Value set iteration for Markov decision processes

This communique presents an algorithm called “value set iteration” (VSI) for solving infinite horizon discounted Markov decision processes with finite state and action spaces as a simple generalization of value iteration (VI) and as a counterpart to Chang’s policy set iteration. A sequence of value functions is generated by VSI based on manipulating a set of value functions at each iteration and it converges to the optimal value function. VSI preserves convergence properties of VI while converging no slower than VI and in particular, if the set used in VSI contains the value functions of independently generated sample-policies from a given distribution and a properly defined policy switching policy, a probabilistic exponential convergence rate of VSI can be established. Because the set used in VSI can contain the value functions of any policies generated by other existing algorithms, VSI is also a general framework of combining multiple solution methods.     

An exact iterative search algorithm for constrained Markov decision processes

This communique provides an exact iterative search algorithm for the NP-hard problem of obtaining an optimal feasible stationary Markovian pure policy that achieves the maximum value averaged over an initial state distribution in finite constrained Markov decision processes. It is based on a novel characterization of the entire feasible policy space and takes the spirit of policy iteration (PI) in that a sequence of monotonically improving feasible policies is generated and converges to an optimal policy in iterations of the size of the policy space at the worst case. Unlike PI, an unconstrained MDP needs to be solved at iterations involved with feasible policies and the current best policy improves all feasible policies included in the union of the policy spaces associated with the unconstrained MDPs.     

A time aggregation approach to Markov decision processes

We propose a time aggregation approach for the solution of infinite horizon average cost Markov decision processes via policy iteration. In this approach, policy update is only carried out when the process visits a subset of the state space. As in state aggregation, this approach leads to a reduced state space, which may lead to a substantial reduction in computational and storage requirements, especially for problems with certain structural properties. However, in contrast to state aggregation, which generally results in an approximate model due to the loss of Markov property, time aggregation suffers no loss of accuracy, because the Markov property is preserved. Single sample path-based estimation algorithms are developed that allow the time aggregation approach to be implemented on-line for practical systems. Some numerical and simulation examples are presented to illustrate the ideas and potential computational savings.     

Actor-critic algorithms for hierarchical Markov decision processes

We consider the problem of control of hierarchical Markov decision processes and develop a simulation based two-timescale actor-critic algorithm in a general framework. We also develop certain approximation algorithms that require less computation and satisfy a performance bound. One of the approximation algorithms is a three-timescale actor-critic algorithm while the other is a two-timescale algorithm, however, which operates in two separate stages. All our algorithms recursively update randomized policies using the simultaneous perturbation stochastic approximation (SPSA) methodology. We briefly present the convergence analysis of our algorithms. We then present numerical experiments on a problem of production planning in semiconductor fabs on which we compare the performance of all algorithms together with policy iteration. Algorithms based on certain Hadamard matrix based deterministic perturbations are found to show the best results.     

Compositional reasoning for weighted Markov decision processes

Weighted Markov decision processes (MDPs) have long been used to model quantitative aspects of systems in the presence of uncertainty. However, much of the literature on such MDPs takes a monolithic approach, by modelling a system as a particular MDP; properties of the system are then inferred by analysis of that particular MDP. In contrast in this paper we develop compositional methods for reasoning about weighted MDPs, as a possible basis for compositional reasoning about their quantitative behaviour. In particular we approach these systems from a process algebraic point of view. For these we define a coinductive simulation-based behavioural preorder which is compositional in the sense that it is preserved by structural operators for constructing weighted MDPs from components.     

Probabilistic opacity for Markov decision processes

Opacity is a generic security property, that has been defined on (non-probabilistic) transition systems and later on Markov chains with labels. For a secret predicate, given as a subset of runs, and a function describing the view of an external observer, the value of interest for opacity is a measure of the set of runs disclosing the secret. We extend this definition to the richer framework of Markov decision processes, where non-deterministic choice is combined with probabilistic transitions, and we study related decidability problems with partial or complete observation hypotheses for the schedulers. We prove that all questions are decidable with complete observation and ω-regular secrets. With partial observation, we prove that all quantitative questions are undecidable but the question whether a system is almost surely non-opaque becomes decidable for a restricted class of ω-regular secrets, as well as for all ω-regular secrets under finite-memory schedulers.     

