Since 2004, NASA has been working to return to the Moon. In contrast to the Apollo missions, two key objectives of the current exploration program are to establish surface infrastructure and an outpost. Achieving these objectives will enable long-duration stays and long-distance exploration of the Moon. To do this, robotic systems will be needed to perform tasks which cannot, or should not, be performed by crew alone. In this talk, I summarize our work at NASA Ames to develop "utility robots" for lunar surface operations, present results and lessons learned from field testing, and discuss directions for future research.