As people are starting to face the prospect of robots becoming part of our everyday lives, it has become increasingly clear that the information robots could gather can be both sensitive and valuable. But the robots may need to gather this information in order to function properly: as elsewhere in our lives, we need to understand how to best reconcile the tension between utility and privacy. The scientific progress made to date on algorithms for planning, control, and coordination of multi-robot systems has been enormous, but it also has paid too little attention to "who knows what." This research effort sets out to understand how the essential computational operations underlying many common robotic tasks can be safely accomplished in circumstances where there is some doubt about the integrity of other elements in the system, including whether they can be trusted to never expose information. This is crucial for autonomous robots operating within socially sensitive settings, as well as contested or adversarial scenarios. Beyond the anticipated impact on robotics research, the project will benefit society by addressing questions of strategic national interest and help facilitate privacy protections. It includes education and outreach activities that serve underrepresented groups, firstly via direct engagement with undergraduate and graduate students at Florida International University and Texas A&M, and secondly, in working with and mentoring high school teachers.
The project will conduct both theoretical and empirical research, through a multi-part research agenda that will enable privacy-preserving filtering and planning in multi-robot scenarios via secure multi-party computation methods. This research endeavor represents a radical departure from present computational assumptions for robots: it aims to introduce abstractions, algorithms, and systems to solve robot tasks in scenarios characterized by collaboration between mutually distrusting robots, this is the first systematic effort to do so. The research will allow multiple robots to coordinate their use of shared resources, without divulging sensitive information that each robot possesses, despite their effective cooperation actually depending on that sensitive information. The research program will produce: (1) a new set of geometric primitives that allow the solution of motion planning (and related) problems in a privacy-preserving fashion; (2) novel filters, constructed to preserve privacy; and (3) constructions for calculating and querying computational topology properties, subject to limits on information shared. Together, these pieces lay groundwork, establishing the research area of secure multi-robot computation.
This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.