SBIR Phase I: GHz Time-of-Flight Depth Sensing for Robotic Bin Picking

SBIR 第一阶段：用于机器人箱拣选的 GHz 飞行时间深度传感

• 批准号: 1647888
• 负责人: John Kohoutek
• 金额: $ 22.48万
• 依托单位: StarSight Inc.
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2016
• 资助国家: 美国
• 起止时间: 2016-12-15 至 2018-08-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1647888&HistoricalAwards=false
• 关键词: SBIR; Phase; GHz; Time; Flight

The broader impact/commercial impact of this project will be to create a depth vision system that acceleratesthe transition to intelligent automation. The intelligent automation movement is driven by the opportunityto increase human productivity, and it can be seen in all domains: from Siemen?s self-organizingmanufacturing plants to Amazon?s proposed delivery drones. This project aims to improve robotic vision,which is currently a limitation to greater autonomy. The research uses innovations in semiconductorphysics to create time-of-flight depth cameras that reach new levels of spatial resolution and fidelity. Thesecameras, in conjunction with rapidly evolving software that uses depth data to model the physical world,give robotic systems the human-like abilities they need to perform more complicated tasks in dynamicenvironments. Robotics will become increasingly effective at the navigation, material handling, and objectmanipulation tasks that are the basis of so many manufacturing and logistics processes. The projectoutcomes will advance the state-of-the-art in depth vision technology, and provide a prototype camera thatcan be used to test a variety of challenging object recognition tasks.This Small Business Innovation Research (SBIR) Phase I project addresses an important limitation in depthcameras that use the phase delay time-of-flight (TOF) method. Existing TOF cameras use ?pixel-leveldemodulation?: specially designed photodetector arrays where each pixel performs both light detection anddemodulation. This design limits the modulation frequency, and thus depth accuracy, because the highfrequency signals create excessive noise and parasitic capacitances. The authors propose an alternative?optical demodulation? design wherein light is demodulated by a single fast optical shutter, and thencollected by a separate conventional image sensor. By separating light modulation from light detection,parasitic capacitances are eliminated and it is possible to reach high frequencies with acceptable powerconsumption and noise. The research program will develop new optical shutters that demonstrate thesecharacteristics, and integrate the shutters into a functioning prototype depth camera. Success will show thatTOF cameras can be produced with substantially higher spatial resolution and design flexibility.

该项目更广泛的影响/商业影响将是创建一个深度视觉系统，加速向智能自动化的过渡。智能自动化运动是由提高人类生产力的机会驱动的，它可以在所有领域看到：从西门子？s self-organizing自组织manufacturing制造plants工厂to Amazon亚马逊？他提议用无人机送货。该项目旨在改善机器人的视觉，这是目前限制更大自主权的一个因素。该研究使用了光电物理学的创新来创建飞行时间深度相机，达到了空间分辨率和保真度的新水平。这些摄像头与快速发展的软件结合使用深度数据来模拟物理世界，使机器人系统具有在动态环境中执行更复杂任务所需的类似人类的能力。机器人在导航、材料处理和物体操作任务方面将变得越来越有效，这些任务是许多制造和物流过程的基础。该项目的成果将推进最先进的深度视觉技术，并提供一个原型相机，可用于测试各种具有挑战性的物体识别任务。这个小企业创新研究（SBIR）第一阶段项目解决了使用相位延迟飞行时间（TOF）方法的深度相机的一个重要限制。现有的TOF相机使用？pixel-leveldemodulation？：专门设计的光电探测器阵列，每个像素执行光检测和解调。这种设计限制了调制频率，从而限制了深度精度，因为高频信号会产生过多的噪声和寄生电容。作者提出了一个替代方案。光解调设计中，光由单个快速光学快门解调，然后由单独的传统图像传感器收集。通过将光调制与光检测分离，消除了寄生电容，并且可以在可接受的功耗和噪声下达到高频。该研究计划将开发新的光学快门，展示这些特性，并将快门集成到功能原型深度相机中。成功将表明TOF相机可以以更高的空间分辨率和设计灵活性生产。