Abstract : We plan a long-term project schedule for which the total budget depends upon the year the project finishes. Each task in the project can begin only when all its predecessor tasks have been completed, and each task has a range of feasible durations with a month-by-month cost profile for each duration. A task start can be delayed, but once started for some chosen duration, a task cannot be interrupted. Each task suffers some risk of delay and changed cost. Ignoring budget constraints, we use Monte Carlo simulation of the duration of each task in the project to infer the probability distribution of the project completion time. We then optimize a deterministic project schedule following budget guidance. Finally, we successively reschedule as the project progresses, simulating annual review of active tasks, and possibly delaying each active task's duration and changing its monthly costs for its forecast duration. We do not require an independence assumption, so we can accommodate learning effects from completed tasks. U.S. Army Future Combat Systems (FCS) is our motivating application. FCS is a complex of information technologies, sensors, and command systems expected to require more than a decade and $16 billion to develop. The U.S. General Accounting Office finds FCS at significant risk of cost and schedule growth, and suggests two alternatives to a baseline Army plan. We analyze these three alternate project plans for FCS to discover which one can most likely be completed soonest and cheapest.