To support the understanding of recycling models applied to plastics, the main objective of this work is to offer a literature review of the different reverse logistics (RL) models for collecting plastic waste (PW). The methodology used for processing the scientific literature was content analysis, using the google scholar search engine. The main keywords used were RL and PW. This article is divided into two parts: the first part discusses the development of circular economy models and RL networks and raises the conceptual framework of the research, and the second part presents mathematical models and exploratory studies, proposed as a solution for RL problems of PW. Articles published between years 2014 and 2019 were reviewed. In total, 102 references were used, 70 of them are part of the literature review. According to our findings, we can state that the most widely used solution method for mathematical modeling is mixed-integer linear programming, and for exploratory studies, it was evaluations. About 93% of studies evaluated raw materials related to PW; only 13% of studies had models with stochastic processes; and 88% of the investigations used continuous variables, being the multiobjective functions one of the most used to provide solutions to RL problems. Regarding the mathematical models, 49% were evaluations, 9% corresponded to multicriteria analysis, 29% to linear and nonlinear programming, and 4% to another type of evaluation or model.

为了支持对应用于塑料的回收模型的理解，这项工作的主要目的是对收集塑料废物 (PW) 的不同逆向物流 (RL) 模型进行文献综述。用于处理科学文献的方法是内容分析，使用谷歌学者搜索引擎。使用的主要关键字是 RL 和 PW。本文分为两部分：第一部分讨论循环经济模型和 RL 网络的发展并提出研究的概念框架，第二部分提出数学模型和探索性研究，提出作为 PW RL 问题的解决方案. 审查了 2014 年至 2019 年间发表的文章。总共使用了 102 篇参考文献，其中 70 篇是文献综述的一部分。根据我们的调查结果，我们可以说，数学建模最广泛使用的求解方法是混合整数线性规划，而对于探索性研究，则是评估。大约 93% 的研究评估了与 PW 相关的原材料；只有 13% 的研究具有随机过程模型；并且 88% 的调查使用了连续变量，这是最常用于为 RL 问题提供解决方案的多目标函数之一。关于数学模型，49% 是评估，9% 对应于多标准分析，29% 对应于线性和非线性规划，4% 对应于另一种类型的评估或模型。只有 13% 的研究具有随机过程模型；并且 88% 的调查使用了连续变量，这是最常用于为 RL 问题提供解决方案的多目标函数之一。关于数学模型，49% 是评估，9% 对应于多标准分析，29% 对应于线性和非线性规划，4% 对应于另一种类型的评估或模型。只有 13% 的研究具有随机过程模型；并且 88% 的调查使用了连续变量，这是最常用于为 RL 问题提供解决方案的多目标函数之一。关于数学模型，49% 是评估，9% 对应于多标准分析，29% 对应于线性和非线性规划，4% 对应于另一种类型的评估或模型。