Driver Reactions to Automated Vehicles: a practical guide for design and evaluation
15th October 2018
Do not be misled by the title; it is not about how drivers of non-automated cars react to driverless vehicles. The research the authors report on looks at the drivers' reactions in automated vehicles: how quickly they respond to warnings and how the driver/system interface can best be managed to ensure quick and safe responses.
The book pays special attention to vehicles with an intermediate level of automation -- those driven with hands and feet off the controls but where the driver might need to intervene at any time to take back direct manual control. For safety reasons these handovers are crucial, especially when drivers need to resume control quickly.
As the authors explain, the book covers situations in which either the driver is not concentrating on the road because the driving is automatic or the automated system has decided it can no longer cope.
Unsurprisingly perhaps, the research found that visual warnings do not provide a sufficient safeguard; audible as well as "haptic" (tactile) prompts are needed too. It also found response times vary from one driver to another and generally depend on the degree to which they are distracted by other things.
Another important conclusion is that the most effective interfaces direct drivers towards a safe action (such as, "it's OK to change lane now") rather than just warning about an approaching hazard -- a stationary or slow-moving vehicle, for example.
The book draws on evidence from aviation showing how sudden autopilot disconnection can leave pilots with only minutes to act before the situation becomes irretrievable -- and fatal. Under pressure, they may do the wrong thing, especially if the instruments give them misleading, ambiguous or excessive information. In the more crowded environment of the public highway, the time available to understand and respond to what is happening could be mere seconds.
A key strand of the research is the use of driving simulators to test expected driver behaviour in a low-risk environment. The book explains how this approach was developed and validated. The conclusions are vital; they show that driving in simulators is a reliable predictor of how people respond in real life. Reassuringly, driver response times in real cars proved slightly better than those in the simulator.