This thesis considers a fixed cycle scheduling problem with periodically-arriving batches of partially-precedent tasks, where a constant number of parallel machines are available during a fixed cycle and each batch (job) can be completed on any machine by processing its including tasks. Under the assumption that each task requires unit period of processing time, the sum of delay time of all the arrived jobs is to be minimized. In the problem analysis, some solution properties for several types of precedence structures between tasks are characterized. These properties are then used to develop heuristic algorithms and a branch-and-bound algorithm.