Multi-Level Driver Workload Prediction using Machine Learning and Off-the-Shelf Sensors

Abstract

The present study aims to add to the literature on driver workload prediction using machine learning methods. The main aim is to develop workload prediction on a multi-level basis, rather than a binary high/low distinction as often found in literature. The presented approach relies on measures that can be obtained unobtrusively in the driving environment with off-the-shelf sensors, and on machine learning methods that can be implemented in low-power embedded systems. Two simulator studies were performed, one inducing workload using realistic driving conditions, and one inducing workload with a relatively demanding lane-keeping task. Individual and group-based machine learning models were trained on both datasets and evaluated. For the group-based models the generalizing capability, that is the performance when predicting data from previously unseen individuals, was also assessed. Results show that multi-level workload prediction on the individual and group level works well, achieving high correct rates and accuracy scores. Generalizing between individuals proved difficult using realistic driving conditions but worked well in the highly demanding lane-keeping task. Reasons for this discrepancy are discussed as well as future research directions.

HeartPy: A novel heart rate algorithm for the analysis of noisy signals

“…The algorithm was validated using two datasets from two different experiments and research domains. The first dataset used was collected with a low-cost PPG sensor in a driving simulator experiment (van Gent et al, 2018). This dataset contains approximately 20.7 h of PPG recordings.…”

“…This makes analysis more difficult. In the traffic domain, PPG sensors have been used by for example (Jarvis, Putze, Heger, & Schultz, 2011;van Gent et al, 2018;Zhai & Barreto, 2006).…”

“…Documentation is available through the GitHub page. The wearable embedded C version is available on GitHub as well ( van Gent, 2018), together with documentation linked there.…”

“…Algorithms are being developed that can estimate the driver's state, whether this is driver workload (Solovey, Zec, Garcia Perez, Reimer, & Mehler, 2014; van Gent, Melman, Farah, van Nes, & van Arem, 2018), driver distraction (Liang, Reyes, & Lee, 2007) or a driver's interruptibility (Kim, Chun, & Dey, 2015). Heart rate is frequently included as an input for predicting a driver's state since it contains information about changes in (driver) workload (Mehler, Reimer, & Coughlin, 2012;Mehler, Reimer, Coughlin, & Dusek, 2010), stress (Healey & Picard, 2005), and general driver state such as drowsiness (Danisman, Bilasco, Djeraba, & Ihaddadene, 2010).…”

“…In a previous study we collected heart rate data with low-cost sensors to develop an affordable driver workload estimation approach (van Gent et al, 2018;van Gent, Farah, van Nes, & van Arem, 2017). Available open source algorithms did not function well on this type of -often noisy-data and couldn't easily be integrated into a real-time system.…”