Echtzeitoptimierung für Ausweichtrajektorien : mittels der Sensitivitätsanalyse eines parametergestörten nichtlinearen Optimierungsproblems

Abstract

In this thesis an approach for the real-time computation of trajectories and controls for collision avoidance in cars traffic is presented. The approach handles the full complexity of a nonlinear model of the host car's dynamics and keeps it stable during the maneuver. A switching point optimization technique reduces the number of variables and constraints in the desired nonlinear optimization problem drastically. Via parametric sensitivity analysis of perturbed nonlinear optimization problems an approach for real-time computation of near-optimal trajectories is developed. This approach leads to the possibility of a rather simple implementation in a car but is still generic enough to cover numerous different collision scenarios. Examples from collision avoidance demonstrate the capabilities of the desired method.