Persistent Mission Planning and Control for Renewably Powered Robotic Systems

The objective of this research is to pioneer new techniques for the mission planning and control of robotic systems that derive their propulsive energy either solely or primarily from renewable resources. Examples of renewably powered robotic systems include tumbleweed rovers for remote terrestrial exploration, solar powered aircraft for aerial observation, and sailing drones for oceanographic surface exploration. By relying on renewable resources for propulsion, such systems can explore hostile and remote regions that cannot be explored through traditional mobile robots due to range limitations, including the surfaces of distant planets, deep waters of the ocean, and the Arctic region. This research effort will focus on the creation and validation of fundamental tools for controlling renewably powered systems through a propulsive resource that varies stochastically in space and time, thereby necessitating a fundamentally new set of control tools relative to traditional mobile robots. The theoretical results from the research will be validated on a small fleet of sailing drones, to be deployed in inland waters. The research effort will be complemented with educational and outreach opportunities involving Autonomous Marine Systems, Inc., the Carolina Sailing Club, and North Carolina State University.

Traditional mobile robotic systems can typically be characterized by very limited range but relatively predictable mobility, wherein the reachable domain of the robotic system (or team of robotic systems) can be characterized with a high degree of certainty at any given time. This research fundamentally reverses that paradigm, focusing on robotic systems with unlimited range but stochastic mobility. Due to the stochastic, spatiotemporal variation in the renewable resource, the application of traditional energy-aware control techniques on such systems will typically result in either ineffective or extremely conservative mission planning strategies. To address this challenge, this research project will pursue a hierarchical mission planning and control framework in which an upper-level mission planner prescribes preferred exploration directions based on statistical characterizations of the propulsive resource, and lower-level dynamic mobility optimizers will maximize expected mobility along preferred directions, taking into account the dynamics of each agent and the stochastic resource model. Gaussian Process modeling will be used to characterize the spatiotemporally evolving resource. Polynomial chaos approximations and stochastic response surface methods will be used to facilitate a receding horizon optimization of search directions at the upper level, whereas stochastic dynamic programming results will be used to extract probabilistically time-optimal waypoint following algorithms at the lower level. Theoretical performance limits will be analyzed in the context of statistical regret bounds. Mission planning and control algorithms will be validated in two settings: (i) a small fleet of instrumented sailing drones to be tested in inland waters and (ii) a larger-scale simulation study wherein the goal of the sailing drones is Gulf Stream resource assessment.

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.