Effects of Automated Systems on Safety (EASY)

Lead Research Organisation: University of Leeds
Department Name: Institute for Transport Studies


This work is intended to examine how some of the new Advanced Driver Assistance Systems, that are envisaged by the car manufacturers, will affect safety. Currently, the most advanced assistance system on the market is Adaptive Cruise Control (ACC) which automates the task of car following. ACC is particularly designed for motorways, but it can also be used on rural and even urban roads. It has deliberate limitations, in that it cannot deal with situations requiring severe braking and in that the ACC forward radar cannot detect stationary objects. The car manufacturers plan to extend the capability of ACC so that it can handle most forward situations. They also plan to provide lane keeping systems which will automate the lateral control of a vehicle (i.e. steering), once again particularly for motorway driving. The combination of longitudinal and lateral control will produce a situation in which a large part of the driving task is automated. As a consequence, there is a risk that drivers will no longer feel a need to pay attention to the road and traffic environment, and therefore may not be aware of impending risk. They may also lose track of when manual control has been resumed, e.g. on exiting from a motorway, and therefore be slower in responding when required to brake or steer.This project will conduct a systematic evaluation of drivers' performance and safety awareness as they experience increasingly greater automation of the driving task. The major tool for this work will be the new driving simulator at the University of Leeds, which will have a complex motion base to provide gravitational feel to the drivers. The forces experienced by the driver in accelerating, braking and steering will be as realistic as possible. The simulator will also use the most up-to-date techniques for image animation to create a fully immersive environment, in which the sensation of driving is as realistic as possible.The initial set of experiments will be designed to identify any safety-related problems that result from driving in a semi-automated vehicle. A wide range of drivers will used, with the major factors in their selection being age, gender and trust in automation. Having identified the problems, a second set of experiments will focus on solutions to those problems, i.e. on ways in which driver alertness and awareness can be enhanced. The results are intended to provide guidance to those governmental organisations that are planning to use new driver assistance systems to increase road capacity and safety. They are also intended to lead to better design of new products by the vehicle manufacturers.


10 25 50
Description EASY conducted two major experiments to examine the impact of semi-automated and fully automated driving on driver attention allocation, situation awareness and ability to react in demanding situations. In the first experiment, drivers were given the freedom to choose how much attention to pay to the driving task and how much attention to pay to various entertainment and information tasks. In the second experiment, behaviour in normal traffic and in more demanding incidents was compared. The final experiment in the project looked at mitigation strategies for bringing the driver back into the loop. A simple time-based procedure was compared to a procedure based on a real-time attention monitor.

In experiment one, 48 participants drove on a motorway under three conditions - manual driving, semi-automated driving (either automated lateral control or automated longitudinal control) and fully automated. Automation was a within-subject condition and 24 participants drove with each of the alternative semi-automated systems. Participants were offered a number of entertainment and information activities as well as grooming and food. They were free to choose as many activities as they wished and were free to engage in a chosen activity at will. Overall, the DVD task was by far the most preferred. Task engagement fit the expected pattern with the highest rate being in the fully automated condition. However, there were substantial differences between the two forms of semi-automated driving. Driving with longitudinal support was overall much closer in its effects to manual driving while driving with lateral support was more similar to fully automated driving. This is a new and potentially important finding, since it indicates that previous theory on levels of automation is insufficient to predict behaviour and driving task performance. Instead, we need a proper understanding of which aspects of the driving task are being taken over by the automated system.

In experiment two, the effect of level of automation on driver behaviour in normal traffic conditions and during incidents was examined. Driving performance in manual, semi-automated and highly automated conditions was compared, both where there was no need for drivers to intervene with the system and in approaching an incident. The effect of an 'artificial' non-visual secondary task on performance in each driving condition was also observed. Results showed that performance with the secondary task deteriorated driving performance, due to the increase in resources required by this task. Safr driving was better maintained by the system than by the drivers themselves. Drivers maintained shorter time headways when they were in control of the vehicle themselves, than when headway was controlled by the longitudinal controller. Drivers' speed was also found to be higher than that imposed by the automated system. Finally, drivers' management of the secondary task and driving performance was found to be worst in the semi-automated conditions, where it was perhaps not clear to drivers whether they or system were in charge.

In experiment three, two mitigation strategies were developed and tested. Drivers were required to take control of the vehicle based on fixed time periods or on dynamic intervals. The approach of using fixed time intervals was based on a review of the literature on adaptive automation. The dynamic intervals were based on drivers' attention allocation - when the driver was detected as being below a threshold of attention to the forward view, the system would then trigger a switch from automated to manual control of the vehicle. The dynamic intervals had a more pronounced effect on keeping drivers 'in-the-loop'. However, both mitigation strategies were equally popular among drivers.
Exploitation Route This research should influence pathways to high automation in driving. There are strong indications that drivers will expect highly automated driving to deliver benefits in the form of freedom to engage in non-driving-related activities. Whe automation is high, maintaining attention to the roadway and traffic is difficult.
Sectors Aerospace, Defence and Marine,Digital/Communication/Information Technologies (including Software),Transport

Description They have been extensively discussed in conferences and other fora with the automotive industry. They have a potential impact on the design of automated vehicles, in particular on expectations of whether or not "drivers" will be attentive to the external road scene under various levels of automation.
First Year Of Impact 2011
Sector Transport
Impact Types Societal,Economic

Description FP7
Amount € 779,700 (EUR)
Funding ID 610428 
Organisation European Commission 
Sector Public
Country European Union (EU)
Start 01/2014 
End 06/2017