Prototyping an AI-powered robotic picker for sorting shredded aluminum

eCycle Solutions在加拿大经营多个资产回收和电子和电气设备回收设施**。eCycle Solutions是一家加拿大私人控股公司，拥有400多名员工，日处理能力300公吨，所有工厂的仓储和处理空间超过35万平方英尺。**传统上，在报废电子电气设备（WEEE）回收设施，过时的废弃电子产品**被粉碎，以释放其材料。目标是根据材料类型将这些碎片分开，以生产纯商品。为了实现这一目标，使用了各种各样的分类技术**来分离这些材料，然而，由于这些技术的效率低下，污染是普遍存在的。为了解决这种污染，工人们手动从物料流中移除碎片。这个过程是劳动密集型的，耗时的，单调的，重复的，不划算的。Conestogas为这个项目提出的解决方案是开发一个拾取机器人作为机械拾取器，它**能够利用卷积神经网络（CNN）对切碎的电子垃圾进行分类。CNN将“训练”机器人，使其模仿人类操作员的能力，并根据经验“学习”。这将提高纯度和加工速度，并将使大部分劳动力能够部署到其他任务中，从而降低加工成本，提高盈利能力。

eCycle Solutions operates multiple asset recovery and electronics and electrical equipment recycling facilities**across Canada. eCycle Solutions is privately held, Canadian owned with over 400 employees and 300 metric**tonnes of daily processing capacity, with over 350,000 sq/ft of warehousing and processing space across all sites.**Traditionally, at Waste Electrical and Electronic Equipment (WEEE) recycling facilities, outdated waste electronics**are shredded in order to liberate their materials. The objective is to then separate these shredded pieces based**on their material type to produce pure commodities. To accomplish this, a variety of sorting technologies are used**to separate these materials, however, contamination is prevalent due to inefficiencies in these technologies. To**address this contamination, workers manually remove pieces from the material stream. This process is labourintensive,**time-consuming, monotonous, repetitive, and not cost effective.**Conestogas proposed solution for this project is to develop a pick-n-place robot as a mechanical picker, which**is capable of sorting shredded waste electronics utilizing a Convolutional Neural Network (CNN). The CNN will**"train" the robot so that it will mimic the capability of the human operator and "learn" based on experience. This**will result in increased purity rates and processing speeds, and will enable much of the labour force to deployed to**other tasks, which will lead to reduced processing costs, and increased profitability.