A team of researchers, including four USC computer science professors, has been awarded a $20 million grant by the National Science Foundation (NSF) to fund the NSF Artificial Intelligence (AI) Institute for Advances in Optimization (AI4OPT). The institute aims to bring together artificial intelligence and mathematical optimization to make new research breakthroughs in automated decision-making--from sustainable ecosystems to supply chain operations.
The five-year grant, led by the Georgia Institute of Technology (Georgia Tech), will unite a multidisciplinary team of thought leaders from six universities across the country, working in close collaboration with public school districts, private companies and community leaders.
USC Computer Science Associate Professor Bistra Dilkina will serve as co-principal investigator for the grant. In addition to USC and Georgia Tech, the interinstitutional team includes researchers and faculty members from the University of California at Berkeley; the University of Texas at Arlington; Clark Atlanta University; and Spelman College.
The USC team will focus on developing optimization and AI tools to create sustainable urban environments and improve logistics and supply chains, while also incorporating ethical and fairness considerations. At the heart of the research is the idea that AI and optimization, both crucial developments in modern computational science independently, are stronger together.
"Truly merging the expertise coming from AI and optimization research can be the big step change we need in AI to solve societal problems by both improving scalability and widening applications," said Dilkina, co-director of the USC Center for AI in Society.
"Merging the expertise coming from AI and optimization research could be the big step change we need in AI to solve societal problems." Bistra Dilkina
"This new institute will lead fundamental AI advances with a sharp focus on benefiting progress on important applications, such as urban sustainability and resilience, while also building skills and enthusiasm for AI among diverse students from K-12 to college and graduate school. We are happy to bring USC's considerable expertise in AI, as well as strong commitment to diversity and outreach to this transformative initiative."
The institute has identified several key use cases, including controlled environment agriculture (CEA) systems to increase food system security for urban communities and warehouse optimization, to address bottlenecks in warehousing and last-mile delivery, using multi-agent planning and optimization systems.
"It is really all about decision making," said Professor Sven Koenig, who will lead a research thrust on creating the algorithmic foundations for the next generation of multi-agent systems in logistics, including warehouse logistics.
"So far, AI and operations research have developed different techniques and often study different applications. The institute is well positioned to take a big step toward developing a unified science of making good decisions in the context of large complex systems and their applications."
Additional USC team members are Assistant Professor Aleksandra Korolova, who will lead research focused on fair and ethical optimization and learning; and Satish Kumar, a research assistant professor at the Information Sciences Institute, who will analyze problems with a focus on theoretical computer science.
Fusing optimization and machine learning
Although optimization largely works behind the scenes, it powers everything from supply chains and flight routes to package deliveries and even disaster response and organ allocation. Without it, "the world would grind to a halt," said Dilkina.
Optimization is also at the core of the machine learning revolution. But, as the world moves faster and becomes increasingly interconnected and uncertain, optimization problems have become more challenging.
"Tackling problems at the scale and complexity faced by society today requires a fusion of optimization and machine learning, with the two technologies working hand-in-hand," said Dilkina, an expert in discrete optimization and machine learning.
"The institute will embark on efforts for much needed deep cross-fertilization between the AI and optimization, so we can see where these two sciences can help each other. For example, I will lead a research thrust focused on developing a new generation of optimization solvers by leveraging data-driven machine learning approaches to 'discover' new algorithmic policies."
A second significant component of the institute is to "inspire and empower a pipeline of future researchers and engineers that are a diverse representation of our society," said the researchers in their proposal, including bringing new education and research programs to Historically Black Colleges and Universities (HBCUs) and Hispanic-serving colleges in Georgia, California, and throughout the US.
This year, for the first time, the NSF program includes collaborations with major corporations-- the NSF Artificial Intelligence Institute for Advances in Optimization is partially funded by Intel Corporation, and industry collaborations represent a key constituency for the institute, driving methodology research with use cases that represent fundamental societal challenges.
The NSF Artificial Intelligence Institute for Advances in Optimization is one of 11 new NSF-led Artificial Intelligence Research Institutes, totaling a $220 million investment.
