We present an algorithm, effective over a broad range of planet populations and instruments, for optimizing integration times of an exoplanet direct imaging observation schedule, to maximize the number of unique exoplanet detections under realistic mission constraints. Our planning process uses “completeness” as a reward metric and the nonlinear combination of optimal integration time per target and constant overhead time per target as a cost metric constrained by a total mission time. We validate our planned target list and integration times for a specific telescope by running a Monte Carlo of full mission simulations using EXOSIMS, a code base for simulating telescope survey missions. These simulations encapsulate dynamic details such as time-varying local zodiacal light for each star, planet keep-out regions, exoplanet positions, and strict enforcement of observatory use over time. We test our methods on the Wide-Field Infrared Survey Telescope (WFIRST) coronagraphic instrument (CGI). We find that planet, Sun, and solar panel keep-out regions limit some target per-annum visibility to <28  %   and that the mean local zodiacal light flux for optimally scheduled observations is 22.79 mag arcsec^−  2. Both these values are more pessimistic than previous approximations and impact the simulated mission yield. We find that the WFIRST CGI detects 5.48  ±  0.17 and 16.26  ±  0.51 exoplanets, on average, when observing two different planet populations based on Kepler Q1-Q6 data and the full Kepler data release, respectively. Optimizing our planned observations using completeness derived from the more pessimistic planet population (in terms of overall planet occurrence rates) results in a more robust yield than optimization based on the more optimistic planet population. We also find optimization based on the more pessimistic population results in more small planet detections than optimization with the more optimistic population.