Definition and Quantification of the Complexity Experienced by Autonomous Vehicles in the Environment and Driving Task

Research on complexity of environment and driving task is critical in testing and validating L3+ autonomous vehicles (AVs). We define the complexity of environment and driving task of AVs as the sum of uncertainties in the prediction process and propose a method to quantify it. Based on the analysis of the effect factors of perception system and the data processing algorithm of the cognition system of AV, and combining with complexity theory and the definition mentioned above, several types of environmental complexity factors liked the amount, variety, and path of motion of objects, the environment conditions are abstracted. Multi-factor analysis quantification is used to generate a quantifying model for the complexity of environment. A survey among five experts in the field of autonomous driving is conducted. The consistency between the average survey results and the quantification by using this methodology is 81%, which proves the effectiveness of this methodology.