With the rapid advance of information technology, network systems have become increasingly complex and hence the underlying system dynamics are often unknown or difficult to characterize. Finding a good network control policy is of significant importance to achieve desirable network performance (e.g., high throughput or low delay). In this work, we consider using model-based reinforcement learning (RL) to learn the optimal control policy for queueing networks so that the average job delay (or equivalently the average queue backlog) is minimized. Traditional approaches in RL, however, cannot handle the unbounded state spaces of the network control problem. To overcome this difficulty, we propose a new algorithm, called Reinforcement Learning for Queueing Networks (RL-QN), which applies model-based RL methods over a finite subset of the state space, while applying a known stabilizing policy for the rest of the states. We establish that the average queue backlog under RL-QN with an appropriately constructed subset can be arbitrarily close to the optimal result. We evaluate RL-QN in dynamic server allocation, routing and switching problems. Simulation results show that RL-QN minimizes the average queue backlog effectively.
Shenzhen Research Institute of Big Data（深圳大数据研究院）；The Chinese University of Hong Kong, Shenzhen（香港中文大学（深圳））；Cornell University（康奈尔大学）
Novel advanced policy gradient (APG) methods, such as Trust Region policy optimization and Proximal policy optimization (PPO), have become the dominant reinforcement learning algorithms because of their ease of implementation and good practical performance. A conventional setup for notoriously difficult queueing network control problems is a Markov decision problem (MDP) that has three features: infinite state space, unbounded costs, and long-run average cost objective. We extend the theoretical framework of these APG methods for such MDP problems. The resulting PPO algorithm is tested on a pa…
Queueing systems appear in many important real-life applications including communication networks, transportation and manufacturing systems. Reinforcement learning (RL) framework is a suitable model for the queueing control problem where the underlying dynamics are usually unknown and the agent receives little information from the environment to navigate. In this work, we investigate the optimization aspects of the queueing model as a RL environment and provide insight to learn the optimal policy efficiently. We propose a new parameterization of the policy by using the intrinsic properties of …
University of Manchester（曼彻斯特大学）；Stanford University（斯坦福大学）
We review the role of information and learning in the stability and optimization of queueing systems. In recent years, techniques from supervised learning, bandit learning and reinforcement learning have been applied to queueing systems supported by increasing role of information in decision making. We present observations and new results that help rationalize the application of these areas to queueing systems. We prove that the MaxWeight and BackPressure policies are an application of Blackwell's Approachability Theorem. This connects queueing theoretic results with adversarial learning. We t…
End-to-end delay is a critical attribute of quality of service (QoS) in application domains such as cloud computing and computer networks. This metric is particularly important in tandem service systems, where the end-to-end service is provided through a chain of services. Service-rate control is a common mechanism for providing QoS guarantees in service systems. In this paper, we introduce a reinforcement learning-based (RL-based) service-rate controller that provides probabilistic upper-bounds on the end-to-end delay of the system, while preventing the overuse of service resources. In order …
Carnegie Mellon University（卡内基梅隆大学）；University of Texas - Austin（德克萨斯大学奥斯汀分校）
In multi-server queueing systems where there is no central queue holding all incoming jobs, job dispatching policies are used to assign incoming jobs to the queue at one of the servers. Classic job dispatching policies such as join-the-shortest-queue and shortest expected delay assume that the service rates and queue lengths of the servers are known to the dispatcher. In this work, we tackle the problem of job dispatching without the knowledge of service rates and queue lengths, where the dispatcher can only obtain noisy estimates of the service rates by observing job departures. This problem …
Approximating Martingale Process (AMP) is proven to be effective for variance reduction in reinforcement learning (RL) in specific cases such as Multiclass Queueing Networks. However, in the already proven cases, the state space is relatively small and all possible state transitions can be iterated through. In this paper, we consider systems in which state space is large and have uncertainties when considering state transitions, thus making AMP a generalized variance-reduction method in RL. Specifically, we will investigate the application of AMP in ride-hailing systems like Uber, where Proxim…
