Boer et al. (2015) outline the role of simulators in developing

automated driving systems. In particular, simulators can be used to

study how drivers of (fully) automated cars use their cars, and how

(multiple) manual drivers and pedestrians interact with automated

cars.

Google was said to be lobbying to have its Driverless Cars certified in a

virtual world instead of in the real world (Risberg, 2014).

Automation is taking over the physical work of manually turning the

wheel and pressing the pedals; more time will be left to interact with

computers in simulated environments. I foresee a blending of physical and

virtual environments.

Boer, E. R., Della Penna, M., Utz., H., Pedersen, L., Sierhuis, M.

(2015). The role of driving simulators in developing and evaluating

autonomous vehicles. Driving Simulation Conference Europe, 3–10.

Risberg, E. (2014). Google lobbies to test self-driving cars in

Matrix-style virtual world. Retrieved from

http://www.theguardian.com/technology/2014/aug/21/google-test-self-driving-cars-virtual-world-matrix

Boer et al. (2015) outline the role of simulators in developing

automated driving systems. In particular, simulators can be used to

study how drivers of (fully) automated cars use their cars, and how

(multiple) manual drivers and pedestrians interact with automated

cars.

Google was said to be lobbying to have its Driverless Cars certified in a

virtual world instead of in the real world (Risberg, 2014).

Automation is taking over the physical work of manually turning the

wheel and pressing the pedals; more time will be left to interact with

computers in simulated environments. I foresee a blending of physical and

virtual environments.

Boer, E. R., Della Penna, M., Utz., H., Pedersen, L., Sierhuis, M.

(2015). The role of driving simulators in developing and evaluating

autonomous vehicles. Driving Simulation Conference Europe, 3–10.

Risberg, E. (2014). Google lobbies to test self-driving cars in

Matrix-style virtual world. Retrieved from

http://www.theguardian.com/technology/2014/aug/21/google-test-self-driving-cars-virtual-world-matrix

In my opinion, driving simulators will always be used for system prototpying as long as there are vehicles on our roads. Safety assessment is not everything our research is about - it is more about the development of new and innovative concepts (that also have to be safe, of course), if you ask me. Just think of the autonomous vehicle as a work space - how would one have to design such a vehicle interieur? And just consider, highly (not fully) automated driving would always involve a take-over concept, which would have to be designed carfully as well.... In the domain of concept prototyping, one will be always faster and cheaper with a driving simulator, so no worries here. 

Any operator in the loop simulator is useful for a variety of purposes but almost all of them come down to being able to answer questions or provide training with respect to the human's interaction with a system.  I would say that as long as human factors issues continue with driving then driving simulators will continue to be useful.  Safety issues are one important aspect.  A second one is training.  For rail vehicles that already operate quite often in driverless modes, operators must be regularly trained on simulators in case they need to go and retrieve a vehicle where the driverless system has failed. 

I think the on-road testing will reveal evidence of problem scenarios for automated vehicles. Certainly, the crash-causation picture will look drastically different with AVs, but there will be one. Then simulators can be used to zoom in on those problem scenarios to study them in detail. I don't think it's how us simulator folks envisioned the process to go, but I think we still have an important role to play. We need to keep educating govt and OEMS as to the benefits of each kind of testing.

I agree with Dr. Richard Romano, that "as long as human factors issues continue with driving then driving simulators will continue to be useful". The on-road testing will reveal problems of new parts or systems of vehicle, or even automated vehicles. I think on-road testing is more likely as the final or last validation method for vehicles, and we can still use the simulator to validate any problem during vehicle system prototyping, especially for developing highly automated vehicle such as ADAS.

First of all, we will enter (or are already entered) in a long transition phase where we will have specific problems: automated PLUS human-driven cars. Then we will still have nichés where human-driven cars are needed. Lastly, for the prototyping phase, e will definitely need devices such as simulators.

Lastly, spoken as an experimental scientist: just to grab data over a very long period of time confronts you with a giant blast of data--often very discrete and clear hypotheses which are then tested quite experimentally in a simulator can be superior to such an effect, because you can control so many factors and you can develop very neat designs.

Cheers!

CCC

Minor caveat here: governments are not pushing automated driving. Usually, they are pretty risk-averse and a lot of testing and data analysis will need to be done before this experimental idea gets a regulatory go-ahead for the consumer market. I agree with CCC that the most likely next stage is the automation of certain driving tasks while keeping the human driver in the loop. 

Simulation remains important because it allows for the testing of "extreme" conditions - the goal being to see when the human will take over (automation is not infallible if conditions are outside the parameters of the system the user/passenger need to be aware and able to take action). Just from a psychological standpoint I anticipate a great deal of push-back as driving is not just a means of transportation, but a sought after and enjoyed activity - there is apt to be some significant trust issues with automated driving that we don't see with other forms of transport. Potentially we can capture these stress points in a simulation.

I agree with all above writing people that simulators will remain important for scientific purposes. We are at the beginning of higher automation in cars.

Aviation is about 30 years ahead of road traffic with respect to automation. Does aviation have many research flight simulators today? Yes! Does aviation have even far more simulators and different simulator types for training today? Yes!

Maybe driving simulators will also be used for training in future.

- There is basic/initial training. Training of manual driving skills, traffic rules (in real situations compared to learning with books), behavior of pedestrians, etc. Parts of this training could be done by the use of simulators.

- There will be training for the use of automation; to operate a kind of autopilot and set different modes. Monitoring is also an important issue: understanding of system boundarys and take over performance.    

- And what about an additional license for highly automated road vehicles? For busses and trucks only? Also for highly automated cars?

- What about additional training for non-automated drivers to understand how other highly automated car behaves in certain situations?

- Recurrent / Refresher training?

I guess some of these issues will be addressed and discussed in future and simulators will remain important tools for different purposes.   

The comparison of road 'traffic' with aviation is only meaningful for cases in which cars only interact with other cars (or the road as such). In all answers above mixed traffic, especially when mixed with bicycles, is ignored. To my mind this is no less than ridiculous. In the Netherlands, and not only there, cars-only roads are a minority, not just in towns ans cities but also in rural and not-so-rural areas. It is not all that helpful to deny the basic complexity involved when discussing automatically driven cars.

Actually, from my point of view the importance of simulator experiments for road traffic safety will increase. The reason is the amount of assisting and distracting devices plus the tendency towards more or even full automation. Keeping the driver 'in the loop' has to be investigated thoroughly in a diversity of situations and a diversity of assisting devices, especially their interaction.

Actually, desasters like the AF 447 show that even professionals are not always able to take manual control after the autopilot has kept them 'out of the loop'. For the normal driver this will be even harder. For this reason simulator experiments might help us to find an optimal balance between technical assistance or partial automation and the drivers' activity.