Micro-mobility modes such as e scooters are gaining increasing popularity, but e-scooter riders have high crash/injury risks. Virtual assessment of traffic safety measures, such as automated driving functions, requires valid models of all road users, including their interactions and responses to threats. This paper focuses on process models of threat response and crash avoidance of e-scooter riders, a crucial aspect of safety assessment. The response latency and the capability for crash avoidance vary considerably among e-scooter riders. We propose threat response models involving cognitive, dynamic and kinematic subprocesses. These models define key parameters that determine threat response characteristics. To capture the individual variability of these features, we present a study in which 36 subjects performed e-scooter riding tasks designed to reproduce the subprocesses required for crash avoidance. The measurements were supported by a semantic questionnaire. The key measured parameter distributions included response times, longitudinal decelerations, lateral accelerations, attainable curvatures and times required for countersteering. Although subjects were clearly motivated to realize their best performance in the e-scooter riding tasks, the proportions of poorer performances (e.g., high latencies, slow braking, or inadequate dodging maneuvers) and their magnitudes are of crucial importance for simulative assessment of safety impacts of automated driving or other traffic measures. The identified characteristics can be used in traffic simulations to represent e-scooter behavior in general and, in incipient conflict situations, to simulate threat response processes, including the failure of braking or dodging maneuvers, with a high degree of realism.
