Satellite communication involves uncertainties including those caused by the environment, regulation, user behavior, and estimation/prediction errors. Current Dynamic Spectrum Access (DSA) schemes do not provide a comprehensive mechanism to represent uncertainties in the decision making process. Designing a robust system to provide performance guarantees under different uncertainty factors in the advanced Low Earth Orbit (LEO) satellite communications will be useful. To address this, IAI and its collaborators at the University of California, Davis, propose to design RADIUS: Reliable and Adaptive Decision makIng under Uncertainty for Spectrum access. RADIUS combines spectrum sensing and sharing, situation awareness, offline/online learning, and active anti-jamming for LEO satellite communications with three types of users, primary users, secondary users and (cognitive) jammers. RADIUS quantifies the uncertainty caused by the environment, regulation, uncontrollable user effects, and prediction, by using either individual observation or collaborative approaches. Statistical modeling of spectrum dynamics and prediction under various sources of uncertainty in communications systems will be used. Reaction strategies will be developed that are robust or opportunistic to uncertainty, or can leverage uncertainty by anticipation. These strategies will be wrapped in reliable DSA algorithms to manage and mitigate uncertainty, based on either offline stochastic optimization or online algorithms. The algorithms' performance will be studied via analysis, simulation or emulation under realistic satellite system models. The feasibility of using software-defined radios (USRPs) to develop and prototype the proposed system will be investigated and tested on IAI’s wireless network emulation platform, RFnest. This approach integrates uncertainty quantification and uncertainty mitigation in common system architecture. The optimal conditions and use cases for each algorithm will be identified and integrated in an adaptive algorithm that switches between offline and online algorithms.