A hierarchical control approach to semi-autonomous systems is presented in the context of RoboFlag, an experimental testbed. The proposed hierarchy includes research at several levels, including estimation and path planning, strategic and resource planning, and operator interface and decision making. The approach is unique in its use of formal techniques at the lower levels of the hierarchy, optimized/randomized techniques at the higher levels, each with appropriate information fusion and requirements up the hierarchy. Specific plays are developed as a function of the baseline technologies. An overview of the approach, specific technologies developed for plays, and demonstration on the RoboFlag game are presented.