- Aforestation, Climate Change Mitigation and Prediction
- Monitoring Effects of Human and Natural Forcings of Critical Zone Dynamics and Evolution
- Geospatial Monitoring of Air Quality and Pollutants
- Sea-level Rise Mitigation and Adaptation Measures
- Modeling and Prediction of Climate Impacts on Snow and Ice
- Drought Monitoring and Prediction in Semiarid Climates
- Optimal Rainwater Harvesting and Irrigation: Monitoring the Soil Water Balance in the Duke Smart Home Garden
Experiments in Engineering and Computer Science
- Intelligent Sensing and Control for Automotive Human Machine Interface
- Integrated Sensor Path Planning and Control
- Undersea Monitoring and Surveillance
- Robotic Saccadic Adaptation and Visually-guided Auditory Plasticity
- Implementing Neuro-Inspired Computing for Sensing and Information Processing
- Radar Inspired Ultra Low-power Communication for Sensors
- Intelligent Robotic Games
- Experiment on Environmental Monitoring with Gas-sensitive Mobile Robots
Point of contact: Gabriel Katul, Duke University
Location: Duke Forest, Durham, NC
This experiment addresses a primary question in climate change pertaining to the mediating role of the biosphere on elevated atmospheric CO2 concentration, and their influence on rainfall and mean air temperature. The ability of terrestrial ecosystems to absorb CO2 is sensitive to atmospheric conditions, and is characterized by feedback loops that, if characterized by intensive sensor data, can lead to far more accurate predictions. WISeNet trainees apply new optimal environmental sensing methods and algorithms to a wide array of wireless sensors, e.g. gas analyzers, anemometers, and sap flux sensors, presently deployed in the Duke Forest to collect measurements of precipitation, soil moisture, vapor pressure deficit, temperature, and, more importantly, photosynthetically active radiation. Using novel scalable simulation and data processing algorithms, and information-driven sensor management and fusion, WISeNet trainees seek to develop improved models that capture the rich spatial variability in ecosystem carbon dynamics, and natural feedback loops from the environmental controls to surface radiative, physiological, and aerodynamic process, to predict their effects on warming potential.
Experiment on Monitoring effects of Human and Natural Forcings of Critical Zone Dynamics and Evolution
Point of Contact: Amilcare Porporato, Duke University
Location: Calhoun Critical Zone Obervatory (CZO), Sumter National Forest, Calhoun, SC
The Calhoun Critical Zone Observatory (CZO) aims to integrate the sciences of water, mineral, and carbon cycles to quantify the natural and human-forced dynamics and evolution of Earth’s Critical Zone. The Calhoun CZ Observatory is located in the Southern Piedmont of the United States, a region with an environmental history that involves some of the most serious agricultural land and water degradation in North America. Trainees will utilize wireless networks to support ecohydrological and biogeochemical observations and measurements and to test the hypotheses of the CZO.
The CZO has three watersheds for which historical data exists, and this project aims to both re-instrument and up-instrument these catchments (see Figure below). Trainees will work with the CZO PIs to plan and install co-located sensors in these watersheds at various soil depths measuring parameters such as soil moisture, temperature, water potential, CO2 content, and redox potential. Additionally, sampling towers containing instruments such as infrared gas analyzers for CO2 and H2O, net radiation sensors, and rain gages will be constructed. Trainees will develop a wireless network to connect one of the soil sensor groupings in the water sheds and the micrometerological towers to allow for real time remote monitoring of the site conditions. Trainees will apply the data obtained from these sensors to coupled ecohydrological-biogeochemical models to interpret the data, hindcast and forecast ecosystem evolution, as well as develop low-dimensional coupled models of ecosystem evolution. The synergy between NSF funded WISeNet IGERT and Calhoun CZO projects is intended to facilitate the ‘field to forecast’ effort of measuring in real time complex environmental interaction to inform mathematical models environmental prediction and guide real time responses.
Point of contact: Gayle Hagler, EPA
Location: Environmental Protection Agency (EPA), Research Triangle, NC
WISeNet trainees participate in ongoing efforts by the EPA National Risk Management Research Laboratory to collect real-time air pollutant, greenhouse gas data, meteorology, and traffic characteristics, using instrumented vehicles and EPA monitoring sites located near major roadways (such as the GMAP program). These efforts concentrate on sensing and predicting air toxics, fine particles, indoor air quality, ozone, and global climate change, with a goal to reducing risks to human health. Trainees develop and implement deployment, fusion, modeling, and prediction algorithms to integrate these measurements with GPS, seeking to detect air pollution sources and patterns for intelligent mitigation and decision making. They also have the opportunity to demonstrate mobile sensor guidance and control algorithms through EPA field experiments on monitoring of landfills, and oil and gas production. The goal is to develop technologies for detecting leaks over large areas using autonomous ground vehicles equipped with onboard instrumentation for in-situ sensing and data analysis.
