Tunnel Boring Machines (TBMs) are powerful tools for tunneling and underground construction, which excavate material and install a segmental concrete tunnel liner for support.However, unknown ground conditions pose a significant risk to tunneling operations and any damage to the machine can be disastrous to a project. As such, there is a need for tools which look ahead of the TBM for potential hazards during tunneling, including water saturated zones, faults, boulders and metal pipes. Geophysical methods offer the capability to image ahead of tunneling in order to prevent damage to the machine or nearby infrastructure, thus improving tunneling operations. In particular, the DC resistivity method is useful because it is sensitive to a large range of conductivity variations in geological and man-made materials.The research presented in my thesis consists of three parts: (1) a laboratory study of a scale model TBM and tunneling environment, (2) a series of forward models studying different survey designs, and (3) the inversion and imaging of synthetic data under different assumptions. I introduce several new survey designs that attach DC resistivity electrodes on a probe or probes, which are then pushed into the earth in front of the machine each time excavation stops. My laboratory data and forward modeling results show that this method reduces interference caused by the metallic TBM body, and increases the distance ahead of the machine at which a target may be detected: depending on the specific survey design, the TBM influence is minimal once the probe is pushed 20% to 55% of the TBM diameter ahead of the machine and targets can be detected up to 60% TBM diameter away. Finally, I invert synthetic data to produce ahead-of-tunneling images using different amounts of prior information (e.g. TBM and host rock resistivity) and perform a time-lapse inversion, which has not been done for DC resistivity on a TBM before. I conclude the inversions with two types of comparisons to the true model and show that including more prior information decreases model error, but does not necessarily improve how well a target can be imaged.