Blind Dynamic Resource Allocation in Closed Networks via Mirror Backpressure

Kanoria, Yash; Qian, Pengyu

doi:10.48550/arxiv.1903.02764

Cited by 5 publications

(1 citation statement)

References 54 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…An intriguing fluctuation bound for MaxWeight is found in Sharifnassab et al (2020). More recent literature on MaxWeight and BackPressure in queueing networks has focused closed systems, such as using it as a decision rule in ride sharing platforms (see Banerjee et al (2018) and Kanoria and Qian (2019)). Another application concerns road traffic signal control (Le et al 2015, Varaiya 2013, Xiao et al 2015, which we will discuss in some more detail in Section 5.…”

Section: Literature Reviewmentioning

confidence: 99%

Learning and Information in Stochastic Networks and Queues

Walton¹,

Xu²

2021

Preprint

View full text Add to dashboard Cite

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 then discuss the requirements of statistical learning for service parameter estimation. As an example, we show how queue size regret can be bounded when applying a perceptron algorithm to classify service. Next, we discuss the role of state information in improved decision making. Here we contrast the roles of epistemic information (information on uncertain parameters) and aleatoric information (information on an uncertain state). Finally we review recent advances in the theory of reinforcement learning and queueing, as well as, provide discussion on current research challenges.

show abstract

Section: Literature Reviewmentioning

confidence: 99%

Learning and Information in Stochastic Networks and Queues

Walton¹,

Xu²

2021

Preprint

View full text Add to dashboard Cite

show abstract

Dynamic Pricing and Matching for Two-Sided Queues

Varma

Bumpensanti

Maguluri

et al. 2020

Proc. ACM Meas. Anal. Comput. Syst.

View full text Add to dashboard Cite

Real-time Approximate Routing for Smart Transit Systems

Banerjee

Hssaine

Périvier

et al. 2021

SIGMETRICS Perform. Eval. Rev.

View full text Add to dashboard Cite

The advent of ride-hailing platforms such as Lyft and Uber has revolutionized urban mobility in the past decade. Given their increasingly important role in today's society, recent years have seen growing interest in integrating these services with mass transit options, in order to leverage the on-demand and flexible nature of the former, with the sustainability and cost-effectiveness of the latter. Our work explores a set of operational questions that are critical to the success of any such integrated marketplace: Given a set of trip requests, and the ability to utilize ride-hailing services, which mass-transit routes should the transit agency operate? How frequently should it operate each route? And how can ride-hailing trips be used to both help connect passengers to these routes, and also cover trips which are not efficiently served by mass transit? We study these under a model in which a Mobility-on-Demand provider (the platform ) has control of a vehicle fleet comprising both single-occupancy and high-capacity vehicles (e.g., cars and buses, respectively). The platform is faced with a number of trip requests to fill during a short time window, and can service these trip requests via different trip options : it can dispatch a car to transport the passenger from source to destination; it can use a car for the first and last legs of the trip, and have the passenger travel by bus in between; or it can use more complicated trips comprising of multiple car and bus legs. Given a set of rewards for matching passengers to trip options, and costs for operating each line (i.e., a bus route and associated frequency), the goal of the platform is to determine the optimal set of lines to operate (given a fixed budget for opening lines), as well as the assignment of passengers to trip options utilizing these lines, in order to maximize the total reward. We refer to this as the Real-Time Line Planning Problem (RLpp). We first demonstrate the computational limits of RLpp by showing that no constant-factor approximation is possible if we relax either one of two assumptions: (i) access to a pre-specified set of feasible lines, and (ii) no bus-to-bus transfers. These assumptions are practically motivated and common in the literature, but our work is the first to formally demonstrate their necessity.

show abstract

Blind Dynamic Resource Allocation in Closed Networks via Mirror Backpressure

Cited by 5 publications

References 54 publications

Learning and Information in Stochastic Networks and Queues

Learning and Information in Stochastic Networks and Queues

Dynamic Pricing and Matching for Two-Sided Queues

Real-time Approximate Routing for Smart Transit Systems

Contact Info

Product

Resources

About