Point of Contact: Stefano Lanzoni, University of Padova
Location: University of Padova, Padova, and Venice, Italy
WISeNet trainees apply novel sensor management and navigation algorithms to a hybrid system of in-situ sonar buoys and satellite remote-sensing networks for measuring turbidity, tide levels, and water quality, in an effort to predict changes to the Venice Lagoon, and develop intelligent mitigation and decision tools. Due to the large extension of estuarine and coastal areas, and to the complexity of bio-morphological processes governing their dynamics, prediction and decision tools must be based on spatially-distributed continuous observations both for direct real-time monitoring, and for hydrodynamics and biomorphological models’ calibration and validation. Similar considerations are valid in the case of large oil spills in oceanic waters, or pollutants discharge in coastal waters. Trainees employ new methods for information-driven environmental sensing and prediction to (i) combine remote sensing data with in-situ measurements, and (ii) optimally deploy a heterogeneous sensor network of autonomous vehicles, stationary multiparametric sensors, and sonic buoys that monitor water sediment, nutrient and pollutant loads. The first activity employs in-situ measurements to calibrate and validate remote sensing retrievals of water column properties, and tackle problems connected with fusing ‘point’ field observations with spatially-averaged remote sensing retrievals. The second activity samples space-time variations over a region of interest using optical and thermal sensors on underwater vehicles to increase the sampling density where large gradients in biophysical observations, or intense deposition and erosion are known to occur.
Point of contact: Marc Parlange, EPFL
Location: EFLUM Laboratory, EPFL, Lausanne, Switzerland
As a result of its high albedo and insulating properties, the seasonal snow cover significantly modifies the surface energy budget, and plays a crucial role in mountain hydrology, water resource management, and the management of natural hazards. The generally small surface roughness and the relatively low skin temperature of the snow influence near-surface air flow and stability conditions of the atmospheric boundary layer. The intensity of atmospheric turbulence in turn impacts the microphysical and thermal properties of the snow pack. WISeNet trainees utilize a heterogeneous network of wireless sensors, e.g. ultrasonic anemometers and fiber optic temperature sensors, presently deployed across the high alpine regions in Switzerland to monitor the physical evolution of the snow and ice pack, and the lower atmosphere. Trainees integrate existing numerical tools, such as Large-Eddy simulations, with new stochastic modeling and information-driven sensor management methods in order to improve basic science, and optimally deploy and refurbish these sensors, in an effort to predict the gradual mass loss due to climate change, and rapid events of avalanches and flood pulses.
Point of contact: John Albertson, Duke University
Location: University of Cagliari, Sardinia, Italy
A secure supply of drinking water is a fundamental human need that goes unmet for much of the world’s population. Although water quality can be ensured through engineered treatment and delivery facilities, the quantity of future water availability remains surrounded by significant scientific uncertainty. WISeNet trainees involved in this field experiment utilize a long-standing NSF-funded broad network of soil moisture sensors and lower atmosphere sensors on the island of Sardinia that are designed to address an important knowledge gap, i.e., how changes in the seasonality of precipitation in semi-arid regions interact with vegetation dynamics to affect available surface-water resources. Trainees develop and test hierarchical reinforcement learning and biologically-inspired algorithms for intelligent data fusion and assimilation in models of hydrologic runoff processes, and of vegetation-rainfall feedbacks. By combining ongoing field experiments, coupled eco-hydrological modeling, and intelligent data assimilation, WISeNet trainees seek to identify the controls of soil and vegetation states on runoff generation and spatial patterns, and obtain results general to the broad class of semi-arid watersheds, particularly those in Mediterranean climates.
Optimal Rainwater Harvesting and Irrigation: Monitoring the Soil Water Balance in the Duke Smart Home Garden
Point of contact: Amilcare Porporato, Duke University
Location: Durham, North Carolina
Duke Smart Home is a live-in research laboratory operated by the Pratt School of Engineering with support from industry partners. The home functions as a laboratory for testing innovative features for future residential building technology. The mission of the home is to advance ``smarter living” by automating everyday habits that encourage energy and water efficiency.