The control of a two-level Markov decision process by time aggregation

The solution of Markov Decision Processes (MDPs) often relies on special properties of the processes. For two-level MDPs, the difference in the rates of state changes of the upper and lower levels has led to limiting or approximate solutions of such problems. In this paper, we solve a two-level MDP without making any assumption on the rates of state changes of the two levels. We first show that such a two-level MDP is a non-standard one where the optimal actions of different states can be related to each other. Then we give assumptions (conditions) under which such a specially constrained MDP can be solved by policy iteration. We further show that the computational effort can be reduced by decomposing the MDP. A two-level MDP with M upper-level states can be decomposed into one MDP for the upper level and M to M(M-1) MDPs for the lower level, depending on the structure of the two-level MDP. The upper-level MDP is solved by time aggregation, a technique introduced in a recent paper [Cao, X.-R., Ren, Z. Y., Bhatnagar, S., Fu, M., & Marcus, S. (2002). A time aggregation approach to Markov decision processes. Automatica, 38(6), 929-943.], and the lower-level MDPs are solved by embedded Markov chains.     

Performance sensitivities for parameterized Markov systems

It is known that the performance potentials (or equivalently, perturbation realization factors) can be used as building blocks for performance sensitivities of Markov systems. In parameterized systerns, the changes in parameters may only affect some states, and the explicit transition probability matrix may not be known. In this paper, we use an example to show that we can use potentials to construct performance sensitivities m a more flexible way; only the potentials at the affected states need to be estimated, and the transition probability matrix need not be known. Policy iteration algorithms, which are simpler than the standard one, can be established.     

An actor-critic algorithm for constrained Markov decision processes

An actor-critic type reinforcement learning algorithm is proposed and analyzed for constrained controlled Markov decision processes. The analysis uses multiscale stochastic approximation theory and the envelope theorem' of mathematical economics.     

Exact finite approximations of average-cost countable Markov decision processes

For a countable-state Markov decision process we introduce an embedding which produces a finite-state Markov decision process. The finite-state embedded process has the same optimal cost, and moreover, it has the same dynamics as the original process when restricting to the approximating set. The embedded process can be used as an approximation which, being finite, is more convenient for computation and implementation.     

Denumerable continuous-time Markov decision processes with multiconstraints on average costs

This article deals with multiconstrained continuous-time Markov decision processes in a denumerable state space, with unbounded cost and transition rates. The criterion to be optimised is the long-run expected average cost, and several kinds of constraints are imposed on some associated costs. The existence of a constrained optimal policy is ensured under suitable conditions by using a martingale technique and introducing an occupation measure. Furthermore, for the unichain model, we transform this multiconstrained problem into an equivalent linear programming problem, then construct a constrained optimal policy from an optimal solution to the linear programming. Finally, we use an example of a controlled queueing system to illustrate an application of our results.     

Policy iteration type algorithms for recurrent state Markov decision processes

We introduce and analyze several new policy iteration type algorithms for average cost Markov decision processes (MDPs). We limit attention to “recurrent state” processes where there exists a state which is recurrent under all stationary policies, and our analysis applies to finite-state problems with compact constraint sets, continuous transition probability functions, and lower-semicontinuous cost functions. The analysis makes use of an underlying relationship between recurrent state MDPs and the so-called stochastic shortest path problems of Bertsekas and Tsitsiklis (Math. Oper. Res. 16(3) (1991) 580). After extending this relationship, we establish the convergence of the new policy iteration type algorithms either to optimality or to within >0 of the optimal average cost.     

Numerical analysis of continuous time Markov decision processes over finite horizons

Continuous time Markov decision processes (CTMDPs) with a finite state and action space have been considered for a long time. It is known that under fairly general conditions the reward gained over a finite horizon can be maximized by a so-called piecewise constant policy which changes only finitely often in a finite interval. Although this result is available for more than 30 years, numerical analysis approaches to compute the optimal policy and reward are restricted to discretization methods which are known to converge to the true solution if the discretization step goes to zero. In this paper, we present a new method that is based on uniformization of the CTMDP and allows one to compute an ε-optimal$\epsilon \text{-optimal}$ policy up to a predefined precision in a numerically stable way using adaptive time steps.     

