In a recently started project, we are developing techniques for intelligent agent control and coordination in a dynamic, real-time, multi-agent setting. The application domain, consisting of teams of autonomous air vehicles (AAVs), is characterized by dynamic environments, real-time response requirements, limited information, and unreliable, low-bandwidth communications. We have developed an initial framework that models decision making, reasoning about constraints, and learning at multiple levels of abstraction and multiple time scales, within and across agents. Decision making in this framework is sensitive to communication availability and costs, tradeoffs among multiple objectives, and reliability of information about other agents (friendly and hostile) in the environment. We plan to focus our research on three key areas: Reasoning about communication constraints to inform action selection during planning and execution. Integrating deliberative planning, reactive planning, and continuous real-time control. Learning methods to improve performance within the AAV environment. The first of these, communication-sensitive decision making, is the focus of this research summary. In the next section, we describe the AAV application domain. We then outline our approach to communication-sensitive decision making and discuss related work.