Driving in the Extreme: Sensing Drifting and the Expert Driver Response

Understanding the professional response to extreme driving events fosters the creation of novel safety procedures for autonomous vehicles (AVs), capable of avoiding and handling critical scenarios. This work proposes a high-frequency sensing system and a data processing methodology for capturing the vehicle dynamics and the associated expert human response in extreme driving conditions, namely, drifting. The system resorts to multiple sensors placed across the vehicle to extract data such as the vehicle’s position, acceleration, heading, and the driver’s steering and pedal inputs. The processing pipeline of each sensing component is presented, to extract the abovementioned features, and combine time series to detect the occurrence of drifting events, and the patterns that precede and succeed them. The system has been tested in various real-life racing environments, and the results demonstrate temporal and spatial coherence with the track challenges and with the intensity of the driving events and responses. The analysis unveils responsive driving behaviors based on the track layout with human response patterns upon critical scenarios. Future analysis of acquired and processed data will target using machine learning (ML) to model the driver’s expert behavior and contribute to enhancing road safety mechanisms.