Strong 1-optimal stationary policies in denumerable Markov decision processes

Usual conditions for existence of stationary average optimal policies in denumerable MDPs with general bounded rewards are shown to be also sufficient for strong 1-optimality. Moreover, we prove that all limit points of discounted optimal stationary policies when the discount factor goes to 1 are strong 1-optimal.     

Simulation-based optimization of Markov decision processes: An empirical process theory approach

We generalize and build on the PAC Learning framework for Markov Decision Processes developed in Jain and Varaiya (2006). We consider the reward function to depend on both the state and the action. Both the state and action spaces can potentially be countably infinite. We obtain an estimate for the value function of a Markov decision process, which assigns to each policy its expected discounted reward. This expected reward can be estimated as the empirical average of the reward over many independent simulation runs. We derive bounds on the number of runs needed for the convergence of the empirical average to the expected reward uniformly for a class of policies, in terms of the V-C or pseudo dimension of the policy class. We then propose a framework to obtain an ?-optimal policy from simulation. We provide sample complexity of such an approach.     

New prioritized value iteration for Markov decision processes

The problem of solving large Markov decision processes accurately and quickly is challenging. Since the computational effort incurred is considerable, current research focuses on finding superior acceleration techniques. For instance, the convergence properties of current solution methods depend, to a great extent, on the order of backup operations. On one hand, algorithms such as topological sorting are able to find good orderings but their overhead is usually high. On the other hand, shortest path methods, such as Dijkstra’s algorithm which is based on priority queues, have been applied successfully to the solution of deterministic shortest-path Markov decision processes. Here, we propose an improved value iteration algorithm based on Dijkstra’s algorithm for solving shortest path Markov decision processes. The experimental results on a stochastic shortest-path problem show the feasibility of our approach.     

Markov decision processes with iterated coherent risk measures

This paper considers a Markov decision process in Borel state and action spaces with the aggregated (or say iterated) coherent risk measure to be minimised. For this problem, we establish the Bellman optimality equation as well as the value and policy iteration algorithms, and show the existence of a deterministic stationary optimal policy. The cost function, while being allowed to be unbounded from below (in the sense that its negative part needs be bounded by some nonnegative real-valued possibly unbounded weight function), can be arbitrarily unbounded from above and possibly infinitely valued.     

Evolutionary policy iteration for solving Markov decision processes

We propose a novel algorithm called evolutionary policy iteration (EPI) for solving infinite horizon discounted reward Markov decision processes. EPI inherits the spirit of policy iteration but eliminates the need to maximize over the entire action space in the policy improvement step, so it should be most effective for problems with very large action spaces. EPI iteratively generates a "population" or a set of policies such that the performance of the "elite policy" for a population monotonically improves with respect to a defined fitness function. EPI converges with probability one to a population whose elite policy is an optimal policy. EPI is naturally parallelizable and along this discussion, a distributed variant of PI is also studied.     

Potential-based online policy iteration algorithms for Markov decision processes

Performance potentials play a crucial role in performance sensitivity analysis and policy iteration of Markov decision processes. The potentials can be estimated on a single sample path of a Markov process. In this paper, we propose two potential-based online policy iteration algorithms for performance optimization of Markov systems. The algorithms are based on online estimation of potentials and stochastic approximation. We prove that with these two algorithms the optimal policy can be attained after a finite number of iterations. A simulation example is given to illustrate the main ideas and the convergence rates of the algorithms.     

Policy iteration for customer-average performance optimization of closed queueing systems

We consider the optimization of queueing systems with service rates depending on system states. The optimization criterion is the long-run customer-average performance, which is an important performance metric, different from the traditional time-average performance. We first establish, with perturbation analysis, a difference equation of the customer-average performance in closed networks with exponentially distributed service times and state-dependent service rates. Then we propose a policy iteration optimization algorithm based on this difference equation. This algorithm can be implemented on-line with a single sample path and does not require knowing the routing probabilities of queueing systems. Finally, we give numerical experiments which demonstrate the efficiency of our algorithm. This paper gives a new direction to efficiently optimize the “customer-centric” performance in queueing systems.