Computers qualify as drivers in automated vehicles
Wed 10 Feb 2016
In a letter posted on the National Highway Traffic Safety Administration website, the US government agency has informed Google that the AI computing system in their driverless vehicles may qualify as a ‘driver’, for the purposes of highway regulation and testing.
Last November, Google submitted a proposed design to the NHTSA for a fully automated vehicle in which an artificially intelligent self-driving system, or SDS, would control the vehicle and respond to the driving environment, so that the vehicle would have no need for a human driver.
Regulation and safety are of paramount concern as auto-makers continue to focus on the development and testing of driverless cars. A radical change in the way we drive would require a radical change in safety parameters and measures for driving; and as government changes are notoriously slow, and the development of automated vehicles is proceeding rapidly, the fear is that regulatory changes will hold back development, testing and rollout of the new technology.
In May 2013 the NHTSA released a preliminary policy on automated vehicle development in which they outlined five levels of vehicle automation. The first, level 0, is a no-automation vehicle, where a human driver is in full control at all times. Level 1 considers function-specific automation, which allows vehicles electronic stability control or pre-charged brakes to assist drivers in certain situations.
Level 2 is combined function automation, for example, cruise control with lane centering; level three is limited automation, where a human driver can turn over control of major safety functions to automated response in proscribed situations, but the driver is still available to take over should it become necessary.
The November 2015 proposal from Google included a design for a Level 4 vehicle with full self-driving automation. A Level 4 vehicle, per the NHTSA, would be “designed to perform all safety-critical driving functions and monitor roadway conditions for an entire trip. Such a design anticipates that the driver will provide destination or navigation input, but is not expected to be available for control at any time during the trip. This includes both occupied and unoccupied vehicles.”
The Google self-driving vehicle as proposed would actually remove steering wheels, gas and brake pedals, and other driver controls from the vehicle. The SDS would control all aspects of driving, and the option of human intervention in vehicle control would be removed completely. The idea is that the SDS would consistently make the best decisions for human occupants of the vehicle as well as pedestrians and other vehicles, and allowing human intervention in vehicle operations would actually be detrimental to the SDS functions.
This created an uncertainty in how the NHTSA would refer to a ‘driver’ in a vehicle in which all conventional ‘driver’ indicators were removed. The NHTSA made the decision to refer to the SDS itself as the ‘driver’ or ‘operator’ for specific Federal Motor Vehicle Safety Standards (FMVSS), including acceleration, braking, and signal operations. While certain regulations must still be officially rewritten before a large-scale deployment of Level 4 driverless vehicles is possible, allowing that an SDS could be considered a driver represents a great step forward in regulatory change required to make that happen.