CITRIS selects 8 multicampus projects for 2022 Seed Awards

Collage of five images representing research projects: An older person wearing a hat, viewed from straight ahead with blurry side views to the left and right; an excavator in the middle of a landfill; a hovering drone with trees in the background; a forest on fire; an electric car connected to an outlet.

The Center for Information Technology Research in the Interest of Society and the Banatao Institute (CITRIS) at the University of California (UC) are proud to announce the 2022 CITRIS Seed Awards recipients. Since 2008, the CITRIS Seed Funding Program has supported more than 240 early-stage, interdisciplinary research projects that show promise to shape the future of their fields.

The eight selected proposals, submitted by multicampus teams from Berkeley, Davis, Merced and Santa Cruz, will receive up to $60,000 for their work, thanks in part to external philanthropic support. The winning projects, which are designed to show results within one year, address various challenge areas within the information technology sector, including aviation, climate resilience, digital health and robotics. 

Sustainability proved to be a common area of interest, as half of the selected projects aim to make advances in energy storage or wildfire mitigation. Three projects will also use uncrewed aerial systems, otherwise known as UAVs or drones, to collect data and monitor terrain.

The awardees represent a diverse array of backgrounds, with 88 percent of the research teams including women or people of color. Over half of the research teams include a pre-tenure faculty member, and 75 percent of the principal investigators are new CITRIS Seed Award recipients.

“The 2022 Seed Award projects represent innovative ideas that have a high probability of meaningful application for the benefit of society in the near term and could shape the future of technical research in these fields for years to come,” said Costas Spanos, director of CITRIS and the Banatao Institute and the Andrew S. Grove Distinguished Professor of Electrical Engineering and Computer Sciences at UC Berkeley. 

The following proposals received 2022 awards:

Activity monitoring to improve caregiver connection and care for older adults living alone with Alzheimer’s disease
Principal Investigators: Alyssa Weakley (Lead PI, UC Davis Health), Shijia Pan (UC Merced), Hao-Chuan Wang (UC Davis)

Alzheimer’s disease is expected to affect 1 in 9 people over the age of 65 in the United States by 2050. Many patients with dementia currently live alone, and the number of people who take on remote caregiving responsibilities will steadily increase in coming years. Family members who provide care from a distance often must balance these duties with their roles as parents and employees, increasing their stress in a situation already rife with challenges. This project intends to create a cohesive digital platform to help long-distance caregivers monitor their loved ones’ everyday activities, such as eating, cleaning and taking medication, and to communicate with them about the data. The technology will use vibration sensors and machine learning to detect deviations from regular behavior, and will provide an interactive data visualization and communication tool, with the ultimate goal of keeping family members with dementia safely in their homes and preventing expensive and burdensome crisis-driven care.

Battery health degradation for electric off-road vehicles
Principal Investigators: Shima Nazari (Lead PI, UC Davis), Scott Moura (UC Berkeley)

About half of the carbon emissions created by construction operations come from heavy machinery such as loaders and excavators, and these emissions disproportionately affect dense, low-income communities of color. Electrification is one of the most promising pathways to addressing this source of pollution. However, the high price and limited battery life of current electric construction equipment impedes its widespread adoption. This project will develop system-level models for excavators, to better understand their duty cycles, and for the batteries themselves, to estimate their life spans under different operating conditions. The researchers will also offer alternative energy storage designs to improve the performance, affordability and sustainability of electric construction vehicles.

BrightBlue: Customizable, decomposable electrical energy storage
Principal Investigators: Eric Paulos (Lead PI, UC Berkeley), Jennifer Parker (UC Santa Cruz)

Nearly all interactive technologies use electrical power, but energy storage is often an afterthought in the design process. This makes unsightly, unsustainable, strung-together battery packs all too common on prototypes, wearable electronics and similar applications. This project will develop and test “Vims”: low-power supercapacitors that can charge rapidly and last for hours, made out of inexpensive, renewable, decomposable and even edible materials, such as graphite, glycerol, egg whites and table salt. The research team will also investigate how Vims could be used across health, fashion, food and other sectors, particularly within the realm of wearables.

