REU Project Opportunities

Biodegradable Sensors, Attachments, and Power Sources for Agricultural Monitoring

(Faculty Mentor: Mark G. Allen, Alfred Fitler Moore Professor of Electrical and Systems Engineering, University of Pennsylvania) 

Location: University of Pennsylvania (Philadelphia, PA)

A key technology enabler for widespread monitoring of agricultural fields is the ability to make sensors and associated elements that are biodegradable; i.e., that will not contribute to pollution or contamination of the agricultural environment once their functional lifetime and utility is completed. The Allen group is working on fabrication technologies for the realization of sensors, power sources, and packages based on all-biodegradable and biocompatible materials. Micromechanical structures with barbs and piercing structures for position stability, pressure sensors capable of RF interrogation, and biodegradable batteries and power sources have all been fabricated from biodegradable/biocompatible materials and large- area fabrication methods such as micromolding and lamination (Fig. 1). 




Figure 1. Structures and sensors fabricated from biodegradable/biocompatible materials. (Left) Wireless pressure sensor from biodegradable polymers and metals; (Center left with inset) biodegradable battery based on Mg and NaCl; (Center right) electrochemical sensor for ion and oxygen concentration measurement based on Au and silicone;(Right) Barbed cylindrical package structure from biodegradable polymers for chaff packaging. All structures were fabricated based on large area techniques such as molding and lamination. 


REU students will work to expand this technology suite to (1) encompass new biodegradable substrates; and (2) integrate new nano-based sensing materials so as to realize smart chaff suitable for crop monitoring using additional sensing schemes such as water, heat, ions, and pH. Students will not only gain appreciation for the microfabrication technologies needed to produce biologically-degradable MEMS structures and sensors, but will also gain understanding of transduction principles and measurement science in this important application realm.

Autonomous Drone Charging Stations for Agricultural Applications

(Faculty Mentor:  David P. Arnold, George Kirkland Engineering Leadership Professor of Electrical and Computer Engineering, University of Florida) 

Location: University of Florida (Gainesville, FL) 

Precision agriculture aims to increase food production while minimizing agriculture inputs of water, energy, and fertilizers. One can envision swarms of aerial drones and ground-based robots that collect data over vast plots of row or tree corps. These robots will collect vast amounts of data over long periods of times. While autonomous robot operation (localization, navigation, and mapping) is quickly becoming a reality, these systems still require a human to plug-in and recharge their batteries. Due in part to this power problem, scaling these approaches to hundreds of robots over vast areas swaths of farmland is hindered by (a) the vast physical distances required, (b) the limited availability of electrical power sources, and (c) the potentially wet /dirty conditions of an outdoor environment. To meet this need, the focus of this project is to develop self-sustaining, field-deployable, aerodromes to support autonomous recharging of quad-copter style aerial drones. These systems will be 100% solar-powered and make use of wireless power transfer, but must also be ruggedized for operation in wet, dusty, field-deployed locations. For this project, the system must source power from a solar panel, and support autonomous recharging without any human intervention, i.e. the drone lands, recharges itself, and then continues on a mission. Additional optional enhancements may include data communication between the drone and charging station to transfer information such as state-of-charge, location/availability of a charging station; simultaneous multi-drone charging using multiple landing sites; a “smart” air-traffic controller that would manage prioritization/sequencing of multiple drones based on state of charge and availability of a charging station.

Micro Solid Oxide Fuel Cell as Drone Power for Extended Duration Flight

(Faculty Mentor:  Sue Ann Bidstrup Allen, Professor of Chemical and Biomolecular Engineering, University of Pennsylvania)

Location: University of Pennsylvania (Philadelphia, PA)

Drones will need to cover large areas for crop inspection and sensor interrogation. However, the flight duration of current drones is usually limited to 30 minutes due to the limited energy density of batteries used in drones. An approach to extend the flight duration of drones is to electrochemically exploit hydrocarbon fuels that have theoretical energy densities several orders of magnitude higher than lithium-ion batteries. Through a flight duration model of agricultural drones, fuel cells that operate on hydrocarbon fuels can easily extend the flight duration over two hours, which is several times longer than that of benchmark lithium-ion batteries. One of the constraints of using fuel cells as the power source for drones is a relatively low specific power density of fuel cells. MEMS technology can significantly reduce the ion transport path and increase the electrochemically active area of electrodes, thus meeting the power requirement for drones. REU students will work with graduate students in engineering the fabrication process of high power density fuel cells and measuring the electrochemical performance of the cell.

