The US Department of Transport has published the first set of rules for self-driving cars and other highly automated vehicles (HAV).
The Vehicle Performance Guidance for Automated Vehicles is intended to set rules for how autonomous vehicles are constructed and operated, and will form the basis of consultation with manufacturers, transport bodies, law enforcement and the public.
The guidelines include requirements for vehicles to collect data in the event of a crash, to build in cybersecurity as part of the vehicle design and to be able to operate to ‘ethically acceptable’ parameters.
Anthony R. Foxx, Secretary of Transportation, said that the DoT believes there is a lot of potential for HAVs, namely in improving safety, but also in providing mobility to groups of people who cannot drive themselves or do not want to own a car. There were 35,000 road deaths in the US in 2015, and 94% of crashes involved some human error or decision, and the DoT said that HAVs could address the “overwhelming majority” of accidents.
Foxx wrote in the introduction to the guidelines: “Technology in transportation is not new. In fact, the airplane, the automobile, the train and the horse-drawn carriage all introduced new opportunities and new complications to the safe movement of people and goods. As the digital era increasingly reaches deeper into transportation, our task at the US Department of Transportation is not only to keep pace, but to ensure public safety while establishing a strong foundation such that the rules of the road can be known, understood, and responded to by industry and the public.
“The self-driving car raises more possibilities and more questions than perhaps any other transportation innovation,” he added.
All automated vehicles will be expected to pass a safety assessment, which includes parameters for data recording and sharing; privacy; system safety; vehicle cybersecurity; human machine interface; crashworthiness; consumer education and training; registration and certification; post-crash behaviour; federal, state and local laws; ethical considerations; operational design domain; object and event detection and response; fall back (minimal risk condition) and validation methods.
Vehicles will be expected to collect and retain data, especially in the event of a crash, like a black box flight recorded so that accident data can be analysed and used to improve safety and to understand specific incidents.
HAV data will be governed by strict privacy policies to regulation, data collection, transparency, and so on; while cybersecurity should be a part of a robust product design processes, to minimise safety risks. NIST and other cybersecurity best practices should be considered in this process.
Human Machine Interfaces (HMI) should be designed to consider how the driver interacts with the car, even when handing over control to the vehicle, and how the vehicle interacts with external factors or road users.
Manufacturers should ensure that car dealers, distributors and end customers are properly trained in how to use the technology.
Policies will also have to be set to govern software and hardware updates, and to ensure compliance with all road laws.
HAVs will also need to be developed to operate to “broadly acceptable” ethical guidelines, which would govern behaviour in the event of circumstances that would create conflict between safety, legality and mobility ie. when can a car pass a parked car that is blocking a single lane road by moving into the opposing lane; or if a crash is unavoidable, what course of action minimises harm to all passengers, pedestrians etc?
Manufacturers will also have to set the specifications for operational conditions for the vehicle, such as which types of road it can drive on in which weather conditions; parameters for Object and Event Detection and Response; capabilities to avoid hazardous ‘pre-crash’ situations and procedures for the car to safely shut down or hand control to a human driver if automated systems are not working properly.
