Systems modelling and ethical decision algorithms for autonomous vehicle collisions
AuthorsPickering, James Edward
MetadataShow full item record
AbstractThere has been an increasing interest in autonomous vehicles (AVs) in recent years. Through the use of advanced safety systems (ASS), it is expected that driverless AVs will result in a reduced number of road traffic accidents (RTAs) and fatalities on the roads. However, until the technology matures, collisions involving AVs will inevitably take place. Herein lies the hub of the problem: if AVs are to be programmed to deal with a collision scenario, which set of ethically acceptable rules should be applied? The two main philosophical doctrines are the utilitarian and deontological approaches of Bentham and Kant, with the two competing societal actions being altruistic and selfish as defined by Hamilton. It is shown in simulation, that the utilitarian approach is likely to be the most favourable candidate to succeed as a serious contender for developments in the programming and decision making for control of AV technologies in the future. At the heart of the proposed approach is the development of an ethical decision-maker (EDM), with this forming part of a model-to-decision (M2D) approach. Lumped parameter models (LPMs) are developed that capture the key features of AV collisions into an immovable rigid wall (IRW) or another AV, i.e. peak deformation and peak acceleration. The peak acceleration of the AV is then related to the accelerations experienced by the occupant(s) on-board the AV, e.g. peak head acceleration. Such information allows the M2D approach to decide on the collision target depending on the selected algorithm, e.g. utilitarian or altruistic. Alongside the EDM is an active collision system (ACS) which is able to change the AV structural stiffness properties. The ACS is able to compensate for situations when AVs are predicted to experience potentially severe and fatal injury severity levels.
PublisherUniversity of Wolverhampton
TypeThesis or dissertation
DescriptionA thesis submitted in partial fulfilment of the requirements of the University of Wolverhampton for the degree of Doctor of Philosophy.
The following licence applies to the copyright and re-use of this item:
- Creative Commons
Except where otherwise noted, this item's license is described as Attribution-NonCommercial-NoDerivatives 4.0 International