We study reinforcement learning methods with linear function approximation under non-Markov state and cost processes. We first consider the policy evaluation method and show that the algorithm converges under suitable ergodicity conditions on the underlying non-Markov processes. Furthermore, we show that the limit corresponds to the fixed point of a joint operator composed of an orthogonal projection and the Bellman operator of an auxiliary Markov decision process. For Q-learning with linear function approximation, as in the Markov setting, convergence is not guaranteed in general. We show, however, that for the special case where the basis functions are chosen based on quantization maps, the convergence can be shown under similar ergodicity conditions. Finally, we apply our results to partially observed Markov decision processes, where finite-memory variables are used as state representations, and we derive explicit error bounds for the limits of the resulting learning algorithms.

Studying the stability of partially observed Markov decision processes (POMDPs) with respect to perturbations in either transition or observation kernels is a significant problem. While asymptotic robustness/stability results as approximate transition kernels and/or measurement kernels converge to the true ones have been previously reported, studies on explicit bounds on value differences and mismatch costs have been limited in scope for POMDPs. In this paper, we provide such explicit bounds under both discounted and average cost criteria. To this end, and also as an independent contribution, we first study the perturbations induced on the filter kernels (that is, the kernels of the belief-MDP reduction of POMDPs) as the transition and measurement kernels are perturbed. The bounds are given in terms of Wasserstein and total variation distances between the original and approximate transition and observation kernels. We then show that control policies optimized for approximate models yield performance guarantees when applied to the true model with explicit bounds. As a particular application, we consider the case where the state space and the measurement spaces are quantized to obtain finite models, and we obtain explicit error bounds which decay to zero as the approximations get finer. This provides explicit performance guarantees for model reduction in POMDPs.

The authors study an approximation method for partially observed Markov decision processes (POMDPs) with continuous spaces. Belief MDP reduction, which has been the standard approach to study POMDPs requires rigorous approximation methods for practical applications, due to the state space being lifted to the space of probability measures. Generalizing recent work, in this paper the authors present rigorous approximation methods via discretizing the observation space and constructing a fully observed finite MDP model using a finite length history of the discrete observations and control actions. The authors show that the resulting policy is near-optimal under some regularity assumptions on the channel, and under certain controlled filter stability requirements for the hidden state process. The authors also provide a Q learning algorithm that uses a finite memory of discretized information variables, and prove its convergence to the optimality equation of the finite fully observed MDP constructed using the approximation method.

We study a multi-agent mean field type control problem in discrete time where the agents aim to find a socially optimal strategy and where the state and action spaces for the agents are assumed to be continuous. The agents are only weakly coupled through the distribution of their state variables. The problem in its original form can be formulated as a classical Markov decision process (MDP), however, this formulation suffers from several practical difficulties. In this work, we attempt to overcome the curse of dimensionality, coordination complexity between the agents, and the necessity of perfect feedback collection from all the agents (which might be hard to do for large populations.) We provide several approximations: we establish the near optimality of the action and state space discretization of the agents under standard regularity assumptions for the considered formulation by constructing and studying the measure valued MDP counterpart for finite and infinite population settings. It is a well known approach to consider the infinite population problem for mean-field type models, since it provides symmetric policies for the agents which simplifies the coordination between the agents. However, the optimality analysis is harder as the state space of the measure valued infinite population MDP is continuous (even after space discretization of the agents). Therefore, as a final step, we provide further approximations for the infinite population problem by focusing on smaller sized sub-population distributions.

The paper focuses on mean-field type multi-agent control problems with finite state and action spaces where the dynamics and cost structures are symmetric and homogeneous, and are affected by the distribution of the agents. A standard solution method for these problems is to consider the infinite population limit as an approximation and use symmetric solutions of the limit problem to achieve near optimality. The control policies, and in particular the dynamics, depend on the population distribution in the finite population setting, or the marginal distribution of the state variable of a representative agent for the infinite population setting. Hence, learning and planning for these control problems generally require estimating the reaction of the system to all possible state distributions of the agents. To overcome this issue, we consider linear function approximation for the control problem and provide coordinated and independent learning methods. We rigorously establish error upper bounds for the performance of learned solutions. The performance gap stems from (i) the mismatch due to estimating the true model with a linear one, and (ii) using the infinite population solution in the finite population problem as an approximate control. The provided upper bounds quantify the impact of these error sources on the overall performance.

