Abstract: Conformant planning is the problem of synthesizing an executable sequence of actions such that the sequence can be executed from an initial state with uncertainty to achieve the specified goal without sensing. The effects of the actions may be non-deterministic. This dissertation investigates the utility of plan reuse in conformant planning. Our planning system is the first to reuse plans for conformant planning. Given a new conformant planning problem, the plan reuse system searches for the best candidate problem in the plan library. It then modifies the retrieved plan to solve the new problem. The ability to reuse plans from past problems may reduce the computational effort needed to solve the given problem from scratch. Ideas such as using domain-dependent knowledge to help problem matching, and a heuristic to decide whether to continue plan reuse, are also introduced by this dissertation. An empirical evaluation of the plan reuse system on more than 100 problems with various sizes from 7 benchmark domains is reported. In general, the plan reuse system outperforms highly efficient plan generative planners like Conformant-FF and MBP. When the plan reuse system is faster than these 2 planners, the maximum speedup is 3913. The reuse-based system is able to solve several problems that conformant planners MBP and Conformant-FF don't. Limitations of the reuse-based system and directions for future work are also discussed. This dissertation shows that plan reuse is an effective to scale up conformant planning.