In this work, we revisit the problem of fairly allocating a number of indivisible items that are located on a line to multiple agents. A feasible allocation requires that the allocated items to each agent are connected on the line. The items can be goods on which agents have non-negative utilities, or chores on which the utilities are non-positive. Our objective is to understand the extent to which welfare is inevitably sacrificed by enforcing the allocations to be fair, i.e., price of fairness (PoF). We study both egalitarian and utilitarian welfare. Previous works by Suksompong [Discret. Appl. Math., 2019] and H\"ohne and van Stee [Inf. Comput., 2021] have studied PoF regarding the notions of envy-freeness and proportionality. However, these fair allocations barely exist for indivisible items, and thus in this work, we focus on the relaxations of maximin share fairness and proportionality up to one item, which are guaranteed to be satisfiable. For most settings, we give (almost) tight ratios of PoF and all the upper bounds are proved by designing polynomial time algorithms.
翻译:在这项工作中,我们再次探讨了公平分配属于多个代理商的若干不可分割物品的问题。可行的分配要求分配给每个代理商的物品在线上连接。这些物品可以是代理商拥有非消极公用事业的物品,也可以是公用事业非积极的杂务。我们的目标是了解通过执行公平分配而不可避免地牺牲福利的程度,即公平价格(PoF)。我们研究平等和实用福利。Suksompong[Discret. Appl. Math, 2019]和H\'ohne和van Stee[Inf.Comput., 2021]以前的工作要求将分配给每个代理商的物品与该线上线上。这些物品可以是代理商拥有非消极公用事业的物品,或公用事业非积极性的杂务。然而,这些公平分配对于不可分割的物品几乎不存在,因此在这项工作中,我们侧重于将最大份额的公平性和相称性降为一项物品,这有保证是值得信赖的。对于大多数环境而言,我们给PoF和所有上层界限的紧凑比率(最接近)通过设计聚式时间算法来证明。