Data Pipeline Engineering and Integration with models

(Faculty Mentor: Dennis R. Buckmaster, Professor of Agricultural & Biological Engineering, Purdue University)

Location: Purdue University (West Lafayette, IN)

Digital Agriculture, at its best, builds upon decades of discipline research with some integration of new IoT sensors and communication pathways as well as public resource data such as weather, soil, and topography. One challenge to be addressed is to more fully document the backstory or fuller context of situations so that artificial intelligence and machine learning can be more complete and robust. Another is the integration of mechanistic (descriptive of the fundamental science) models that might be biological, physical, chemical, logistical, economic, etc. in origin. The better parameterization of these models and even auto-population of initial conditions can stem from data sets and data streams. In this project, the student will extract biophysical model(s) from literature and other simulations to meld model + data. It will require interoperability focus and that involves wise choices of data architecture and an integration with data pipelines (often based on open source tools). The end game is to provide better insight (including probabilities, when applicable) for tactical and strategic cropping decisions while preserving security and privacy.

Design of an IoT4Ag Robotic Sensor Deployment System

(Faculty Mentor:  David J. Cappelleri, Professor of Mechanical Engineering, Purdue University)

Location: Purdue University (West Lafayette, IN)

The goal of this project is to design an IoT4Ag sensor deployment system for autonomous agricultural ground robot. Two types of IoT sensors must be deployed by the robotic platform. Chaff sensors need to be distributed on the surface of soil at locations with designated spacing to ensure appropriate spatial coverage for the field of interest.  The second type of sensors similarly need to be spread about the field but require them to be inserted into the soil at a depth of approximately 3” deep. Thus, the developed sensor deployment system should be able to 1. Store the sensors that need to be deployed; 2. Distribute sensors at a designated spacing above the soil; and 3. Insert the sensors into the ground at a designated spacing in the soil; and 4. Log the type of sensor that has been distributed, its sensor ID, and its placement location. This project will require the mechanical design of the deployment systems, mechatronic system design for operating and controlling the systems, and integration and interfacing with the agricultural ground robot for execution and tracking of sensor deployment locations. Field tests will be conducted at the Purdue University Agronomy Center for Research and Education (ACRE) facility.

Integrated remote sensing systems for high resolution monitoring in agriculture

(Faculty Mentor: Melba M. Crawford, Professor of Agronomy, Purdue University) 

Location: Purdue University (West Lafayette, IN)

With the advances in science and technology, it has become possible to generate thousands of genotypes of a plant and produce the seeds at low cost. To evaluate the performance of the plant varieties, numerous physical and agronomic traits are measured in the field during the growing season (phenotyping), and the varieties with the most desirable traits are selected. High-throughput phenotyping using non-invasive remote sensing (RS) technologies in combination with machine learning based analytics has become a critical part of the plant breeding chain focused on reducing the time and cost of the selection process for the “best” genotypes with respect to the trait(s) of interest. In this interdisciplinary project, we will investigate the use of RGB and multi/hyperspectral cameras and LiDAR integrated with ground-based and aerial robotics platforms to measure and monitor variables of interest (e.g. plant count, plant height, leaf area, and spectral response) throughout the growing season to evaluate varieties and management practices.  Undergraduate students will work with teams of graduate students in engineering, agronomy, and aviation technology to develop and calibrate robotic systems, acquire and process the remote sensing data, and collect field reference data for predictive models.

Use of AI Images and machine learning in precision agriculture

(Faculty Mentor: William R. Eisenstadt, Professor of Electrical and Computer Engineering, University of Florida) and (Faculty Mentor: Ian Small, Professor of Plant Pathology, University of Florida) 

Location: University of Florida (Gainesville, FL) 

The engineering and agricultural researchers at UF and IFAS have been successfully developing systems and AI for precision agriculture including sensors, stationary and roving cameras, powerful computers, communications, and AI analysis to evaluate the crop condition and disease activity. There are great opportunities and challenges in applying these techniques in the field. In this project, the student will employ existing stationary camera systems to evaluate cotton Hardlock in the field and will perform machine learning image analysis of the data from these systems. Cotton Hardlock adversely effects harvesting cotton balls and can be controlled if detected during the growth of the cotton plants.  The student will upgrade the existing AI systems for Hardlock by doing one or more of the following, 

  • Assemble the hardware components in an environmentally protected package on a post that provides good visual camera access as the plants grow. 
  • Improve the camera quality and camera water resistance.
  • Develop a method to keep the camera clean of debris.
  • Include a sensor or imager that detects the top of the canopy and communicates with a mechanical device that moves the height of the camera. This sensor could double as a method to measure plant height over time as an indicator of plant growth rate. 
  • Add a solar power system to provide system power. 
  • Transmit Hardlock assessment results through a long-distance wireless data transfer system, such as LoRaWAN to a central WiFi, cell modem or ethernet node.
  • Examine the use of Starlink to receive the image data from the Hardlock detection system.
  • Develop web server software to interface with the Hardlock AI imaging system.
  • Improve and assess model accuracy when detecting Hardlock under field conditions.

