Autonomous Vehicles: Issues Affecting Future Development
Of primary importance in the future of AV is the development of technological systems which allow for reliable operation in atypical conditions (e.g. traffic, difficult weather, high speed). At present, diverse sensor technologies are being explored to counteract issues arising in previous tests using LIDAR, radar, and cameras.
In the 2017 paper, ‘Managing the Transition to Electrical and Autonomous Vehicles’, Todorovic proposed necessary future stages for AV on the basis of Anderson’s 1990 model of technological change. This portrayed innovation as a discontinuity followed by a cyclic process, culminating in one dominant design. While this process is reflective of many well-known case studies and may be true for AV, a distinction should be drawn between the individual technologies which make up such vehicles, and the composite technology which is the autonomous vehicle as a whole. Certainly, each individual technology is likely to converge towards a single dominant design, however a core component of successful machine learning (upon which many AV processes will depend) is collection of a broad, multi-faceted data set.
The challenge seems likely to be in creating dynamic and responsive systems which successfully fuse inputs from a wide sensors array (shown on the TRM), to allow collection of more data than strictly necessary (and different data by different companies). The scale of this challenge suggests that there, at least, convergence is unlikely.
If true, failure to achieve a dominant design, and resulting divergence in proprietary systems between companies leads to the next key problem – the requirement for AV cooperation. Traditionally tech companies such as Google are likely to have different aims to traditional automotive companies, yet despite these differences integration will be key. The weaknesses of the Internet of Things approach to automation is already laid bare in the lack of integration between competing eco-systems in the field of home automation, and the disjointed network of proprietary refuelling stations for electric vehicles in the UK and US.
In the case of connected autonomous vehicles, which must communicate with and ‘listen’ to their surroundings to ensure safety, this issue is even more pressing. As a result, and in light of the ever-increasing complexity of technological ecosystems (which are, overall, a recent development) there is an opportunity and a need for further innovation management research into methods to successfully balance competition and cooperation. AV technology will require a reconciliation of processes of creation for private enterprise, developing a competitive framework which self-regulates and standardises as common practice, minimising the need for external regulation yet ensuring safety and consumer confidence to reach critical mass.
Though there are theoretical frameworks which allow for the formation of both collaborative and competitive relationships between the same two companies as noted in ‘How firms navigate cooperation and competition in nascent eosystems’ (Hannah and Eisenhardt., 2018), empirical evidence on the development and aggregation of these relationships shows a cumulative effect in which companies will, in general, make an early choice to operate as collaborators or as competitors which is then ‘fixed’ (Sytch, 2013). This is unlikely to allow efficient adoption of AV technology, reducing value for all key stakeholders. As such, the importance of the ‘universal AV communication’ target outlined on the TRM cannot be overstated.
Finally, the most commonly recognised challenge facing AVs, relates to questions of ethical responsibility, societally accepted standards, and legal responsibility in the event of system failures and accidents. There is a clear risk, noted in Marchant (2012), that questions of liability acts as a deterrent to the development of the technology among manufacturers. This is particularly worrisome given that, for all the discussion of the safety of self-driving vehicles, the limited available data from past road-tests demonstrates the opposite.
In the US at present road traffic deaths sit at a little over 1 per 100 million miles driven (IIHS, 2018 data accessed 17 January 2021). Yet since 2015 AVs developed by Uber and Tesla have been involved in 5 deaths, failures of the technology directly contributing or causing the accident in some of these. Given that the developer with the most road-miles, Google’s Waymo, have driven only 20 million miles as of January 2020 these 5 crashes likely represent a considerably higher ‘deaths per 100 million miles’ figure for AVs.
Interestingly, however, much research into public perception of autonomous vehicles (limited though it is by a lack of first-hand consumer experience with level 5 automation) identifies factors such as ownership aspirations and ease of use as more pressing influences in their perceived usefulness (Kum Fai, 2020). As a result, market research into factors affecting social acceptability and legislative backing were noted on the TRM as key stages in gaining consumer confidence, but a requirement for safety messaging was not. It may be that such discussions are better confined to the legislative sphere and, as far as possible, overlooked in messaging to a public which has already spent some considerable time sitting with the ethical questions relating to AVs and remains, generally, positive about the concept.
