Improving Environmental Sensing

Researchers seek to reduce the cost and improve environmental sensing efficiency by developing algorithms to deploy on drones.

Drone Flying through air

John Lipor, an assistant professor of electrical and computer engineering at Portland State University, recently received a five-year, $550,000 CAREER award from the National Science Foundation. The grant supports the development of algorithms to reduce the cost and increase the efficiency of using autonomous vehicles for environmental sensing.

In recent years, autonomous vehicles (e.g., drones) have revolutionized environmental science. Drones equipped with specialized sensors and other sample-gathering technologies provide scientists an affordable, easy-to-use means to gather and analyze environmental data in disciplines ranging from geomorphology to biodiversity and including domains such as air and water quality monitoring. Lipor's research focuses on developing technologies that improve scientists' ability to collect data while lowering the associated costs.

"The primary aim of our research is to improve efficiency and reduce the cost of environmental sampling for the problem of level set estimation," Lipor said.

According to Lipor, level set estimation seeks to find regions where a quantity of interest lies above some important threshold. For example, air quality scientists may wish to find all regions where particulate matter lies above the “unhealthy” threshold. The algorithms Lipor and his team plan to develop will address improvements in three level set structures used by scientists deploying drones to gather data. The first is boundary smoothness, where the boundary of the level set is assumed to vary by only small amounts over a small spatial region. The second is known similarity structure, where there may be similarities between one sample site and another that can guide autonomous sampling to help determine the extent of the similarities. And finally, unknown cluster structure, where information about each location can be incorporated to help guide sampling.

"By developing algorithms that address these factors of level set estimation, we believe we can address cost issues associated with determining where samples should be collected, and, thus, the time required to collect samples," Lipor said.

Among those who may benefit are scientists at the US Geological Survey (USGS). Lipor has partnered with Burns, a research hydrologist at the USGS facility in Portland, OR, on a project that aims to deploy machine learning to study sites with the potential to produce geothermal energy.

In addition to improving drone-facilitated environmental sensing, the project also promotes STEM education at all levels. Lipor has partnered with James Madison High School in Portland, where the project helps fund curriculum for a drone programming course for students. Undergraduate students at PSU will play a role in programming and testing the algorithms on drones. The project also provides funding for a Ph.D. student.

Beyond opportunities for students, the NSF award will support a citizens science 5K running/walking event for the community in Portland's Forest Park. The "5K to Nowhere," as Lipor calls it, will engage citizens in environmental sensing by having participants cover sections of the park's 70-mile trail system to identify conditions that impede trail navigability.


Related Links: