The development of autonomous vehicles by automotive and technology companies, is an exciting and active area of research receiving much media attention. Fascination extends from the independence of these vehicles, gained by measuring their local environments using a number of sensing and imaging technologies. The widely adopted SAE International’s Levels of Driving Automation define a six-level scale describing vehicle independence, with higher levels indicating more sophisticated automation. Contemporary research has recently developed vehicles rated at level-3 in which vehicles can operate autonomously but do not support all driving modes, such as high speed highway driving or low speed traffic jams and, in an emergency, require driver intervention. To attain higher levels of driving automation vehicles must be able to reason about their environment to be able to support all driving modes and to react in emergencies. To achieve this, research must extend to fully-connected vehicles, in which vehicles become more aware of their environment, at greater distances, by communicating with other vehicles and transport infrastructure.
The speed and latency communication requirements of fully-connected vehicles preclude their use of 4G and even 5G telephony networks. Instead, vehicles will communicate using localised and temporarily-formed vehicle ad-hoc networks (VANETs). Without a centralised authority, VANETs are susceptible to malicious attacks from vehicles and even roadside infrastructure, which can transmit falsified information about vehicles’ motion and road conditions. Such attacks have the potential to disable individual vehicles, dramatically increase congestion, and even cause potentially fatal accidents. Past research has attempted to address this problem using roadside units (RSUs), devices which supply services such as temporary authentication or systems to prevent identified malicious actors from communicating. However, these units will incur a large cost to set up, will have a transition period before they are operational and, particularly in a large country such as Australia with remote areas, cannot offer full coverage.
To better support the detection of falsified or simply errant information, we present a decentralised system which establishes localised reputation- and trust-based networks between fully-connected vehicles. Information received by each vehicle is verified by calculations that match data from physical systems on the vehicle and the degree of consensus about the same information also received by neighbouring vehicles.
HEDGES, J. and McDONALD, C. (2017). Establishing Trust Between Fully-Connected Autonomous Vehicles. Paper presented at CAITR 2017 (35th Conference of Australian Institutes of Transport Research), Rottnest Island, WA, Australia.