It is often claimed that autonomous vehicles (AVs) hold the answer to improving vehicle safety. To support this belief, the commonly quoted figure is ‘AVs will eliminate 90% of accidents that are caused by human error’. But is this true? Where did this figure come from?
The figure seems to have arisen from the release of a 2015 NHTSA report which, on face-value, appears to support this perception. Closer interpretation of the report, however, shows that the situation is far more complex and that a more detailed analysis is needed before any realistic comparison between an automated system and a human driver can be made.
AVs (SAE Levels 3-4) are being developed primarily with convenience in mind, with their increased safety occurring as an inherent bonus, whereas ADAS (Levels 1-2) is mainly aimed at making cars safer. But, is attention and funding moving to AVs before ADAS has reached its full potential?
Looking to the UK as an example, there has been a downward trend in road fatalities ever since the first deployment of ABS in the 1970s, and this has continued over the last decade thanks to ADAS. One of the most effective current ADAS systems is Autonomous Emergency Braking (AEB). A 2015 study across six European countries, released by the UK insurance industry, identified that vehicles equipped with City-AEB systems were effective in preventing 38% of front-to-rear collisions, calling it ‘probably the most significant development in car safety since the seatbelt’. Recently, OEMs have benefited greatly from the accolades received for their ADAS. For example, Subaru’s EyeSight system resulted in all equipped models receiving the highest possible safety rating for front crash prevention in the US in 2019 testing. Furthermore, ITARDA (the Japanese Government’s road safety institute) has reported a 61% decrease in accidents resulting in injury or death for vehicles equipped with EyeSight.
Despite the proven benefit ADAS has on vehicle safety, the deployment of ADAS is still relatively low, leaving a lot of room for increasing fitment. Research shows that, in 2018, only around half (52%) of new cars sold in Europe were fitted with Collision Avoidance systems. In the developing world, fitment rates of all ADAS are much lower and a huge benefit to vehicle safety can be gained by a focus on increasing ADAS fitment in these markets.
Many ADAS systems are still relatively immature, and the technology deployed can severely limit their operating domain. For example, many AEB systems only provide full collision avoidance between 10 and 50 kph, and even then, it’s only for very specific scenarios. Increasing the range of operation to also accommodate motorway speeds would significantly reduce fatalities. Also, LDW (lane departure warning) systems are currently only capable of operating optimally on clearly marked carriageways. If we could extend the operating range so that they can recognise road edges for all road types, the system could be operational for close to 100% of drive time.
It is also worth highlighting that some ADAS are only active on demand (e.g. LDW), so their effectiveness depends on the driver actively engaging them. The industry therefore needs to also consider making ADAS on-by-default and (controversially) make it difficult or impossible to de-activate. This means that ADAS usability needs to significantly improve so that annoyance is kept to a minimum.
AVs are still a long way away, so why are we waiting?
NHTSA data from 2015 shows that the split between accidents with fatalities, vs those with injuries, is around 0.5% and 27% of all reported accidents respectively. The vast majority (over 72%) of accidents involve no injury, consisting of only property damage. Investment in ADAS will push accidents down this severity pyramid, resulting in significant reductions in fatalities and serious injuries.
This benefit is available predictably and immediately, unlike that promised from AVs. Simply put, to wait for the arrival of AVs to improve vehicle safety is not sensible as Level 4 or 5 AVs are still a very long way off. In fact, some voices are beginning to express doubts as to whether they will appear in the medium-term… if they ever get here at all.
– CEO of Waymo, January 2019.
This may appear to be unduly pessimistic, but in any case, it is clear that the number of AVs on our roads is unlikely to surpass manually driven vehicles until well beyond 2050 and there is still significant doubt whether they can reduce accidents by anything close to widely expected levels.
“Large amounts of money are being spent on the development of autonomous technology, the safety benefits of which are still poorly defined and a long way off” says Dr Alain Dunoyer, Head of Autonomous Car at SBD Automotive. “A balanced shift towards investing in both AVs and technology improvements for current ADAS will enable these systems to address a much wider range of accidents, and will have a larger impact on reducing road fatalities.”