Teleoperation has been suggested as a means of overcoming the high uncertainty in automated driving by letting a remote human driver assume control of an automated vehicle when it is unable to proceed on its own. Remote driving, however, suffers from restricted situational awareness and network latency, which can increase the risk for accidents. In most cases though, the automation system will still be functional, which opens the possibility for shared control where the human and machine collaborates to accomplish the task at hand. This thesis develops one such method in which an operator demonstrates their intentions by maneuvering a virtual vehicle in a virtual environment. The real vehicle observes the virtual counterpart and acts autonomously to mimic it. For this, the vehicle uses a genetic algorithm to find a path that is similar to that of the virtual vehicle, while avoiding potential obstacles. Longitudinal control is performed with a method inspired by Adaptive Cruise Control (ACC) and lateral control is achieved with a model predictive controller (MPC). The method is evaluated with user trials in simulation and is seen to perform better in the presence of large delays when compared to a direct control approach. With negligible delay, it performs on par with or slightly worse than direct control, but test participants still reported higher confidence when using the new method as well as an overall preference for the same. ; Teleoperation har föreslagits som metod för att hantera den höga osäkerheten som förekommer hos självkörande fordon. När de inte klarar av att på egen hand planera sin rutt kan kontrollen istället överföras till en fjärrlokaliserad förare. Direkt fjärrstyrning innebär dock flera nackdelar i form av exempelvis begränsad lägesuppfattning och nätverksfördröjningar, vilket kan innebära en förhöjd risk för olyckor. I det flesta fall kommer dock fordonets automationssystem fortfarande vara aktivt, vilket öppnar för att använda delad styrning där människan och fordonet samarbetar för att ...