal control: approximations/large-scale problems. Queues, algorithms: control of queueing networks.)

This paper presents a reinforcement learning framework for continuoustime dynamical systems without a priori discretization of time, state, and action. Based on the Hamilton-Jacobi-Bellman (HJB) equation for infinitehorizon, discounted reward problems, we derive algorithms for estimating value functions and for improving policies with the use of function approximators. The process of value function estimation is formulated as the minimization of a continuous-time form of the temporal difference (TD) error. Update methods based on backward Euler approximation and exponential eligibility traces are derived and their correspondences with the conventional residual gradient, TD(0), and TD() algorithms are shown. For policy improvement, two methods, namely, a continuous actor-critic method and a value-gradient based greedy policy, are formulated. As a special case of the latter, a nonlinear feedback control law using the value gradient and the model of the input gain is derived....

This paper addresses the problem of planning under uncertainty in large Markov Decision Processes (MDPs). Factored MDPs represent a complex state space using state variables and the transition model using a dynamic Bayesian network. This representation often allows an exponential reduction in the representation size of structured MDPs, but the complexity of exact solution algorithms for such MDPs can grow exponentially in the representation size. In this paper, we present two approximate solution algorithms that exploit structure in factored MDPs. Both use an approximate value function represented as a linear combination of basis functions, where each basis function involves only a small subset of the domain variables. A key contribution of this paper is that it shows how the basic operations of both algorithms can be performed efficiently in closed form, by exploiting both additive and context-specific structure in a factored MDP. A central element of our algorithms is a novel linear program decomposition technique, analogous to variable elimination in Bayesian networks, which reduces an exponentially large LP to a provably equivalent, polynomial-sized one. One algorithm uses approximate linear programming, and the second approximate dynamic programming. Our dynamic programming algorithm is novel in that it uses an approximation based on max-norm, a technique that more directly minimizes the terms that appear in error bounds for approximate MDP algorithms. We provide experimental results on problems with over 10^40 states, demonstrating a promising indication of the scalability of our approach, and compare our algorithm to an existing state-of-the-art approach, showing, in some problems, exponential gains in computation time.

An important application of reinforcement learning (RL) is to finite-state control problems and one of the most difficult problems in learning for control is balancing the exploration/exploitation tradeoff. Existing theoretical results for RL give very little guidance on reasonable ways to perform exploration. In this paper, we examine the convergence of single-step on-policy RL algorithms for control. On-policy algorithms cannot separate exploration from learning and therefore must confront the exploration problem directly. We prove convergence results for several related on-policy algorithms with both decaying exploration and persistent exploration. We also provide examples of exploration strategies that can be followed during learning that result in convergence to both optimal values and optimal policies.

1 RoboCup simulated soccer presents many challenges to reinforcement learning methods, in-cluding a large state space, hidden and uncertain state, multiple independent agents learning simultaneously, and long and variable delays in the effects of actions. We describe our appli-cation of episodic SMDP Sarsa(λ) with linear tile-coding function approximation and variable λ to learning higher-level decisions in a keepaway subtask of RoboCup soccer. In keepaway, one team, “the keepers, ” tries to keep control of the ball for as long as possible despite the efforts of “the takers. ” The keepers learn individually when to hold the ball and when to pass to a teammate. Our agents learned policies that significantly outperform a range of benchmark policies. We demonstrate the generality of our approach by applying it to a number of task variations including different field sizes and different numbers of players on each team.

Abstract. TD(λ) is a popular family of algorithms for approximate policy evaluation in large MDPs. TD(λ) works by incrementally updating the value function after each observed transition. It has two major drawbacks: it may make inefficient use of data, and it requires the user to manually tune a stepsize schedule for good performance. For the case of linear value function approximations and λ = 0, the Least-Squares TD (LSTD) algorithm of Bradtke and Barto (1996, Machine learning, 22:1–3, 33–57) eliminates all stepsize parameters and improves data efficiency. This paper updates Bradtke and Barto’s work in three significant ways. First, it presents a simpler derivation of the LSTD algorithm. Second, it generalizes from λ = 0 to arbitrary values of λ; at the extreme of λ = 1, the resulting new algorithm is shown to be a practical, incremental formulation of supervised linear regression. Third, it presents a novel and intuitive interpretation of LSTD as a model-based reinforcement learning technique.

TD() is a popular family of algorithms for approximate policy evaluation in large MDPs. TD() works by incrementally updating the value function after each observed transition. It has two major drawbacks: it makes inefficient use of data, and it requires the user to manually tune a stepsize schedule for good performance. For the case of linear value function approximations and = 0, the Least-Squares TD (LSTD) algorithm of Bradtke and Barto (Bradtke and Barto, 1996) eliminates all stepsize parameters and improves data efficiency. This paper extends Bradtke and Barto's work in three significant ways. First, it presents a simpler derivation of the LSTD algorithm. Second, it generalizes from = 0 to arbitrary values of ; at the extreme of = 1, the resulting algorithm is shown to be a practical formulation of supervised linear regression. Third, it presents a novel, intuitive interpretation of LSTD as a model-based reinforcement learning technique.