Co-design of Optical Sensor – Detector Systems

(Faculty Mentor: Cherie Kagan, Professor of Electrical & Systems Engineering, University of Pennsylvania)  

Location: University of Pennsylvania (Philadelphia, PA)

Current remote sensing techniques utilized in precision agriculture involve RGB, multi/hyperspectral, and LiDAR imaging using aerial and ground robot systems. These systems cannot directly measure plant stress indicators such as transpiration rate and leaf temperature. Low-cost, biodegradable colorimetric sensors are being developed in IoT4Ag to directly measure plant stress indicators and communicate this information through optical signals detectable with current imaging systems. We propose an REU project to define specifications for, and co-design optical sensor-detector systems. These sensor-detector systems are composed of highly-reflective optical sensors consisting of nanostructured metasurfaces embedded in adaptive polymers deployed on soil and leaf surfaces, and optical detectors mounted on ground and aerial robots for sensor readout. Metasurface resonances shift as the adaptive polymer responds to environmental stimuli, yielding a change in color to be read by cameras. We aim to design high quality-factor (Q) metasurfaces by engineering the size, shape, composition, and arrangement of constituent nanostructures, such that the signal-to-noise ratio is maximized. The project also involves developing a testbed to assess the detector sensitivity to camera orientation and distance from the sensor, sensor reflectivity, camera resolution, natural/artificial illumination etc. to inform metasurface sensor design criteria. Image analysis will be conducted to derive quantitative metrics of sensor-detector performance and to select cameras with suitable size, weight, and operating spectral range for deployment.

Development of ISOBlue: An Open Source Platform for Edge Computing and Communications to Enable At-Scale Data Collection from Novel Agricultural Sensors

(Faculty Mentor: James V. Krogmeier, Professor of  Electrical & Computer Engineering, Purdue University) 

Location: Purdue University (West Lafayette, IN)

This project focuses on leveraging existing ag machines (tractors, combines, sprayers, etc.) as mobile platforms for sensor data collection, processing, and cloud connection. Because these sorts of machines travel widely over a farm and have ample power and space for communications and edge computing they are ideal for early deployment of IoT4Ag sensors. This work builds on the existing ISOBlue project to extend its influence beyond simple logging of a machine’s controller area network (CAN) bus to general purpose edge computing that can also support a host of radios, software defined radios (SDRs), and backhauls. Over several development cycles the ISOBlue project has evolved to be a software stack, called Avena, that provides open source developers a telematics and edge computing base for creation of farm machine software applications. In order to provide abstraction and a predictable software-machine interface, a collection of open source tools is used in its implementation including Docker, for consistent software-runtime and easy distribution, Wireguard,  for secure and flexible network access,, for message queue supporting inter-process communication and micro-service style pipelines, and Ansible, for consistent deployments. An REU student will work on the ISOBlue project to add the capability to perform as a LoRaWAN gateway and/or to incorporate CBRS/TVWS and SDR style radios as a backhaul for machine and other nearby sensor data and run machine-to-machine communication experiments.

Precision Agriculture in Orchards

(Faculty Mentor:  Vijay Kumar, Professor of Mechanical Engineering and Applied Mechanics, Computer and Information Systems, and Electrical and Systems Engineering, University of Pennsylvania) 

Location: University of Pennsylvania (Philadelphia, PA)

Precision agriculture, namely continuous, high spatial resolution monitoring of the acreage of a field and the crops thereon, promises better management of irrigation, energy intensive fertilizers, and other plant nutrients, as well as identification of crop stress. Remote sensing satellites and airborne sensing with winged aircraft have allowed scientists and farmers to map large farmlands and forests through acquisition of multispectral imagery and 3-D structural data.  However, data from these platforms are costly and typically lack the spatio- temporal (e.g., cm-scale, day-to-day) resolution necessary for precision agriculture. A REU student will work with doctoral students in Kumar’s group to develop the integrated hardware and software to enable the automated acquisition of data and the translation to actionable information for farming of high-value crops.

