Railway level crossings create serious potential conflict points for collisions between road vehicles and trains. Governments, the rail industry and others have been applying a variety of countermeasures for many years to improve railway level crossing safety which including upgrading passive crossings to conventional active systems. However, the cost to eliminate or upgrade passive crossings is very high. Estimates of installing conventional (track circuit) flashing lights or boom barriers (active protection) is not cost-effective. On-going maintenance costs would also likely to be considerable in view of the remote location of many passive crossings. Furthermore, vehicle collisions at crossings are frequently attributed to driver behaviour (i.e., non-compliance) in response to the warning device. These situations raise the question of whether upgrading passive to conventional active crossings would produce the desired improvement in safety, looking also at the cost incurred and system effectiveness gained. In view of that, researching cost-effective alternative systems is a worthwhile undertaking. Considerable research and innovation has occurred in some countries into the development of low-cost warning systems for level crossing. Immediate application in Australia is possible subject to its effectiveness and adaptation to the localised (Australian) conditions. The effectiveness of these alternative systems needs to be assessed to reflect safety improvements at crossings. However, to date, there has been no systematic approach available to evaluate these systems for implementation other than before-and-after implementation studies which are expensive.
This research proposes a methodology, through simulation, for evaluating alternative warning devices at level crossing to identify cost-effective devices with respect to the engineering requirements and particularly the driver responses prior to expensive prototype testing at field. The methodology involves multi-criteria analysis to short-list devices according to required engineering criteria; data collection of driver behaviour using driving simulator; development of driver behavioural models and application of such models associated with surrogate safety measures and traffic conflict technique in traffic micro-simulation modelling.
In this research, two potential alternative warning devices, rumble strips (passive) and in-vehicle auditory warning device (active), were studied. Driver behaviour in response to these two alternative devices were evaluated and contrasted with two conventional devices, stop sign (passive) and flashing-red-light (active) in driving simulator. In addition, field survey of driver behaviour at level crossings with conventional warning systems (stop sign, flashing light and boom barrier) at/near Brisbane in Australia was conducted for comparison. Driver behaviour investigated includes driver stopping compliance and driver reacting positions such as accelerator release distance, initial braking distance and final braking distance.
A binary logistic regression model was established for predicting the probability for a driver to stop or cross at a railway crossing, and mixed regression models for predicting drivers’ reacting positions before stopping at the crossing. In summary, these models describe driver behaviour with respect to the various types of alternative warning devices reasonably well. Contributing factors, such as age, gender, speed and types of warning devices were found significant at different approaching stages to the level crossings. As expected, speed affects drivers’ reaction in the early stage when approaching the level crossings and ‘types of warning devices’ has higher influence at the final stages of stopping. The influence of ‘age’ and ‘gender’ occurs only at the later stage of stopping prior to the stop line. Overall distinctions in behaviour within the two passive or two active systems were relatively small. The large differences occurred between the passive and active devices. Violation results indicated that the active systems produce much higher levels of driver compliance than passive devices. The results of the analysis revealed a potential weakness of the passive warning system in obtaining drivers’ compliance. Although similar driver behaviour was observed with alternative and conventional devices, the considerably lower costs of application of the alternative systems provide extra motivation for their use.
The behavioural models were then applied in a microscopic simulation to produce safety measures such as collision likelihood and time-to-collision for comparison among the potential devices. The modelling approach of using drivers’ stopping compliance and reacting positions enables relative safety evaluation of alternative traffic devices. This provides a more detailed evaluation of alternative devices, using simulated outcomes which have a more concrete relationship to road safety, including the numerical prediction of collision likelihood and severity. However, more research into other potential influences, such as familiarity with the crossing, and personality factors such as propensity to engage in risky behaviour, would certainly capture more of the variance in warning compliance. More complex road geometry, random approaching speed or additional scenarios may contribute to differences in observations. The driving simulator experimental design can
also be extended to testing the influence of other human factors such as distraction and fatigue, as well as obtaining other measurements such as perception-reaction time and deceleration rates. Other alternative devices (i.e., automated camera photo enforcement) may also be explored in future studies using driving simulator.