Erweiterung eines Pilotenassistenzsystems für Hubschrauber um einen parallelisierten Algorithmus zur Kollisionsvermeidung und Bahnfolgeregelung

Kurzfassung

This thesis deals with the efficient implementation of a trajectory control with collision avoidance for a helicopter. In a closed-loop simulation, trajectories are predicted using a non-linear model. It is then possible to evaluate the predicted trajectories using various evaluation criteria such as collision avoidance, vehicle control and path tracking. A control input for the helicopter is then generated from these evaluated paths. This sampling based Model Predictive Control (MPC) algorithm is implemented using CUDA (NVIDIA's GPU programming platform). In this work, a focus is placed on the performance of the program. For this purpose, different algorithms for sub-problems are discussed in order to increase the performance of the implementation.