To quantify human driving behavior, comparative metrics such as velocity and acceleration profiles, Drive Quality Evaluation, pedal busyness, and energy consumption were selected. The analysis was conducted using a dataset with 280 measurements obtained from human-driven WLTC type approval tests, covering all used types of powertrains (BEV, ICE, PHEV, THEV). It was confirmed that the deviations from the mean of the measurements at each sampling point of the cycle follow a normal distribution, enabling the application of the empirical rule to the data. Based on this, a tolerance range was developed around the mean curve for each comparative metric, effectively filtering out outliers. If the robot stays within all tolerance ranges, its behavior is confirmed to closely resemble that of a human driver. The latest developments steps deliver a comprehensive taxonomy for evaluation of a certain control behavior in terms of “human-like”-criteria and the detection, if the control behavior is systematically better or worse than the average human driving behavior. The control algorithm of the driving robot was parameterized and optimized. Specific driving phenomena and dynamic powertrain effects are discussed and the related improvement in the robot’s control behavior is demonstrated. Based on the comparative metrics, the system's ability to replicate human-like driving behavior was subsequently evaluated. ..activities were conducted both on chassis dyno testbeds and in simulation...