Human-drone-robot teaming for wildfire detection: Technology and workforce development
Principal Investigators: Alice Agogino (Lead PI, UC Berkeley), Becca Fenwick (UC Santa Cruz)

More than 50 million households in the United States are under threat from wildfires due to their presence in the wildland-urban interface (WUI), where wilderness transitions to the human-built environment. This risk is exacerbated by a historic labor shortage within fire response organizations. While a record amount of data is now available to help fire professionals battle wildfires, they lack a streamlined method to obtain and use the information to create more effective plans. This project has two aims: to develop a system of sensor-equipped robots that can be deployed from drones to collect data on wildfire-prone areas, and to build an educational program that trains fire response professionals to pilot drones and use the survey data in their work and planning. These efforts will serve to improve emergency response time, and to strengthen the firefighting workforce by helping its personnel harness emerging technologies.

Joint UAV- and robot-optimized approach to quantify methane emissions and energy losses from landfills 
Principal Investigators: Dimitrios Zekkos (Lead PI, UC Berkeley), Stavros Vougioukas (UC Davis)

Methane is a powerful greenhouse gas and a major driver of climate change, and municipal solid waste landfills are the third-largest source of methane emissions caused by human activity. Current methods of measuring methane emissions from landfills are sparse and infrequent, and likely lead to underestimations. This project will use UAVs, autonomous ground robots and topographical models to create a constantly updating map of landfill methane concentrations. The data will offer more accurate measurements of methane levels to help verify the extent to which landfills are contributing to climate change. The sensor system will also pave the way for “smart landfills,” where methane leaks can be easily detected and captured to generate sustainable energy.

Restoring speech communication with a multimodal decoder-synthesizer
Principal Investigators: Lee Miller (Lead PI, UC Davis), Daniel Cates (UC Davis Health), Ahmed Arif (UC Merced)

More than 1 in 100 people worldwide have lost their ability to produce natural and comprehensible speech — a condition called dysarthria — due to cancer, stroke or another cause.  Right now, there is no technology to truly restore speech, only a few inadequate workarounds. This project will develop an assistive device that combines recordings of a person’s facial expressions and muscle movements and uses neural networks to synthesize and produce fluent speech in their own voice. The user will be able to move their mouth silently, as if they were speaking, to activate the device and generate their voice. This decoder-synthesizer will help restore quality of life for the tens of millions of individuals with dysarthria, allowing them to more easily and effectively interact with their co-workers, friends and loved ones.

Toolkit for the assessment of ecosystem resilience and functional diversity in fire-affected landscapes, using remote sensing data
Principal Investigators: Gary Bucciarelli (Lead PI, UC Davis), Todd Dawson (UC Berkeley), Andrew Latimer (UC Davis), Shane Waddell (UC Davis), Derek Young (UC Davis)

Although wildfire’s devastation may be felt most keenly in its destruction of physical and social infrastructure, it radically changes natural environments as well. At present, only a few resources are available to help experts assess whether fire-damaged ecosystems will recover. This project will create new drone- and satellite-based remote sensing tools to collect data about vegetation at six sites in the UC Natural Reserve System that were damaged by wildfires in 2020. The research team will use this information, as well as aerial and ground survey data previously collected at the sites, to train a neural network to estimate the post-fire resilience of each ecosystem. The team will also build an open-source toolkit to instruct other research groups on how to use drone data to conduct similar ecological assessments.

Trust aware human-machine teaming using real-time neurophysiological data
Principal Investigators: Kosa Goucher-Lambert (Lead PI, UC Berkeley), Zhaodan Kong (UC Davis)

Trust is essential for effective collaborations between humans and machines — particularly those involving artificial intelligence. Misaligned trust can have terrible consequences, such as the car accidents caused by an overreliance on vehicle autopilot systems. This project will develop a real-time measurement of human-machine trust by recording physiological signals in the brains of experiment participants as they interact with high- and low-performing robots in a tool-sorting task. The participants will rate their levels of trust in the robots, and the research team will use the data to build and compare different trust-prediction models. This work will set the foundation for research on trust in more complex interactions between people and machines.