Purpose

In the reduction of food waste and the provision of food to the hungry, food banks play critical roles. However, as they are generally run by charitable organisations that are chronically short of human and other resources, their inbound logistics efforts commonly experience difficulties in two key areas: 1) how to organise stocks of donated food, and 2) how to assess the donated items quality and fitness for purpose. To address both these problems, the authors aimed to develop a novel artificial intelligence (AI)-based approach to food quality and warehousing management in food banks.

Design/methodology/approach

For diagnosing the quality of donated food items, the authors designed a convolutional neural network (CNN); and to ascertain how best to arrange such items within food banks' available space, reinforcement learning was used.

Findings

Testing of the proposed innovative CNN demonstrated its ability to provide consistent, accurate assessments of the quality of five species of donated fruit. The reinforcement-learning approach, as well as being capable of devising effective storage schemes for donated food, required fewer computational resources that some other approaches that have been proposed.

Research limitations/implications

Viewed through the lens of expectation-confirmation theory, which the authors found useful as a framework for research of this kind, the proposed AI-based inbound-logistics techniques exceeded normal expectations and achieved positive disconfirmation.

Practical implications

As well as enabling machines to learn how inbound logistics are handed by human operators, this pioneering study showed that such machines could achieve excellent performance: i.e., that the consistency provided by AI operations could in future dramatically enhance such logistics' quality, in the specific case of food banks.

Originality/value

This paper’s AI-based inbound-logistics approach differs considerably from others, and was found able to effectively manage both food-quality assessments and food-storage decisions more rapidly than its counterparts.

中文翻译：

---

基于人工智能的食品银行入库物流食品质量和仓储管理

目的

在减少食物浪费和向饥饿者提供食物方面，食物银行发挥着至关重要的作用。然而，由于它们通常由长期缺乏人力和其他资源的慈善组织运营，其入境物流工作通常在两个关键领域遇到困难：1）如何组织捐赠的食品库存，2）如何评估捐赠的食品物品的质量和用途的适用性。为了解决这两个问题，作者旨在开发一种基于人工智能 (AI) 的新型方法，用于食品银行的食品质量和仓储管理。

设计/方法论/途径

为了诊断捐赠食品的质量，作者设计了一个卷积神经网络（CNN）；为了确定如何在食物银行的可用空间内最好地安排这些物品，使用了强化学习。

发现

对拟议的创新 CNN 的测试表明，它能够对五种捐赠水果的质量提供一致、准确的评估。强化学习方法不仅能够为捐赠的食物设计有效的存储方案，而且比其他一些已提出的方法需要更少的计算资源。

研究局限性/影响

作者发现期望确认理论作为此类研究的框架非常有用，从期望确认理论的角度来看，所提出的基于人工智能的入库物流技术超出了正常预期，并实现了积极的反确认。

实际影响

这项开创性的研究不仅使机器能够学习人类操作员如何处理入站物流，还表明此类机器可以实现出色的性能：即，人工智能操作提供的一致性在未来可以显着提高此类物流的质量，在特定领域食品银行的案例。

原创性/价值

本文基于人工智能的入库物流方法与其他方法有很大不同，并且被发现能够比同类方法更快地有效管理食品质量评估和食品存储决策。