To promote water use efficiency and conservation, the home features eight cisterns that collect rainwater: six 400 gallon cisterns in the basement store water for the toilets and washing machine, while two 1200 gallon outdoor cisterns store water for the garden irrigation system. Using Wireless Intelligent Sensor Networks (WISeNets) developed by IGERT trainees, we plan to monitor the water balance in the Smart Home garden to quantify the efficiency of the smart irrigation system and to develop innovative strategies for rainwater collection system design and management.
Our monitoring system, located in the Smart Home backyard, continuously tracks soil and atmospheric states, such as temperature, humidity/moisture, carbon dioxide concentration, solar and terrestrial radiation, wind speed, cistern water depth, and rainfall depth. Trainees will develop a wireless network to monitor these observations in real-time from a smartphone or computer via the web. With these observations, trainees will develop quantitative models to analyze the water balance in the backyard of Smart Home and to assess the water use efficiency of the smart irrigation system. In combination with the city water use inside the Smart Home, trainees will evaluate the economic benefit of the whole water system and develop strategies for scaling up the system from a single home to an entire community.
Point of contact: Silvia Ferrari, Duke University
Location: Ferrari, S.p.A., Maranello, Italy
Modern sensing and information technologies are revolutionizing the possible functionalities of software interfaces between the driver and the automobile. Trainees involved in this experiment develop new methods inspired by modern control theory, computational intelligence, and data mining, for the design of human machine interfaces that adapt the dynamic response of the automobile to the driver, in order to enhance the driver’s experience and the automobile’s performance in the closed loop. In particular, trainees specialize on developing new remote sensors and information processing algorithms for sensing and inferring the driver state, and on developing adaptive feedback control systems. Trainees also collaborate with engineers involved in the development of displays and ambience design to integrate the sensor and data processing algorithms with new information and entertainment technologies.
Point of contact: Lorenzo Marconi, University of Bologna
Location: CASY, University of Bologna, Bologna, Italy
One of the main challenges facing the next generation of autonomous aerial robots is coordinating and executing sensing and communication operations next to loosely-structured infrastructure and obstacles, where interactions with the environment must be explicitly considered. In collaboration with Marconi, WISeNet trainees utilize ground robots in conjunction with an unmanned Vertical Take-Off and Landing (VTOL) aerial vehicle that maintains communications with a human operator, for teleoperation and monitoring through UAVs (more information). Trainees use information-driven guidance, coordination, and control methods to perform VTOL operations robustly and maintain communications, while maximizing detection probability, and minimizing fuel consumption.
Point of contact: Thomas A. Wettergren, NUWC
Location: NUWC, Newport, RI
In this combined laboratory/field experiment, WISeNet trainees have access to high-fidelity sensing and environmental modeling tools developed by the Navy, and to a small fleet of unmanned underwater vehicles with on-board sensors operating in a safe and restricted coastal region at NUWC. Through an educational partnership agreement (EPA) on Sensor Technologies between Duke and NUWC, trainees can be granted authorized access to NUWC computational and experimental facilities and, consequently, be able to test new sensor modeling, coordination, and control methods, as well as new environmental modeling and prediction methods, using real Navy systems and software for ocean modeling, coastal surveillance, and defense applications.
In this experiment, WISeNet trainees test the biologically-inspired sensor fusion, learning, and adaptation methods on a robot comprised of a servo-mounted video camera, microphone, and sound card, equipped with a robotic arm. The camera sends the visual input to the robot’s computer (e.g. coffee mug), and the computer rotates the camera at saccadic velocities. The sensorimotor system is simulated using a neuronal sheet structure designed with the program Topographica, and the experiment is set up to examine presaccadic remapping, and mediation of our sense of visual continuity while we move our eyes. Trainees test new fusion and sensory-based motor control algorithms on the robot, and seek to show that adaptation and plasticity can be achieved in response to time-varying stimuli and environments. The Turing test and other experiments will be conducted to show that, using these novel approaches, the robot will be able to continually use its video and auditory inputs, while grasping and manipulating objects, even as saccades cause the video camera to lurch in arbitrary directions, and as the sensorimotor system is colocalizing visual and auditory inputs.
Point of contact: Ingo Fischer, IFISC (UIB-CSIC), Palma de Mallorca, Spain
Location: IFISC (UIB-CSIC), Palma de Mallorca, Spain
In this laboratory experiment, WISeNet trainees will work in an interdisciplinary environment to develop smart sensing solutions based on neuro-inspired computing. The trainees will get acquainted with valuable expertise on novel information processing schemes. The IFISC Research Institute aims at understanding the emergent cognitive functions of the brain by combining experts on neuro-science and complex systems theory. Trainees will be involved in developing novel approaches to Reservoir Computing using networks of electronic oscillators. They will experimentally implement and utilize state-of-the-art dedicated hardware solutions based on custom-designed electronic boards and FPGA platforms, which allow for autonomous operation and low power consumption. Different sensing and information processing tasks, including medical and auditory signal classification, will be explored. One goal beyond state-of-the-art is to introduce and implement plasticity rules into the electronic reservoir computing systems to mimic adaptability as present in the brain.