We consider the problem of scheduling in constrained queueing networks with a view to minimizing packet delay. Modern communication systems are becoming increasingly complex, and are required to handle multiple types of traffic with widely varying characteristics such as arrival rates and service times. This, coupled with the need for rapid network deployment, render a bottom up approach of first characterizing the traffic and then devising an appropriate scheduling protocol infeasible. In contrast, we formulate a top down approach to scheduling where, given an unknown network and a set of sch…
Reinforcement Learning for Opportunistic Routing in Software-Defined LEO-Terrestrial Systems
School of Electrical and Mechanical Engineering, University of Adelaide（阿德莱德大学电气与机械工程学院）；Singapore University of Technology and Design（新加坡科技设计大学）；Electronics and Telecommunications Research Institute（电子电信研究院）
Indian Institute of Science（印度科学学院）；Technion（以色列理工学院）
We consider an improper reinforcement learning setting where a learner is given $M$ base controllers for an unknown Markov decision process, and wishes to combine them optimally to produce a potentially new controller that can outperform each of the base ones. This can be useful in tuning across controllers, learnt possibly in mismatched or simulated environments, to obtain a good controller for a given target environment with relatively few trials. \par We propose a gradient-based approach that operates over a class of improper mixtures of the controllers. We derive convergence rate guarantee…
Novel advanced policy gradient (APG) algorithms, such as proximal policy optimization (PPO), trust region policy optimization, and their variations, have become the dominant reinforcement learning (RL) algorithms because of their ease of implementation and good practical performance. This dissertation is concerned with theoretical justification and practical application of the APG algorithms for solving processing network control optimization problems. Processing network control problems are typically formulated as Markov decision process (MDP) or semi-Markov decision process (SMDP) problems t…
Indian Institute of Science（印度科学理工学院）；Purdue University（普渡大学）
Multicasting in wireless systems is a natural way to exploit the redundancy in user requests in a Content Centric Network. Power control and optimal scheduling can significantly improve the wireless multicast network's performance under fading. However, the model based approaches for power control and scheduling studied earlier are not scalable to large state space or changing system dynamics. In this paper, we use deep reinforcement learning where we use function approximation of the Q-function via a deep neural network to obtain a power control policy that matches the optimal policy for a sm…
Reinforcement Learning for Intensity Control: An Application to Choice-Based Network Revenue Management
Department of Systems Engineering and Engineering Management, The Chinese University of Hong Kong（香港中文大学系统工程与工程管理系）；Rotman School of Management, University of Toronto（多伦多大学罗特曼管理学院）
Autonomous navigation has recently gained great interest in the field of reinforcement learning. However, little attention was given to the time optimal velocity control problem, i.e. controlling a vehicle such that it travels at the maximal speed without becoming dynamically unstable (roll-over or sliding). Time optimal velocity control can be solved numerically using existing methods that are based on optimal control and vehicle dynamics. In this paper, we use deep reinforcement learning to generate the time optimal velocity control. Furthermore, we use the numerical solution to further impr…
Finding optimal policies which maximize long term rewards of Markov Decision Processes requires the use of dynamic programming and backward induction to solve the Bellman optimality equation. However, many real-world problems require optimization of an objective that is non-linear in cumulative rewards for which dynamic programming cannot be applied directly. For example, in a resource allocation problem, one of the objectives is to maximize long-term fairness among the users. We notice that when an agent aim to optimize some function of the sum of rewards is considered, the problem loses its …
We consider an improper reinforcement learning setting where a learner is given $M$ base controllers for an unknown Markov decision process, and wishes to combine them optimally to produce a potentially new controller that can outperform each of the base ones. This can be useful in tuning across controllers, learnt possibly in mismatched or simulated environments, to obtain a good controller for a given target environment with relatively few trials. Towards this, we propose two algorithms: (1) a Policy Gradient-based approach; and (2) an algorithm that can switch between a simple Actor-Critic …
University of Illinois at Urbana-Champaign（伊利诺伊大学厄巴纳-香槟分校）；Zhejiang University（浙江大学）
We propose a novel training algorithm for reinforcement learning which combines the strength of deep Q-learning with a constrained optimization approach to tighten optimality and encourage faster reward propagation. Our novel technique makes deep reinforcement learning more practical by drastically reducing the training time. We evaluate the performance of our approach on the 49 games of the challenging Arcade Learning Environment, and report significant improvements in both training time and accuracy.