This paper investigates robustness and stability properties to incorrect transition and measurement kernel models and their approximations in partially observable Markov decision processes (POMDPs). In particular, we analyze the sensitivity of filter kernels to simultaneous perturbations in both the transition and observation kernels, and provide explicit quantitative bounds in terms of total variation and Wasserstein metrics. As a result, we obtain robustness and stability bounds for optimal stochastic control in POMDPs under discounted and average cost criteria.

We study reinforcement learning with linear function approximation and finite-memory approximations for partially observed Markov decision processes (POMDPs). We first present an algorithm for the value evaluation of finite-memory feedback policies. We provide error bounds derived from filter stability and projection errors. We then study the learning of finite-memory based near-optimal Q values. Convergence in this case requires further assumptions on the exploration policy when using general basis functions. We then show that these assumptions can be relaxed for specific models such as those with perfectly linear cost and dynamics, or when using discretization based basis functions.

We obtain conditions for the Wasserstein regularity of non-linear filters, which serve as reduced MDP models for Partially Observable Markov Decision Processes (POMDPs). In particular, we establish a new set of assumptions: instead of requiring total variation continuity of the transition kernel as in earlier work, we show that Wasserstein continuity of the transition kernel, together with total variation continuity of the observation channel, is sufficient to ensure key results. These conditions lead to: (i) unique ergodicity and (ii) geometric ergodicity of the control-free non-linear filter kernel, as well as (iii) existence of optimal policies and (iv) their rigorous approximations and Q-learning methods for both discounted and average cost optimal stochastic control problems. We further identify complementary conditions under which finite-window policies remain near-optimal and provide refined error bounds. Our framework unifies and extends prior results while offering verifiable conditions for analysis and design in POMDPs.

Finding optimal policies for Partially Observable Markov Decision Processes (POMDPs) is challenging due to their uncountable state spaces when transformed into fully observable Markov Decision Processes (MDPs) using belief states. Traditional methods such as dynamic programming or policy iteration are difficult to apply in this context, necessitating the use of approximation methods on belief states or other techniques. Recently, in (Journal of Machine Learning Research, vol. 23, pp. 1-46, 2022) and (Mathematics of Operations Research, vol. 48, pp. 2066-2093, Nov. 2023), it was shown that sliding finite window based policies are near-optimal for POMDPs with standard Borel valued hidden state spaces, and can be learned via reinforcement learning, with error bounds explicitly dependent on a uniform filter stability term involving total variation in expectation and sample path-wise, respectively. In this paper, we refine these performance bounds and (i) extend them to bounds via uniform filter stability in expected Wasserstein distance leading to an error bound in expectation, and (ii) complementary conditions bounds via uniform filter stability in sample path-wise total variation distance leading to a uniform error bound. We present explicit examples. Our approach thus provides complementary and more refined bounds on the error terms in both total variation and Wasserstein metrics, offering more relaxed and stronger bounds over the approximation error in POMDP solutions on the performance and near optimality of sliding finite window control policies.

We study mean-field control problems in discrete-time under the infinite horizon average cost optimality criteria. We focus on both the finite population and the infinite population setups. We show the existence of the average cost optimality equation (ACOE) and the existence of optimal stationary Markov policies for finite population problems under (i) a minorization condition that provides geometric ergodicity on the collective state process of the agents, and (ii) under standard Lipschitz continuity assumptions on the stage-wise cost and transition function of the agents when the Lipschitz constant of the transition function satisfies a certain bound. Furthermore, we provide a relative value iteration algorithm that approximates the optimal solution of the problem. For the infinite population problem, we establish the existence of the ACOE, and the existence of optimal policies under the continuity assumptions on the cost and the transition functions, and we provide a convergent relative value iteration method for the infinite population problem as well. Finally, we prove that the optimal value of the finite population problem converges to the optimal value of the infinite population problem as the number of agents grows to infinity. This result has been proven for finite horizon and infinite horizon discounted cost problems, however, it is the first result for the average cost optimality criteria to our knowledge.

