The evolution towards semi-autonomous and eventually driverless vehicles will progressively remove human error as the leading cause of vehicle accidents, significantly lowering vehicle accident frequency rates. Contracting personal motor insurance lines are anticipated as a result and liability is projected to shift towards Original Equipment Manufacturers (OEMs), tier 1 and 2 suppliers1 and software developers, generating potentially large loss exposures for these industries. This transition may force a departure from existing underwriting and actuarial methods to ensure risks are correctly identified and priced so that safe deployment of semi-autonomous and driverless vehicles is facilitated. Semi-autonomous and connected vehicles will generate enormous amounts of driving data. Such telematics and naturalistic driving data can be utilised to better understand emerging risk exposures and enable vehicle advancements and deployment through the provision of product liability and recall coverages. In order to develop insights into the risks involved, this paper will examine likely risk trajectories across the 6 levels of vehicle automation, from a risk management perspective. The application of driver performance and behaviour analysis to semi-autonomous vehicles is demonstrated using Bayesian Networks as a novel machine learning method for risk estimation. Hundreds of millions of autonomous miles are required to sufficiently estimate autonomous accident frequency and severity rates. With this in mind, we propose a method to analyse any deviations from a defined normal/safe driving performance, matched with contextual factors, in order to better assess technology limitations and capabilities. Deviations include any sudden or unexpected change in driving patterns whether that is swerving, braking, system disengagement etc. The resulting likelihood and severity of these deviations can be used to model expected claims loss for personal motor, product liability and product recall insurance coverages.
This work is supported by the VI-DAS (Vision Inspired Driver Assistance Systems) a European Commission, Horizon 2020 research consortium [grant number690772].
1 Tier 1 suppliers are direct suppliers to OEM’s whereas Tier 2 supply Tier 1.