This thesis considers the problem of a robot with complex dynamics navigating a partially discovered environment to satisfy a temporal logic formula consisting of both a co-safety formula component and a safety formula component. We employ a multi-layered synergistic framework for planning motions to satisfy a temporal logic formula, and we combine with it an iterative replanning strategy to locally patch the robot's discretized internal representation of the workspace whenever a new obstacle is discovered. Furthermore, we introduce a notion of closeness'' of satisfaction of a linear temporal logic formula, defined by a metric over the states of the corresponding automaton. We employ this measure to maximize partial satisfaction of the co-safety component of the temporal logic formula when obstacles render it unsatisfiable. For the safety component of the specification, we do not allow partial satisfaction. This introduces a general division between soft'' and ``hard'' constraints in the temporal logic specification, a concept we illustrate in our discussion of future work.

The novel contributions of this thesis include (1) the iterative replanning strategy, (2) the support for safety formulas in the temporal logic specification, (3) the method to locally patch the discretized workspace representation, and (4) support for partial satisfaction of unsatisfiable co-safety formulas. As our experimental results show, these methods allow us to quickly compute motion plans for robots with complex dynamics to satisfy rich temporal logic formulas in partially unknown environments.