With the enrichment of large-scale group decision-making (LSGDM) methods, the decentralized consensus reaching process (CRP) has demonstrated many advantages. However, when the probabilistic linguistic preference relation (PLPR) is utilized in the decentralized CRP, its consistency and readability are hardly to maintain. Besides, the low-cost consensus adjustment and non-cooperative behaviors of subgroups are still not considered simultaneously in the decentralized CRP. In order to solve these problems, this article proposes a decentralized feedback-based consensus model to support CRP in LSGDM based on PLPR with complete readability. First, to maintain the consistency of PLPR throughout the LSGDM process, an additive expected consistency for PLPR is specifically defined. This definition enables the automatic consistency maintenance of PLPR during linear-weight-based clustering, opinion adjustment, and opinion aggregation. Given that the existing consistency adjustment method always destroys the readability of the original PLPR, a definition of complete readability for PLPR, reflected by a reasonable probability distribution of its elements, is proposed. This is followed by a consistency-improving optimization model that considers both the adjustment cost and complete readability. Subsequently, in order to support a more realistic CRP, a decentralized feedback-based minimum cost consensus model is established to improve the group consensus level while addressing the non-cooperative behaviors of subgroups. Furthermore, an illustrative example of the selection of expressway repair plans is presented to testify the practicality of the proposed methods and demonstrate the distinctive characteristics in comparison with the existing approaches.

随着大规模群体决策 （LSGDM） 方法的丰富，去中心化共识达成过程 （CRP） 展示了许多优势。然而，当在分散式 CRP 中使用概率语言偏好关系 （PLPR） 时，其一致性和可读性几乎难以保持。此外，在去中心化 CRP 中，低成本的共识调整和子组的非合作行为仍然没有同时考虑。为了解决这些问题，本文提出了一种基于去中心化反馈的共识模型，以支持基于 PLPR 的 LSGDM 中的 CRP，具有完全可读性。首先，为了在整个 LSGDM 过程中保持 PLPR 的一致性，专门定义了 PLPR 的加性预期一致性。此定义支持在基于线性权重的聚类、意见调整和意见聚合期间自动维护 PLPR 的一致性。鉴于现有的一致性调整方法总是破坏原始 PLPR 的可读性，因此提出了 PLPR 完全可读性的定义，该定义体现在其元素的合理概率分布上。接下来是一个提高一致性的优化模型，该模型同时考虑了调整成本和完全可读性。随后，为了支持更现实的 CRP，建立了基于反馈的去中心化最低成本共识模型，以提高群体共识水平，同时解决子群体的不合作行为。此外，还给出了高速公路修复方案选择的说明性示例，以证明所提出的方法的实用性，并展示与现有方法相比的显著特点。