As vehicles become more autonomous, it is crucial to also take a closer look at technological deficiencies that will inevitable occur. Such issues will undoubtedly occur when entering SAE level 3, where the driver would be allowed to drive hands-off and eyes-off. The official definition for Level 3 (“eyes off”) according to SAE is as follows: The driver can safely turn their attention away from the driving tasks, e.g. the driver can read a text or watch a movie. The vehicle will handle situations that call for an immediate response, like emergency braking but the driver must still be prepared to intervene within some limited time, specified by the manufacturer, when called upon by the vehicle to do so. The very interesting part of this definition is that the driver must be prepared to take over control within some limited time. It seems also that this time, not specified by now, can be defined by the manufacturer and not by an authority or legal body.
I had a discussion on this topic recently with other professionals in the industry but mainly with my fellow assistant. I was pretty sure that for level 3, 1 sec reaction is a given and for level 4, 10 seconds would be adequate. However my colleague was almost certain that 10 secs are given for level 3 and not level 4. I hence decided to conduct some primary research on this and I found out no exact value is defined at the moment. One reason is that there are not sufficient studies on the human-machine interaction of autonomous vehicles carried out to provide the evidence and information required. I found some interesting papers such as from “  Kelly Funkhouser and Frank Drews University of Utah, Proceedings of the Human Factors and Ergonomics Society 2016 Annual Meeting, Paper. The study presents research on reaction times with interesting figures. As the tests are only done with 24 people and only in a simulator, the results cannot be taken for the definition of the official time limit. Another study shows reaction times of up to 25 seconds  [Alexander Eriksson and Neville Stanton, University of Southampton]. Another research article [ Autonomous Vehicles: Disengagements, Accidents and Reaction Times, Vinayak V. Dixit , Sai Chand, Divya J. Nair, Published: December 20, 2016 on Plos.org Article] describes the results of test drives from Google, Mercedes and Tesla in California and also presents a discussion on reaction times. The reaction times where quite fast, but as there are no detailed conditions or parameters on the actual tests – especially for reactions times – I am quite sure the drivers always paid attention and did not drive “eyes-off”.
Coming back to the topic of taking back control by the human driver when the driver is completely “eyes-off” which could – in the future – entail a number of tasks such as reading a book or mails, attending a video conference, or watching a movie. Even if autonomous technology is well designed and tested, the interaction between the human and the machine is not yet designed well enough hence consequently causing the whole system to fail. Though mechanical failures constitute only parts of autonomous system failures, human factors continue to be the leading reason for failures .
In order to make the complete “system”, comprising of the autonomous system plus the human driver” safe, it has to be assured that the human driver is in any situation capable of what to do, when to do it and how to do it in order to get back full control of the vehicle. The reaction time itself depends on criteria such as, age of the driver, how often the driver uses an autonomous vehicle and of course the traffic situation and weather conditions. By saying all this it seems that nobody as a clear vision about the fact in how many seconds a human driver would be able to take back control safely.
There are two (2) main areas that would have to be considered to solve this problem:
1. Human-machine interface, warn and inform the human driver to take over and how the actual situation looks like (Simulation could be an option)
2. The algorithm which decides when to inform the human driver to take over the control
Just keep in mind that modern autonomous systems are designed to cover various traffic situations. In case the system gives-up it could be assumed that this is a very complex situation, or one of the system components fails. In case the system decides to give back control to the human driver, it must consider the reaction time, plus an additional time to really assure that the human driver is able to take over within the defined time limit.
In order to prepare the driver, the system would have to provide all necessary information of the traffic situation, system errors or failures, potential scenarios to react so as to equip him or her to be able to take over full control “properly”. This question does however brings to the question of whether drivers would need special training and.or licenses to operate in such conditions as mentioned above. This will have an impact on the design of the human interface!
Conclusively, it is not sufficient to design autonomous technology with the human interface, when vehicles will be developed with a steering wheel and pedals. In other words, the system is as safe as the human, as long as he/she is involved.