Mobile Edge Testbed for Driving Behavior Data Collection and Cognitive Impairment Analysis

Studying cognitive impairment and its impact on driving behaviors is crucial for enhancing public safety. To facilitate cognitive impairment studies, we devise a testbed for real-world driving data collection using ubiquitous mobile edge devices (i.e., smartphones) [4]. Toward this end, we develop an application for autonomous data collection using smartphones. To enable robust data collection in real-world driving scenarios, we design a coordinate alignment method that automatically aligns the smartphone’s coordinate system with the vehicle’s by continuously detecting stationary and straight-line acceleration periods. We also design a two-step segmentation algorithm that first utilizes gyroscope readings to segment rotation-based behaviors (e.g., turning) and then employs accelerometer data to segment non-rotation-based behaviors (e.g., braking). The processed data is then uploaded to a cloud server through WiFi connections for further analysis. In this work, we show an example application of our testbed in studying cognitive impairment from substances such as cannabis. Cannabis use has become increasingly prevalent due to evolving legal and societal attitudes in recent years. The psychoactive compound tetrahydrocannabinol (THC) in cannabis is known to disrupt central nervous system functions, resulting in impairments such as reduced attention and slower reaction speeds [3, 6]. Although the effects of cannabis are beneficial in medical treatments, they raise significant concerns about its impact on public safety, particularly in the context of driving. Existing studies on cannabis-influenced driving mainly rely on driving simulators or pre-installed special equipment [1, 2, 5, 7]. However, simulation-based approaches cannot fully capture the complexity of real-world driving environments, including road texture, weather conditions, and dynamic traffic events. Additionally, specialized equipment, such as cameras or optical devices, incurs high costs and raises privacy concerns, making it challenging to achieve scalable deployment in real-world scenarios.