Opportunities

Safety via Better Driving

The improvements to road safety represents one of the brightest hopes of self-driving technology.
The WHO estimated that in 2018, the number of road fatalities for that year had reached 1.35 million (World Health Organization, 2018). It has become the leading cause of death for persons aged 5-29, and the eighth leading cause of death for all age groups, outpacing historical killers such as disease and famine.

Research into road accidents tends to place human error as a cause more than 90% of the time. This may represent a bias against human judgement, with any failure not attributable to mechanical issues or an environmental hazard such as a landslide defaulting to ‘blame the driver’. Improvements to road safety following the ‘Vision Zero’ principles - placing the responsibility of safety on road designers rather than road users – have seen many countries slash their road toll. Sweden leads the world in road safety, having reduced its number of road deaths by 66% between X and Y (World Health Organization, 2018).
That Vision Zero has not reduced road fatalities to zero suggests that human error is undoubtedly still a significant contributor to road accidents. Automated driving systems already assist in many of the sub-tasks that are necessary to drive a vehicle safely. It will likely be preferable that, at some stage, the entire responsibility of the car’s safe passage is handed over to such a system. Even before that takes place, it’s likely we will begin to see the benefits of partially automated vehicles as they lift some of the burden from drivers.

Augmenting and eventually replacing capable and responsible but fallible drivers is a worthy cause. But where self-driving cars will likely have the most to give is as an alternative to irresponsible behaviour such as driving while fatigued, distracted, intoxicated or under the effects of drugs. Another potential use of automated driving systems is the ability to act as a co-pilot. This might be a particularly useful way to glean benefits from such systems in instances of cultures and individuals who resist giving up control to a machine.
An AI co-pilot may also be a vital training tool for inexperienced drivers and serve to resharpen the skills of trained drivers. In the case of the latter, particular care would have to be taken to ensure the system is useful without being a nuisance. Too aggressive in its instruction and it will be quickly disabled, even in cases where it might be useful.

Even the model human driver, cautious and attentive, cannot escape the decay in skill that comes with age, and like all skills, there are diminishing returns and increasing costs on further instruction to improve human driving ability.
Inversely, new advances in AI software and hardware will likely see self-driving cars continue to improve generationally, or even over the course of the life of the vehicle via software and hardware updates.

The first laser cost as much as a small house (Downs, 1999); 40 years later, the cheapest costs less than a dollar. Therefore, it is conceivable that history may repeat itself, where sensors that are prohibitively expensive for autonomous systems now may in future be cost-effective.

All but entirely replacing the human driver may also present secondary opportunities to improve road safety. Currently, $300 million NZD is spent on policing New Zealand’s roads. In a future in which NZ’s roads are dominated by AVs designed to obey the rules of the road almost to a fault, this could be cut back significantly. Savings such as this could then be reinvested into other measures to improve road safety.

Safety via Vehicular Design

The most straightforward improvement AVs offer over human-driven vehicles is the lack of blind spots and the ability to observe all their surroundings at once. This is particularly significant for cyclists and motorcyclists who are much more likely to be missed by a human driver and are naturally far more fragile than their four-wheeled counterparts.

As the need to design cars around a human driver drops away, vehicular design may change to better accommodate existing or new needs, such as comfort or safety. Much of a contemporary vehicle’s surface area is glass because this is a safety necessity for a human driver. With self-driving vehicles comes the need to externalise sensors, so once the human driver is disestablished from the vehicle’s design, glass becomes a safety risk rather than a safety necessity.

Whilst still “a dream”, as Barla (2023) puts it, vehicle-to-vehicle networking is very much an avenue that will be explored once self-driving vehicles are firmly established, and as networking technologies and algorithms improve in accuracy and efficiency.
One application of inter-vehicular networking is sensor fusion. Sensor fusion is already used by cutting-edge militaries like the United States to provide a more complete picture of the battlespace to combatants, allowing them to make more informed decisions.
Within civilian transport infrastructure, inter-vehicular networking could be used to enhance an automated vehicle’s decision-making. It could also be used to shore up an individual vehicle’s view of its surroundings if, for example, part of its sensor suite becomes incapable of providing the system with useful data in the event of damage, interference, jamming or blinding (Shalev-Shwartz et al., 2017).

Vehicle networking may pave the way for a more unified traffic management system. Such a system could influence the course of vehicles away from potentially hazardous areas, and its capabilities could even extend beyond directing traffic, such as assisting with safety and simplifying liability.
Emergency services with access to the system could access vehicular cameras remotely to assess accident sites on or near the road to make more informed decisions when dispatching first responders.

Safety via Predictability

Autonomous vehicles will still share the road with human-controlled vehicles well into the foreseeable future. One advantage that may benefit the non-autonomous vehicles is predictability. The ability to anticipate other vehicles is an important skill, and one that gets easier when the vehicle being observed can be trusted to behave in a particular manner.
With safety as a primary concern for consumers, it is reasonable to suspect, in the short term, that AVs will be overly accommodating, rather than aggressive, towards other road users.