The thesis is concerned with online mapping and scheduling of tasks on multiprocessor embedded systems in order to improve the reliability subject to various constraints regarding e.g. time, or energy. To evaluate system performances, the number of rejected tasks, algorithm complexity and resilience assessed by injecting faults are analysed. The research was applied to: (i) the primary/backup approach technique, which is a fault tolerant one based on two task copies, and (ii) the scheduling algorithms for small satellites called CubeSats. The chief objective for the primary/backup approach is to analyse processor allocation strategies, devise novel enhancing scheduling methods and to choose one, which significantly reduces the algorithm run...