Abstract: Mission and flight planning problems for uncrewed aircraft systems (UAS) are typically large and complex in space and computational requirements. In size, weight, and power (SWaP) constrained UAS complex decision problems may need to be computed offline and then executed during flight. But dynamic environments and responsiveness to changes in the mission demand online adaptation and replanning. In this talk I will present a novel approach to solving Markov Decision Process (MDP) problems at runtime in practical planning applications for SWaP-constrained UAS. To do this we leverage information derivatives of nearby states to determine how to solve the MDP. The approach allows for dynamic decision making and replanning by a UAS in an uncertain environment by incorporating new information or objectives into the MDP model.

Short Bio: Justin Bradley is a Richard L. and Carol S. McNeel Associate Professor in the School of Computing at the University of Nebraska-Lincoln. He holds a B.S. in computer engineering (2005) and M.S. in electrical engineering (2007) from Brigham Young University, and M.S. (2012) and Ph.D. (2014) degrees in aerospace engineering from the University of Michigan. He has worked with Uncrewed Aircraft Systems (UAS) for over 17 years, starting at the MAGICC lab at BYU, the A2Sys lab at the University of Michigan, and most recently as a co-director of the NIMBUS lab since 2015. He is a recipient of a 2021 NSF CAREER award and is an AIAA Associate Fellow.