SBIR Phase I: Robotic vertical glass sorter powered by machine vision and artificial intelligence for accurate, accessible, sustainable, and cost-effective glass recycling

这个小型企业创新研究(SBIR)第一阶段项目的更广泛影响将解决美国每年处置的1000多万吨玻璃的问题。虽然100%的玻璃是可回收的，但实际上只有33%被回收，因为在回收之前对旧玻璃进行分类和运输是具有挑战性的。这些关键功能的传统方法大大降低了玻璃的价值，因为运输成本和将玻璃与其他废物分类所需的仓库空间，并将透明、棕色和绿色分开。即使是现代的美国玻璃分类设施也需要超过25万平方英尺的仓库空间，用于水平跑道和传送带系统，以清理和分类足够好的废物流，以便自动玻璃分拣机继续运行。这导致了高成本、体力劳动、对机器的负面环境影响，以及高水平的未回收玻璃。拟议的项目推进了一个机器人系统，以有效地进行分类。这个小型企业创新研究第一阶段项目的目标是一个原型漏斗和执行器控制系统，用于检测和分类的人工智能引擎，以及现场性能研究。主要的开发挑战包括增强原型平台的摄像头和传感器网络；开发和验证用于快速分类落地玻璃的人工智能引擎；开发执行器控制，以及按照规范进行系统集成和功能测试。首先，将对样机进行改进，包括机械设计、物流、传感器网络和分拣机构的执行器。接下来，使用预先录制的不同条件下玻璃坠落和垃圾的视频，将开发并验证人工智能引擎，以准确地对玻璃颜色进行分类并触发致动器序列。一旦算法被离线验证，硬件和算法将在系统级别进行集成，并在实时操作中进行性能验证。该奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

The broader impact of this Small Business Innovation Research (SBIR) Phase I project will address the over 10 million metric tons of glass disposed of annually in the U.S. While 100% of glass is recyclable, only 33% is actually recycled, as it is challenging to sorting and transport used glass prior to recycling. Traditional methods for these key functions dramatically reduce the value of the glass due to transportation costs and required warehouse space for sorting glass from other waste and separate the colors of clear, brown, and green. Even modern U.S. glass sorting facilities require warehouse space of more than 250,000 square feet for horizontal runway and belt systems to clean and sort waste streams well enough for an automated glass sorter to continue. This leads to high costs, manual labor, negative environmental impact for the machines, and high levels of glass that is not recycled. The proposed project advances a robotic system to sort efficiently. This Small Business Innovation Research Phase I project will target a prototype hopper and actuator control system, AI engine for detection and sorting, and field performance study. The main development challenges include enhancing the camera and sensor network of the prototype platform; developing and validating an AI engine for rapid classification of falling glass; and developing actuator controls, and system integration and functional testing to specifications. First, the prototype will be enhanced including the mechanical design, material flow, sensor network, and actuators for the sorting mechanisms. Next, using prerecorded videos of falling glass and waste under different conditions, an AI engine will be developed and validated to accurately classify glass colors and trigger an actuator sequence. Once algorithms have been validated offline, the hardware and algorithms will be integrated at a system level and validated for performance in real-time operation.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.