We consider the problem of reinforcement learning (RL) with unbounded state space motivated by the classical problem of scheduling in a queueing network. Traditional policies as well as error metric that are designed for finite, bounded or compact state space, require infinite samples for providing any meaningful performance guarantee (e.g. $\ell_\infty$ error) for unbounded state space. That is, we need a new notion of performance metric. As the main contribution of this work, inspired by the literature in queuing systems and control theory, we propose stability as the notion of "goodness": t…
Transfer learning significantly accelerates the reinforcement learning process by exploiting relevant knowledge from previous experiences. The problem of optimally selecting source policies during the learning process is of great importance yet challenging. There has been little theoretical analysis of this problem. In this paper, we develop an optimal online method to select source policies for reinforcement learning. This method formulates online source policy selection as a multi-armed bandit problem and augments Q-learning with policy reuse. We provide theoretical guarantees of the optimal…
This paper studies the adaptive optimal stationary control of continuous-time linear stochastic systems with both additive and multiplicative noises, using reinforcement learning techniques. Based on policy iteration, a novel off-policy reinforcement learning algorithm, named optimistic least-squares-based policy iteration, is proposed which is able to find iteratively near-optimal policies of the adaptive optimal stationary control problem directly from input/state data without explicitly identifying any system matrices, starting from an initial admissible control policy. The solutions given …
We study the scheduling polices for asymptotically optimal delay in queueing systems with switching overhead. Such systems consist of a single server that serves multiple queues, and some capacity is lost whenever the server switches to serve a different set of queues. The capacity loss due to this switching overhead can be significant in many emerging applications, and needs to be explicitly addressed in the design of scheduling policies. For example, in 60GHz wireless networks with directional antennas, base stations need to train and reconfigure their beam patterns whenever they switch from…
We consider the age-of-information in a multi-class $M/G/1$ queueing system, where each class generates packets containing status information. Age of information is a relatively new metric that measures the amount of time that elapsed between status updates, thus accounting for both the queueing delay and the delay between packet generation. This gives rise to a tradeoff between frequency of status updates, and queueing delay. In this paper, we study this tradeoff in a system with heterogenous users modeled as a multi-class $M/G/1$ queue. To this end, we derive the exact peak age-of-Informatio…
Motivated by medium access control for resource-challenged wireless Internet of Things (IoT), we consider the problem of queue scheduling with reduced queue state information. In particular, we consider a time-slotted scheduling model with $N$ sensor nodes, with pair-wise dependence, such that Nodes $i$ and $i + 1,~0 < i < N$ cannot transmit together. We develop new throughput-optimal scheduling policies requiring only the empty-nonempty state of each queue that we term Queue Nonemptiness-Based (QNB) policies. We propose a Policy Splicing technique to combine scheduling policies for small netw…
In modern computer systems, jobs are divided into short tasks and executed in parallel. Empirical observations in practical systems suggest that the task service times are highly random and the job service time is bottlenecked by the slowest straggling task. One common solution for straggler mitigation is to replicate a task on multiple servers and wait for one replica of the task to finish early. The delay performance of replications depends heavily on the scheduling decisions of when to replicate, which servers to replicate on, and which job to serve first. So far, little is understood on ho…
In this paper, we study the problem of reducing the delay of downloading data from cloud storage systems by leveraging multiple parallel threads, assuming that the data has been encoded and stored in the clouds using fixed rate forward error correction (FEC) codes with parameters (n, k). That is, each file is divided into k equal-sized chunks, which are then expanded into n chunks such that any k chunks out of the n are sufficient to successfully restore the original file. The model can be depicted as a multiple-server queue with arrivals of data retrieving requests and a server corresponding …
A two-sided matching system is considered, where servers are assumed to arrive at a fixed rate, while the arrival rate of customers is modulated via a price-control mechanism. We analyse a loss model, wherein customers who are not served immediately upon arrival get blocked, as well as a queueing model, wherein customers wait in a queue until they receive service. The objective is to maximize the platform profit generated from matching servers and customers, subject to quality of service constraints, such as the expected wait time of servers in the loss system model, and the stability of the c…
The optimal tradeoff between average service cost rate, average utility rate, and average delay is addressed for a state dependent M/M/1 queueing model, with controllable queue length dependent service rates and arrival rates. For a model with a constant arrival rate $λ$ for all queue lengths, we obtain an asymptotic characterization of the minimum average delay, when the average service cost rate is a small positive quantity, $V$, more than the minimum average service cost rate required for queue stability. We show that depending on the value of the arrival rate $λ$, the assumed service cost …
In this thesis, we study the optimal tradeoff of average delay, average service cost, and average utility for single server queueing models, with and without admission control. The continuous time and discrete time queueing models that we consider are motivated by cross-layer models for noisy point-to-point links, with random packet arrivals. We study the above tradeoff problem for a class of admissible policies, which are monotone and stationary and obtain an asymptotic characterization of the minimum average delay as a function of the average service cost and average utility constraints.
Motivated by the increasing popularity of learning and predicting human user behavior in communication and computing systems, in this paper, we investigate the fundamental benefit of predictive scheduling, i.e., predicting and pre-serving arrivals, in controlled queueing systems. Based on a lookahead window prediction model, we first establish a novel equivalence between the predictive queueing system with a \emph{fully-efficient} scheduling scheme and an equivalent queueing system without prediction. This connection allows us to analytically demonstrate that predictive scheduling necessarily …
The presented study elaborates a multi-server catastrophic retrial queueing model considering preemptive repeat priority policy with phase-type (PH) distributed retrial times. For the sake of comprehension, the scenario of model operation prior and later to the occurrence of the disaster is referred to as the normal scenario and as the catastrophic scenario, respectively. In both scenarios, the arrival and service processes of all types of calls follow marked Markovian arrival process (MMAP) and PH distribution with distinct parameters, respectively. In the normal scenario, the incoming hetero…
We consider assignment policies that allocate resources to users, where both resources and users are located on a one-dimensional line. First, we consider unidirectional assignment policies that allocate resources only to users located to their left. We propose the Move to Right (MTR) policy, which scans from left to right assigning nearest rightmost available resource to a user, and contrast it to the Unidirectional Gale-Shapley (UGS) matching policy. While both policies among all unidirectional policies, minimize the expected distance traveled by a request (request distance), MTR is fairer. …