UAV Trajectory Optimization for Communications Coverage

(Faculty Mentor:  David Love, Professor of Electrical and Computer Engineering, Purdue)

Location: Purdue University (West Lafayette, IN)

Precision agricultural systems depend on high-rate communications for controlling the myriad of agricultural control and sensing applications. To provide this connectivity, new network deployment approaches are needed.  In this project, we look at how a UAV can be used to offer connectivity over a large coverage area.  We optimize the UAV trajectory subject to aerodynamic, power, and rate constraints.  Our focus is primarily on fixed-wing UAVs which can be used for other precision agriculture applications in addition to communications.

Farming for the Future: Data-driven Precision Plant Phenotyping with Controlled Environment Agriculture

(Faculty Mentor: Rahul Mangharam, Professor of Electrical and Systems Engineering, University of Pennsylvania)

Location: University of Pennsylvania (Philadelphia, PA)

Farming for the future needs to go beyond conventional field agriculture which is subject to the vagaries of climate, drought, blight and results in significant use of pesticides and fertilizers. In this project, we develop indoor controlled environment farms to execute ‘grow recipes’ which investigate the optimal growing conditions for plant yield, nutrition and taste. You will be involved in developing the operating system to conduct dozens of concurrent experiments which quantitatively sense plant growth, yield, and abiotic/biotic stress through a machine learning based computer vision pipeline; control growing conditions of lighting, nutrient supply, air quality, etc. and adapt grow recipes to find the best farm outcomes. The ideal candidate would have a background in computer vision, machine learning and likes hands-on work in developing bio-physically constrained machine learning models for future farming.

Wireless Communication with Subsurface Agricultural Sensors

(Faculty Mentor: Troy Olsson, Professor of Electrical and Systems Engineering, University of Pennsylvania)

Location: University of Pennsylvania (Philadelphia, PA)

To massively deploy agricultural sensors key breakthroughs are needed in the cost, volume, biodegradability, and subsurface operation of wireless sensor interfaces. The objective of this project is to develop radio frequency (RF) hardware and waveforms for remotely and passively interrogating buried agricultural sensors. The Olsson lab is exploring passive wireless sensor architectures that minimize the number and size of the wireless components, are amenable to interfacing with a broad range of sensors, and are constructed entirely from biodegradable and/or benign materials. In particular, we are exploiting the recent breakthroughs in the performance of RF microelectromechanical systems (MEMS) devices developed in our lab to miniaturize and extend the range of passive wireless sensors.  In this REU project, the student will participate in the design of passive wireless sensors, their components, and interrogation waveforms. The REU student will also characterize the passive wireless sensors in both a laboratory and subsurface field environment, providing valuable feedback to improve the designs.

Robust Deep Learning for Agricultural Data

(Faculty Mentor: George Pappas, Professor of Electrical and Systems Engineering, University of Pennsylvania) 

Location: University of Pennsylvania (Philadelphia, PA)

The goal of this project is to make deep learning approaches for agricultural prediction more robust to environmental variations (snow, fog, rain, low lighting, new sensors). Our group has developed a new approach for robust deep learning for physically induced changes in the data. The specific technical goal of this paper is to apply this deep learning technique to agricultural datasets. The successful applicant would be very well versed in Python, PyTorch, GitHub, deep learning algorithms, and have an interest in applying their skills in data-driven prediction for data in standard agricultural datasets as well as datasets developed by other groups in the IoT4Ag consortium.

Assessing aflatoxin risk points in peanuts via polymer science and chemical sensing

(Faculty Mentor:  Brent Sumerlin,  Professor of Chemistry, University of Florida)

Location: University of Florida (Gainesville, FL) 

Precision agriculture is an emerging field which uses advanced technological tools for assisting farming management decision and production to secure and improve crop yields. Monitoring agriculture crops in terms of soils nutriments, pathogen presences and environmental changes is essential to increase farming profitability, quality, and production. Therefore, sophisticated low-cost sensors capable of detecting negative impacts in soil at an early stage and over extended areas, are highly sought. In this project, we will focus on the development of a sensor able to selectively detect aflatoxins, one of the most carcinogenic compounds in peanuts and in soil. The sensor design will be based on the detection of specific small molecules by molecularly imprinted polymer (MIP) networks immobilized to a plasmonic lattice array nanostructure. The detection of toxins will induce a detectable change in the refractive index of the hydrogels, resulting in a spectral shift of the resonances of the array. An eventual goal is for such optical changes to enable data collection and communication of the information from the site of disease in the field to swarms of drones.

