Sensors: Multi-Sensor Information Processing with Automotive Applications

汽车发动机、变速器和排放后处理系统的技术进步为开发复杂的信息处理和控制策略带来了新的挑战，从而以更低的成本优化车辆性能。传感器信息处理在这一追求中至关重要。低成本传感器的最佳利用、多传感器的信息协调以及使用能力和精度有限的传感器估计内部状态和系统参数已成为关键设计考虑因素之一。以汽车系统为关键平台，以汽油直喷发动机及其后处理系统为方法论开发和实施，本计画将探讨下列感测器资讯处理的基本问题：（1）如何利用低成本且只能提供有限资讯与精确度的感测器来辨识系统参数，进而估计内部状态？（2）如何使多传感器系统的信息效用最大化？(3)传感器位置、配置和特性对性能优势和成本有何影响？由于高度非线性、大的不确定性和传感器的局限性，这些问题极具挑战性，需要新的方法和实现技术来推动系统辨识和实际传感器信息处理的前沿。拥有超过2.2亿辆注册车辆和2.8万亿英里的年行驶里程，美国汽车工业对国家和全球经济、安全和环境都有着巨大的影响。本项目将在PI及其合作者过去广泛研究工作的基础上，开发一种创新的多传感器识别和估计方法。通过与汽车行业的研究人员合作，我们将采用和实施这项调查的结果，以设计更好的汽车系统控制和适应策略，从而提高性能。该项目的成功完成还将引入新的识别和传感器信息方法，这些方法远远超出了这些领域的已知方法。特别是，它将导致新的多传感器识别方法，使用二进制值和其他非线性传感器，以及对实际传感器信息处理的实现问题的新理解。鉴于在许多应用领域（例如气体含量传感器、生物传感器、无线医疗传感器、纳米传感器等）中的巨大传感器开发努力，这些都将需要先进的和新的信息处理技术，以最大限度地发挥其能力和效用，本项目预计这种新兴的要求传感器信息处理方法，并开发通用的方法，将看到增加效用时，新的传感器被开发。广泛的影响：除了汽车动力系统平台之外，该项目的研究结果将在广泛的应用中具有直接效用，包括车辆侧翻预测、燃料电池系统和车辆的故障诊断、计算机网络交通控制、医疗传感器信息处理和过程控制问题。PI和他的行业合作者目前正在这些领域进行方法学改进、技术转让和设备开发。该研究项目包括跨广泛学科的基础研究和技术开发。它直接针对广泛而重要的汽车应用，涉及数学建模，传感器信号处理和系统识别，并利用福特汽车公司最先进的设备。因此，它为参与的本科生和研究生提供了一个极好的机会，让他们接触到大量的科学和技术前沿，掌握基本的方法论开发，动手设计技能和行业经验。该项目的研究成果和课程材料将在专业会议和期刊上广泛传播。本项目所产生的软件包将通过本项目的主页免费发布给公众使用。

Technology advancement in automotive engine, transmission, and emission aftertreatment systems hasushered in new challenges in developing sophisticated information processing and control strategies tooptimize vehicle performance at reduced costs. Sensor information processing is of vital importance inthis pursuit. Optimal utility of low-cost sensors, information coordination of multiple sensors, andestimation of internal states and system parameters using sensors of limited capability and accuracy havebecome one of the key design considerations.Using automotive systems as a key platform, and gasoline direct injection engine and its aftertreatmentsystem for methodology development and implementation, this project will investigate the followingfundamental issues on sensor information processing: (1) How can one identify system parameters orestimate internal states by using low-cost sensors that provide only limited information and accuracy? (2)How can one maximize the information utility of multiple sensor systems? (3) What is the impact ofsensor location configuration and characteristics on performance benefits and costs? Due to highnonlinearity, large uncertainty, and sensor limitations, these issues are extremely challenging, anddemand new methodologies and implementation technologies that will push forward the frontiers ofsystem identification and practical sensor information processing.Intellectual Merit of the Project: With over 220 million registered vehicles and 2.8 trillion miles ofannually traveled distances, the US automotive industry bears an enormous impact on the national andglobal economy, safety, and environment. This project will develop an innovative methodology of multi-sensoridentification and estimation, on the basis of extensive past research effort from the PI and hiscollaborators. In collaboration with researchers from the automotive industry, findings from thisinvestigation will be employed and implemented to design better control and adaptation strategies inautomotive systems for improved performance.Successful completion of this project will also introduce new identification and sensor informationmethods that go much beyond what is known in these fields. In particular, it will lead to new multi-sensoridentification methods that use binary-valued and other nonlinear sensors, and new understanding ofimplementation issues on practical sensor information processing. In light of tremendous sensordevelopment effort in many application areas, such as gas content sensors, biosensors, wireless medicalsensors, nanosensors, etc., which will all demand advanced and new information processing techniques tomaximize their capabilities and utilities, this project anticipates such emerging requirements for sensorinformation processing methodologies, and develops generic methods that will see increased utility whennew sensors are developed.Broad Impact: Beyond the automotive powertrain platforms, the findings from this project will havedirect utility in a wide array of applications, including vehicle rollover prediction, fault diagnosis of fuelcell systems and vehicles, computer network traffic control, medical sensor information processing, andprocess control problems. The PI and his industry collaborators are currently pursuing methodologyenhancement, technology transfer, and device development in these areas.This research project encompasses fundamental research and technology development across a widerange of disciplines. It targets directly at a broad and important automotive application; involvesmathematics modeling, sensor signal processing, and system identification; and utilizes the mostadvanced facility at Ford Motor Company. As such it provides participating undergraduate and graduatestudents an excellent opportunity to be exposed to a large spectrum of scientific and technology frontierswith fundamental methodology development, hand-on design skills, and industry experience. Theresearch findings and course material from this project will be widely disseminated in professionalconferences and journals. Software packages resulted from this project will be released through theproject homepage for the public use free of charge.