The objective of this work is to describe guidelines for measuring degraded human performance based on driver state and competences from a real-time driver monitoring perspective. The guidelines integrate state-of-the-art knowledge from the literature with knowhow from the industry and practical results from the Mediator project. The formulated guidelines are defined based on functionality, technological possibilities, safety relevance and feasibility.
In summary, an ideal driver monitoring system should have the following general features:
- Minimally obtrusive sensors. Camera-based systems have several advantages here, since they have the potential to capture rich information about humans, objects, and their interaction. Unobtrusive sensing is needed to facilitate high adoption rates, to avoid deactivation, and to avoid interfering with drivers’ operation of the vehicle.
- Real-time operation and timely detections. Impairment detection, and subsequent interventions, have different demands on acceptable latencies. Detection of early signs of fatigue is not time critical (order of minutes) while severe fatigue, microsleep, and long offroad glances are time critical (order of seconds or less). In some situations, discomfort can be detected offline several minutes in advance, for example when approaching harsh weather or a traffic jam. Proactive impairment interventions, in contrast to reactive detection/intervention, is favourable. This requires forecasts of drivers’ future readiness levels.
- Robustness to environmental conditions. System performance should not be significantly influenced by environmental conditions such as traffic, landscape, weather, and darkness.
- Automation level dependent. The drivers’ responsibilities change with the level of vehicle automation, which in turn affects the requirements for a driver monitoring system. As an example, continuous distraction detection is highly relevant in manual and assisted driving. In higher levels of automation, where non-driving related task engagement is allowed, it is sufficient to ensure that the driver is attentive in relation to transitions of control.
- Situational awareness. A driver/vehicle-unit should have sufficient situational awareness to be able to drive safely. With higher levels of automation, the responsibilities for situational awareness are gradually shifted from the human to the vehicle. Similarly, to be able to provide relevant impairment detections, a driver monitoring system should also be situationally aware and take contextual factors into account. For example, fatigue warning systems would benefit from knowledge about sleep history and driving time, and distraction detection systems would benefit from knowledge about which areas in the surroundings that needs to be sampled to gain sufficient situational awareness.
- Ecological validity. Final evaluations/testing of driver monitoring systems should be conducted in ecologically valid settings with naturalistically induced impairments. Lab testing can and should be used in earlier evaluation stages, for example, when testing if an eye tracking system provides high quality tracking throughout a broad range of the population.
- Minimal intrusion on privacy. Driver monitoring systems should avoid privacy intrusions. For example, video data should be deleted continuously and should not be stored beyond what is needed for impairment detections.
Since MEDIATOR has focused on driver distraction, driver fatigue, and driver comfort, the guidelines are restricted to these three states. In addition to comfort, distraction, and fatigue, many researchers, legislators, developers, and users also mention sudden sickness and intoxication as important impairments. These impairments have therefore been covered as well in the state-of-theart review and in the interviews, but not in the actual guidelines.
The goal of driver monitoring is to increase road safety. Achieving this goal depends not only on the performance of the driver monitoring system, but also on the intervention strategy and how the intervention is communicated to the driver. Guidelines on intervention strategies established in the Mediator project are described in van Grondelle (2023).
Though the formulation of definitive operational guidelines for driver monitoring systems still suffer from a lack of knowledge, this should not prevent or delay the introduction and implementation of such systems. Instead, available technologies should be used to address and mitigate impairments to the extent possible, starting with severe behaviours such as incapacitation, alcohol intoxication, microsleeps, and long glances away from the road.
MEDIATOR has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement No 814735.