As a primary contribution, we present a convergence theorem for stochastic iterations, and in particular, Q-learning iterates, under a general, possibly non-Markovian, stochastic environment. Our conditions for convergence involve an ergodicity and a positivity criterion. We provide a precise characterization on the limit of the iterates and conditions on the environment and initializations for convergence. As our second contribution, we discuss the implications and applications of this theorem to a variety of stochastic control problems with non-Markovian environments involving (i) quantized approximations of fully observed Markov Decision Processes (MDPs) with continuous spaces (where quantization break down the Markovian structure), (ii) quantized approximations of belief-MDP reduced partially observable MDPS (POMDPs) with weak Feller continuity and a mild version of filter stability (which requires the knowledge of the model by the controller), (iii) finite window approximations of POMDPs under a uniform controlled filter stability (which does not require the knowledge of the model), and (iv) for multi-agent models where convergence of learning dynamics to a new class of equilibria, subjective Q-learning equilibria, will be studied. In addition to the convergence theorem, some implications of the theorem above are new to the literature and others are interpreted as applications of the convergence theorem. Some open problems are noted.

In this review/tutorial article, we present recent progress on optimal control of partially observed Markov Decision Processes (POMDPs). We first present regularity and continuity conditions for POMDPs and their belief-MDP reductions, where these constitute weak Feller and Wasserstein regularity and controlled filter stability. These are then utilized to arrive at existence results on optimal policies for both discounted and average cost problems, and regularity of value functions. Then, we study rigorous approximation results involving quantization based finite model approximations as well as finite window approximations under controlled filter stability. Finally, we present several recent reinforcement learning theoretic results which rigorously establish convergence to near optimality under both criteria.

The average cost optimality is known to be a challenging problem for partially observable stochastic control, with few results available beyond the finite state, action, and measurement setup, for which somewhat restrictive conditions are available. In this paper, we present explicit and easily testable conditions for the existence of solutions to the average cost optimality equation where the state space is compact. In particular, we present a new contraction based analysis, which is new to the literature to our knowledge, building on recent regularity results for non-linear filters. Beyond establishing existence, we also present several implications of our analysis that are new to the literature: (i) robustness to incorrect priors (ii) near optimality of policies based on quantized approximations, (iii) near optimality of policies with finite memory, and (iv) convergence in Q-learning. In addition to our main theorem, each of these represents a novel contribution for average cost criteria.

Reinforcement learning algorithms often require finiteness of state and action spaces in Markov decision processes (MDPs) (also called controlled Markov chains) and various efforts have been made in the literature towards the applicability of such algorithms for continuous state and action spaces. In this paper, we show that under very mild regularity conditions (in particular, involving only weak continuity of the transition kernel of an MDP), Q-learning for standard Borel MDPs via quantization of states and actions (called Quantized Q-Learning) converges to a limit, and furthermore this limit satisfies an optimality equation which leads to near optimality with either explicit performance bounds or which are guaranteed to be asymptotically optimal. Our approach builds on (i) viewing quantization as a measurement kernel and thus a quantized MDP as a partially observed Markov decision process (POMDP), (ii) utilizing near optimality and convergence results of Q-learning for POMDPs, and (iii) finally, near-optimality of finite state model approximations for MDPs with weakly continuous kernels which we show to correspond to the fixed point of the constructed POMDP. Thus, our paper presents a very general convergence and approximation result for the applicability of Q-learning for continuous MDPs.

We study a Q learning algorithm for continuous time stochastic control problems. The proposed algorithm uses the sampled state process by discretizing the state and control action spaces under piece-wise constant control processes. We show that the algorithm converges to the optimality equation of a finite Markov decision process (MDP). Using this MDP model, we provide an upper bound for the approximation error for the optimal value function of the continuous time control problem. Furthermore, we present provable upper-bounds for the performance loss of the learned control process compared to the optimal admissible control process of the original problem. The provided error upper-bounds are functions of the time and space discretization parameters, and they reveal the effect of different levels of the approximation: (i) approximation of the continuous time control problem by an MDP, (ii) use of piece-wise constant control processes, (iii) space discretization. Finally, we state a time complexity bound for the proposed algorithm as a function of the time and space discretization parameters.

For infinite-horizon average-cost criterion problems, there exist relatively few rigorous approximation and reinforcement learning results. In this paper, for such problems, we present several approximation and reinforcement learning results for Markov Decision Processes with standard Borel spaces. Toward this end, (i) we first provide a discretization based approximation method for fully observed Markov Decision Processes (MDPs) with continuous spaces under average cost criteria, and we provide error bounds for the approximations when the dynamics are only weakly continuous (for asymptotic convergence of errors as the grid sizes vanish) or Wasserstein continuous (with a rate in approximation as the grid sizes vanish) under certain ergodicity assumptions. In particular, we relax the total variation condition given in prior work to weak continuity as well as Wasserstein continuity conditions. (ii) We provide synchronous and asynchronous Q-learning algorithms for continuous spaces via quantization (where the quantized state is taken to be the actual state in corresponding Q-learning algorithms presented in the paper), and establish their convergence; for the former we utilize a span semi-norm approach and for the latter we use a direct contraction approach. (iii) We finally show that the convergence is to the optimal Q values of the finite approximate models constructed via quantization, which implies near optimality of the arrived solution. Our Q-learning convergence results and their convergence to near optimality are new for continuous spaces, and the proof method is new even for finite spaces, to our knowledge.

