The aim of determining fitness to drive is to achieve a balance between minimising any driving-related road safety risks for the individual and the community and maintaining the driver's lifestyle and employment-related mobility independence. Driving a car is a complex and dynamic task and there is a wide range of conditions that temporarily affect the ability to drive safely like consuming substances or fatigue. Professional drivers are particularly affected by fatigue. The main effect of fatigue is a progressive withdrawal of attention from the road and traffic demands leading to impaired driving performance. The particular practice of professional drivers includes working long hours, prolonged night work, working irregular hours, little or poor sleep, and early starting times which in many cases lead to fatigue. Fatigue causes reduced alertness, longer reaction times, memory problems, poorer psychometric coordination, and less efficient information processing. The results of different surveys world-wide show that over 50% of long-haul drivers have at some time almost fallen asleep at the wheel. The project will design, implement and test a new AI-based framework, for the monitoring and evaluation of driving performance, with particular regards to mental fatigue. The system will create neurophysiological models able to detect the onset of abnormal drivers' fitness based on data obtained while driving, in order to potentially trigger on-board intelligence. Artificial Intelligent models will associate different kinds of anomalous behaviour to its most probable cause: fatigue in particular, but potentially also stress, alcohol and drugs effects.
The project impact would be ground-breaking, since AI-DRIVE project by employing forefront data science and neuroscience methodologies will directly tackle road safety concerns.
The research work about driver's fatigue and drowsiness detection algorithm has great significance to improve traffic safety. Presently, there are many fruits and literature about driving drowsiness detection methods. However, these solutions are still not considered in the applied field mainly because of monitoring systems invasiveness, and the lack of standardized procedures. Current methods are devoted to find a universal drowsiness detection system, while ignore the individual driver differences.
The work of all agencies working within road safety is limited by the lack of objective and multidimensional data, statistically valid, on driving while impaired, and the effect of collateral variables such as age, substances consumption, etc. AI-DRIVE will pave the way for a new approach linking the "unfit" causes (i.e. alcohol, drugs, fatigue, stress), the immediate hidden unconscious psychophysiological alterations (changes in brain activity, cognitive capacities, etc.), and the resulting unsafe driving behaviours. This project will propose a real-time driving "fitness" monitoring algorithm (validated with fatigue and to be exploited with other mental impairments) able to profile individual behavioural patterns and to detect eventual deviations from the "normal habits".
Only a real-time driver physiological monitoring technology could recognise cues of the incoming impairment, thus alerting and/or supporting the driver timely. AI-DRIVE research will contribute to realize a ground-breaking device able to detect and even predict sudden driver's impairments and will pave the way for a new generation of devices for the driver mental state measurement, able to be fully embedded on future vehicles.
Besides this clear short-term impact, AI-DRIVE would enable also long-term implications with respect to the field of Automated Vehicles (AVs). In fact, AVs do not imply the end of road traffic accidents. Crashes will still occur, particularly during the long transitional period to full automation, if the relationship between the driver and the vehicle is not improved. In fact, autonomous vehicles able to drive in SAE 3 level are expected to shortly enter into the market: the characteristics of such a level (the driver should take the control when requested by the system) need different degrees of human attention in a single trip, according to the traffic/infrastructure/weather conditions. Let us consider for example that the mortal accident occurred while testing the Tesla AV has been demonstrated to be caused also by a human error: because of a loss of situational awareness, the test driver was not able to counteract to a wrong autopilot manoeuvre. For this transition to occur safely, it is imperative that drivers react in an appropriate and timely manner. It is therefore important that the car's system is able to know and monitor driver's fitness conditions, in order to assess his/her ability to correct car decisions or predict driving performance and eventually take proper countermeasures. These may range from audio signals, flashing visual cues adapted to catch the driver attention, up to a safe stop of the vehicle on the right border of the road in the worst cases. Starting with SAE-level 2, psychological constructs such as vigilance, fatigue, drowsiness, sleepiness and how they are interlinked with the overall driver state, play an important role for all efforts in improving road safety. In SAE-level 2, the driver needs to monitor the vehicle's guidance as well as the surrounding environment, whereby the automation requires the human to intervene immediately, if a system limit is met. However, basic psychological research indicates that these vigilance tasks are very stressful and accompanied with human failures. Laboratory experiments show that humans miss stimuli after about 15 minutes using standardized vigilance tasks. One possible explanation could be an increasing fatigue level, caused by monotonous tasks transmitting so called passive task-related fatigue. The AI-DRIVE technology could be easily integrated in the current and future Human Machine Interfaces (HMI) of semi-autonomous and autonomous vehicles and provide this relevant function.
Lastly, the possibility of monitoring driver's fitness, including his/her cognitive and perceptive abilities, while driving would produce relevant benefit for the whole automotive industry in the design phase: the recent disciplines of Neuroergonomics and the concept of Human-Centered Design will allow industry to design car interiors and instrumentation actually able to improve drivers' cognitive performance (e.g. devices easy-to-use, alarms in the optimal position to catch drivers' attention, lights colour and sounds not inducing stress, etc.).