No Agent Left Behind: Dynamic Fair Division of Multiple Resources

Abstract

Recently fair division theory has emerged as a promising approach for allocation of multiple computational resources among agents. While in reality agents are not all present in the system simultaneously, previous work has studied static settings where all relevant information is known upfront. Our goal is to better understand the dynamic setting. On the conceptual level, we develop a dynamic model of fair division, and propose desirable axiomatic properties for dynamic resource allocation mechanisms. On the technical level, we construct two novel mechanisms that provably satisfy some of these properties, and analyze their performance using real data. We believe that our work informs the design of superior multiagent systems, and at the same time expands the scope of fair division theory by initiating the study of dynamic and fair resource allocation mechanisms.

Cake Cutting Algorithms

“…While the model is somewhat different, it has much in common with the cake cutting model. Other papers that study this research direction include papers by Dolev et al (2012), Gutman and Nisan (2012), Parkes et al (2014), and Kash et al (2014).…”

Dynamic fair node spectrum allocation for ad hoc networks using random matrices

“…Similar to Kash et al [16], we denote the number of agents with link requirements as N = {1,…, n}, and the amount of resources agent i requires as Ri. Our example is different in that we consider some links as being extensively used while others are limited.…”

“…A dynamic allocation mechanism is dynamic envy free (DEF) if at any step an agent i envies an agent j only if j arrived before i did and j has not been allocated any resources since i arrived [16].…”

“…Our aim is to design irrevocable resource allocation such that more assets can be allocated to an agent, but can never be taken back until no longer needed. For fairness, envy freeness means that agents like their own allocation best and are truthful needs [16].…”

Multi-Objective Bottleneck-Aware Allocation of Multiple Resources

“…The model provides clients that are bottlenecked on the same device with allocations that are proportional to their fair shares, while allowing allocation ratios between clients in different bottleneck sets to be set by the allocator to maximize the utilization. No agent left behind [7] proposes a dynamic resource allocation mechanism. Also, all these approaches make the assumption that all jobs and/or resources are continuously divisible.…”