In this paper, for partially observed Markov decision problems (POMDPs), we provide the convergence of a Q learning algorithm for control policies using a finite history of past observations and control actions, and consequentially, we establish near optimality of such limit Q functions under explicit filter stability conditions. We present explicit error bounds relating the approximation error to the length of the finite history window. We establish the convergence of such Q learning iterations under mild ergodicity assumptions on the state process during the exploration phase. We further show that the limit fixed point equation gives an optimal solution for an approximate belief Markov decision problem (MDP). We then provide bounds on the performance of the policy obtained using the limit Q values compared with the performance of the optimal policy for the POMDP, in which we also present explicit conditions using recent results on filter stability in controlled POMDPs. Whereas there exist many experimental results, (i) the rigorous asymptotic convergence (to an approximate MDP value function) for such finite memory Q learning algorithms and (ii) the near optimality with an explicit rate of convergence (in the memory size) under filter stability are results that are new to the literature to our knowledge. Funding: This research was supported in part by the Natural Sciences and Engineering Research Council (NSERC) of Canada.

We study continuity and robustness properties of infinite-horizon average expected cost problems with respect to (controlled) transition kernels, and applications of these results to the problem of robustness of control policies designed for approximate models applied to actual systems. We show that sufficient conditions presented in the literature for discounted-cost problems are in general not sufficient to ensure robustness for average-cost problems. However, we show that the average optimal cost is continuous in the convergences of controlled transition kernel models where convergence of models entails (i) continuous weak convergence in state and actions, and (ii) continuous setwise convergence in the actions for every fixed state variable, in addition to either uniform ergodicity or some regularity conditions. We establish that the mismatch error due to the application of a control policy designed for an incorrectly estimated model to the true model decreases to zero as the incorrect model approaches the true model under the stated convergence criteria. Our findings significantly relax related studies in the literature which have primarily considered the more restrictive total variation convergence criteria. Applications to robustness to models estimated through empirical data (where almost sure weak convergence criterion typically holds, but stronger criteria do not) are studied and conditions for asymptotic robustness to data-driven learning are established.

In the theory of Partially Observed Markov Decision Processes (POMDPs), existence of optimal policies have in general been established via converting the original partially observed stochastic control problem to a fully observed one on the belief space, leading to a belief-MDP. However, computing an optimal policy for this fully observed model, and so for the original POMDP, using classical dynamic or linear programming methods is challenging even if the original system has finite state and action spaces, since the state space of the fully observed belief-MDP model is always uncountable. Furthermore, there exist very few rigorous value function approximation and optimal policy approximation results, as regularity conditions needed often require a tedious study involving the spaces of probability measures leading to properties such as Feller continuity. In this paper, we study a planning problem for POMDPs where the system dynamics and measurement channel model are assumed to be known. We construct an approximate belief model by discretizing the belief space using only finite window information variables. We then find optimal policies for the approximate model and we rigorously establish near optimality of the constructed finite window control policies in POMDPs under mild non-linear filter stability conditions and the assumption that the measurement and action sets are finite (and the state space is real vector valued). We also establish a rate of convergence result which relates the finite window memory size and the approximation error bound, where the rate of convergence is exponential under explicit and testable exponential filter stability conditions. While there exist many experimental results and few rigorous asymptotic convergence results, an explicit rate of convergence result is new in the literature, to our knowledge.

We study an approximation method for partially observed Markov decision processes (POMDPs) with continuous spaces. Belief MDP reduction has been the standard approach to study POMDPs, which, due to its uncountable state space and strict regularity properties however, requires rigorous approximation methods for practical applications. In this work, we focus on an approximation procedure via discretizing the observation space and constructing a fully observed finite MDP model using a finite length history of the discrete observations and control actions. We show that the resulting policy is nearly optimal under some regularity assumptions on the channel, and under certain controlled filter stability requirements for the hidden state process. We also provide a Q learning algorithm that uses a finite memory of discretized information variables, and prove its convergence to the optimality equation of the finite fully observed MDP constructed using the approximation method.