This thesis discusses the topic of high-performance embedded systems, their widespread usage, and the need to optimise their hardware utilisation concerning energy consumption and time. Throughout the thesis, various techniques are suggested to execute problems on less powerful hardware, thus saving resources and expenses. Alternatively, these techniques can be utilised to execute more complex problems using the same hardware.First, the measurement setup used throughout the dissertation is detailed, as well as a set of experiments that determine the importance of the sampling rate for accurate energy measurement of high-performance embedded systems. Second, a novel energy model, system model, and energy-aware scheduler are introduced. They outperform existing schedulers. Third, various ranking algorithms are explored and compared. Fourth, a new task model is presented, which divides tasks into multiple phases, allowing for a more fine-grained separation of workloads. Lastly, a ranking-independent Reinforcement-Learning(RL)-based scheduler is presented. While performing similarly to a greedy heuristic, experiments show that including Graph Convolutional Neural Networks in the RL-scheduler improves the final schedules significantly.