This thesis investigates whether simulation-based testing serves as a credible complement for evaluating selected automated driving functions. Two Advanced Driver Assistance Systems (ADAS) functions were studied: lane keeping and parking assistance. A literature-based mapping was first conducted to identify relevant evaluation aspects and reference values from previous research. Based on this mapping, both functions were implemented and tested in a steer-by-wire simulation environment built in Unreal Engine 5. The evaluation was structured around three shared performance dimensions: accuracy, stability, and responsiveness. For lane keeping, the system achieved a mean RMS lateral error of 8.29 cm on straight roads, which is consistent with the lower end of values reported in previous literature under comparable conditions. On curved roads, mean lateral errors increased to approximately 20-21 cm, where a systematic directional offset revealed the known limitation of a proportional controller lacking an integration term. For parking assistance, all 60 runs were completed successfully. The strongest final pose accuracy was observed at a 15 m start distance, with a mean lateral error of approximately ±3 cm and a mean yaw error below 1^◦. Lateral jerk RMS values ranged from 3.6 to 4.0 m/s³ across all parking scenarios, and mean obstacle response times ranged from 0.204 to 0.252 s. The results show that the simulation environment generated structured, repeatable, and measurable data for both functions, and that the three research questions were addressable through simulation-based evaluation. Responsiveness was the most consistent dimension across all tested conditions, while accuracy and stability were more sensitive to scenario variation. The study concludes that simulation-based testing is useful as an early-stage complement for evaluating selected ADAS functions, but cannot replace physical validation, since factors such as physical vehicle dynamics, sensor uncertainty, actuator delays, weather variation, and complex traffic behavior were not fully represented.