MIP network construction will be achieved by carrying out polymerizations of multifunctional (f ≥ 3) vinyl monomers (i.e., crosslinkers) and mixtures of monofunctional (f = 2) vinyl monomers, a fraction of which contain functional groups that promote supramolecular interactions with the target of detection. Crosslinking will occur in the presence of sacrificial templates that mimic aflatoxin. Subsequent isolation and purification will result in crosslinked polymeric materials with templated voids that enable specific binding with the target. Working with ERC Director Kagan, the MIP will be designed to be embeddable within plasmonic arrays, with the target of identifying and producing an optical sensor constructed by coating the molecular imprinted hydrogel to interface with a plasmonic nanorod lattice array-based system.

Biodegradable nanocellulose-based agricultural sensors

(Faculty Mentor: Kevin T. Turner, Professor of Mechanical Engineering and Applied Mechanics, University of Pennsylvania) 

Location: University of Pennsylvania (Philadelphia, PA)

The objective of this project is to develop a new class of agricultural moisture sensors that are low cost and transient (i.e., biodegrade and disappear over time).  The combination of low-cost and biodegradability will allow these sensors to be broadly deployed in agricultural fields and enable distributed sensing.  Cellulose nanomaterials are natural materials that can be assembled into bulk films and these films have previously been used as substrates for flexible electronics.15 In this project, the high moisture sensitivity of nanocelluose materials will be exploited to realize a new class of high performance sensors.  As the aim is to realize sensors that are biodegradable, the focus will be on passive sensors (i.e., no battery) that use the hygroexpansivity of nanocellulose and patterned nanostructures to produce a change in optical response that can be interrogated remotely via an autonomous vehicle.  In this REU project, the student will develop fabrication methods, build devices, and characterize the performance of the sensors.

Thermal Stress Sensing for Corn Production Systems

(Faculty Mentor: Tony J. Vyn, Professor of Agronomy, Purdue University) 

Location: Purdue University (West Lafayette, IN)

Crop canopy temperatures are modulated by transpiration of water vapor from leaf surfaces when water exits via leaf stomata. Although thermal sensors are being deployed on drones and autonomous robots, too little is known about the relationship between evaporative cooling and stomatal conductance that can be measured directly via leaf photosynthesis assessment (e.g. with a LiCor 6400 or 6800 portable photosynthesis system). The simultaneous and direct measurement of thermal properties of corn canopies from above the canopy and below the canopy is suggested here to coincide with leaf photosynthesis measurements. The project goal is to investigate the differential between air temperatures and both upper and lower leaf temperatures via both thermal sensors and leaf stomatal conductance assessment for corn plots under a range of water deficit conditions. Knowing these relationships could help guide the timing (diurnal and weekly frequency) for thermal canopy assessments at different growth stages. Field experiments will be established in spring 2022 at the Agronomy Center for Research and Education near West Lafayette, IN.  Corn treatments may include both hybrid and management variables intended to create a spectrum of crop water stress. Corn biomass measurements will also be taken to study crop growth rates occurring in the actual range of “water productivity” treatments.

Autonomous Interactive Plant Root Phenotyping Applications

(Faculty Mentor:  Alina Zare,  Professor of Electrical and Computer Engineering, University of Florida) 

Location: University of Florida (Gainesville, FL) 

In order to understand how to increase crop yields, breed drought tolerant plants, investigate relationships between root architecture and soil organic matter, and explore how roots can play in a role in greenhouse gas mitigation, we need to be able to study plant root systems effectively. However, we are lacking high-throughput, high-quality sensors, instruments and techniques for plant root analysis. Techniques available for analyzing root systems in field conditions are generally very labor intensive, allow for the collection of only a limited amount of data and are often destructive to the plant. Once root data and imagery have been collected using current root imaging technology, analysis is often further hampered by the challenges associated with generating accurate training data.  Most supervised machine learning algorithms assume that each training data point is paired with an accurate training label. Obtaining accurate training label information is often time consuming and expensive, making it infeasible for large plant root image data sets. Furthermore, human annotators may be inconsistent when labeling a data set, providing inherently imprecise label information. Given this, often one has access only to inaccurately labeled training data. To overcome the lack of accurately labeled training, an approach that can learn from uncertain training labels, such as Multiple Instance Learning (MIL) methods, is required. In this REU project, we will investigate and advance approaches for characterizing and understanding plant roots using methods that focus on alleviating the labor intensive, expensive and time consuming aspects of algorithm training and testing.