One of the key challenges facing wireless sensor networks is the energy cost associated with frequent communication. This experiment explores the utilization of radar systems for the purpose of ultra-low power communication and localization of sensors. In the simplest of terms, the idea of backscatter signaling is the modern radar equivalent of the 19th century optical “heliograph” wherein the sender used mechanical tilts of a mirror to modulate sunlight reflected to a distant receiver in patterns corresponding to Morse code. In our concept, the “sun” is replaced by RF transmitter, mechanical tilting of a mirror is replaced by extremely low-power digital antenna impedance load modulation circuits, and the optical observer is replaced by a microwave receive system. In this experiment, WISeNet trainees will develop multi-disciplinary methods intersecting radar theory, communications and coding theory. Trainees will have a unique opportunity to study, experience and measure communication effects and structured interferences and test their methods in an indoor environment.
Point of contact: Martin McGinnity, ISRC
Location: ISRC, University of Ulster, Magee Campus, Londonderry, UK
In this laboratory experiment, WISeNet trainees utilize a state-of-the-art cognitive robotics laboratory equipped with more than 30 mobile robots (Pioneer, Scitos and Robotino), the world’s largest electric floor -- which charges the robots -- and a high-precision VICON motion tracking system, and computer interface. Trainees implement new intelligent robotic sensor control and coordination algorithms on benchmark robotic games and competitions that may involve mobile robots, such as Marco Polo and pursuit-evasion games. Trainees also have the unique opportunity to test the robustness, adaptation, and feasibility of these methods in a real indoor environment, where these robots can cooperate and compete with each other, as well as with humans. State-of-the-art robotics software will also be available to control a Schunk 7DoF manipulator, and a Shadow Robot 5 finger hand. Trainees can thus develop biologically-inspired computational models and sensorimotor simulations for these robots, and investigate sensorimotor learning and adaptation in robot-human interactions, and in competitive settings for reward-accelerated learning.
Point of contact: Achim J. Lilienthal, Örebro University
A major challenge for environmental monitoring is dense sampling of airborne pollutants and particulate matter. In order to address current environmental challenges, higher and more adaptive sampling densities are needed. A current example of high significance is monitoring of fugitive methane emissions. Methane is considered to be a bridge fuel towards a decarbonized global energy system  but a switch to methane has a positive net effect on the global climate only if efficient monitoring systems allow to keep leakage rates low. Autonomous inspection robots equipped with gas, wind and other environmental sensors can play a major role in future monitoring solutions. On-board or external computational resources allow on-line fusion of all the measurements into a continuously updated environment model. This model can be used to promptly communicate consolidated results to human operators (e.g. pollution models, source location hypotheses, estimated total emission rates) and to identify the most informative measurement positions so as to achieve dense and adaptive sampling. In addition, autonomous inspection robots tirelessly carry out the required repetitive measurement procedure and minimize the exposure of human operators to hazardous compounds.
In this experiment we investigate gas distribution modelling, emission estimation and gas source localization with a mobile robot. We are especially interested in sensor fusion, data analysis and predictive modelling. WISeNet trainees get access to the gas-sensitive mobile robots available at the MR&O Lab, especially the Gasbot platform [http://aass.oru.se/Research/mro/gasbot/index.html]. In collaboration with the researchers at Örebro University and working with different sensing principles, trainees develop novel algorithms for gas identification, gas distribution modelling and gas source localization. In addition to laboratory tests, WISeNet trainees have the possibility to get access to sites where the developed algorithms can be evaluated in a realistic application scenarios. This includes monitoring of biogas emissions at a landfill site and pollution monitoring in foundries.
 A. R. Brandt, G. A. Heath, E. A. Kort, F. O'Sullivan, G. Pétron, S. M. Jordaan, P. Tans, J. Wilcox, A. M. Gopstein, D. Arent, S. Wofsy, N. J. Brown, R. Bradley, G. D. Stucky, D. Eardley, and R. Harriss, "Methane Leaks from North American Natural Gas Systems," Science, vol. 343, no. 6172, pp. 733–735, Feb. 2014.