Can a Bus Driver Simulator Improve Driver Behaviour?

The Cranfield University Bus Driver Studies

In 2002, Cranfield University began a major research programme funded by Arriva Bus UK under the Knowledge Transfer Programme. The research aimed to identify the factors that increased bus crash risk and then put in place a raft of measures aimed at reducing that risk. The impetus for the work was the fact that bus drivers are a special group of professional drivers with increased risk of being involved in a crash at work for a number of reasons. Bus drivers have responsibility for passengers’ lives driving large, heavy vehicles that are constantly pulling in and out of traffic, mostly in built-up areas. They have a much higher annual mileage than private motorists and organisational factors such as bus schedules exert a strong influence on bus driver behaviour.

Added to this background had been the publication of data showing that whilst novice bus drivers are older and more experienced compared with novice car drivers, previous car driving experience seemed to have no significant impact on the chances of being involved in a bus crash offering little help in whether a bus driver might go on to be a safe driver (Wåhlberg, 2002). It appeared that driving a bus then represented quite a different set of problems to that of driving a private car, but little was known about what could be done to address bus driver safety.

The Cranfield studies began by investigating all bus crashes, whether culpable or not. Analyses showed that lack of experience of driving a bus was more influential than age of the driver in the first year of operational bus driving but after the first 2 or 3 years, age was found to be more influential.  Young and older drivers were more at risk of being involved in a crash compared with middle-aged drivers (Dorn and Wåhlberg, 2008a). We also found that about 50% of all bus crashes were contributed to by novice bus drivers in their first year of service and that most of these at-fault crashes take place at bus stops and junctions (Dorn, Garwood and Muncie, 2002).

Further, analyses showed that some drivers had a disproportionate number of bus crashes compared with others (Wåhlberg and Dorn, 2007). It was concluded that any training intervention should focus on novice drivers as the evidence suggested that there were deficient in hazard perception skills at specific roadway locations. Crash-involved drivers, on the other hand, seemed to demonstrate certain personal characteristics that increased their risk of crash involvement (Wåhlberg and Dorn, 2008b). A psychometric assessment of bus driver behaviour – the Bus Driver Risk Index™ – was developed for use in group discussions to address these personal characteristics Garwood and Dorn, 2003; Dorn and Garwood, 2005).

Using a Bus Simulator for Novice Bus Drivers

Given that the increased risk of novice driver crash involvement was the most pressing need for Arriva, it was decided that a simulator would be the most beneficial training intervention to address decision making at critical events. Hazard perception and anticipation have been singled out as especially important skills for road safety. Drivers with better hazard perception skills are described as having a more effective predictive mental model of the driving environment. Since novice bus drivers have had less contact with traffic and less time to develop and refine their mental models, they are less able to correctly predict the development of traffic situations than experienced bus drivers (Dorn and Stannard, 2006). A simulator would enable:-

●      Presentation of high risk scenarios impossible in the real system for reasons of danger

●      Whole task support including conditions of bus driver workload and time pressure

●      Standardisation of training intervention in a controlled environment

●      Analysis of task performance to enhance learning with repeated trials

●      Repeated performance of the same task until a benchmark is reached

●      Presentation of instructions dependent on previous responses by the driver

Despite these obvious advantages, there are few peer-reviewed published studies that support the benefit of using simulators for training purposes and organisations are often nervous about investing in a system without evidence of effectiveness.  Companies have many questions regarding what kind of simulator would best serve their purpose. Do trainees take driving simulators seriously? How realistic do they need to be for learning to take place? What is the evidence that learning acquired on a driving simulator is transferred to operational driving? The answers to these questions are hard to come by, but research at Cranfield University has gone some way to investigate important issues in simulator-based bus driver training.

The ABS was designed as a fixed base wide field of view simulator, adapted from a STISIM PC-based driving simulator. A bus cab sits in the middle of a 180° curved screen, 6m in diameter and 2.75m high from which the participant drives the simulator.  Simulator scenarios were developed from a bus accident analysis of the most high risk road features and events. The simulator training included multiple exposures to a high rate of hazardous and time critical encounters with traffic, pedestrians, signalised junctions, varying roadway configurations and traffic control devices (signs and markings). The ABS also captures up to 60 variables of performance data per 0.1 second. Performance measures included elements such as lane and speed deviations, speed limit and traffic signal violations, accidents, run completion time and median time to collision for all vehicle and pedestrian encounters.

The Arriva bus simulator has been validated and is capable of discriminating between novice and experienced bus drivers (Dorn and Stannard, 2006) showing that experienced drivers are slower on approach to a junction and avoid sudden braking compared with novice drivers.

Psychological Fidelity

Do trainees take driving simulators seriously?  Simulator researchers believe that if the user perceives a simulation as being more realistic, their behaviour is more likely to mimic that performed in the operational environment, but this is not necessarily the case. Perception of simulator realism is dependent on the level of experience in the real system. For the expert or experienced driver, even a high fidelity vehicle is not judged to perform like the real thing. For the novice driver though, their lack of experience with the real vehicle means they are less distracted by the artificial nature of a simulator. In our work, we find that whether or not a driver considers the simulation to be realistic does not seem to have any impact on the way they drive. Those drivers with a more negative attitude to the Arriva bus simulator did not significantly differ in their driving performance compared with those drivers with a more positive attitude (Muncieand Dorn, 2003). Previous research using flight simulators has revealed similar findings.

Physical Fidelity

Secondly, do driving simulators need to be high fidelity for training effectiveness? Physical fidelity is achieved if there is a high correlation between the physical features of a simulator and the road environment being replicated. With perfect fidelity, a training environment would be indistinguishable from the actual task environment. High-fidelity simulators are characterised by very sophisticated visual image generation systems, advanced vehicle dynamic models and complex motion bases.  Engineers and system designers assume that the greater the physical fidelity the greater the training effectiveness, but the importance of physical fidelity appears to depend on the nature of the task being trained. For instance, if decision-making skills are being evaluated, then (based on lessons from the aviation literature) high fidelity in vehicle handling characteristics may not be critical. Very realistic experiences can be achieved without fully replicating the real world – so long as the features of interest for training purposes are replicated. This is referred to as ‘functional fidelity’. So, providing vital elements of the tasks are simulated, high fidelity in all components of the simulator may not be necessary. Indeed, there is growing evidence from the aviation literature that intentional departures from reality by abstracting or augmented displays might actually enhance skill acquisition by focusing the trainees’ attention on specific task-related issues during training – especially for novices. This poses the problem that although simulators are currently categorised according to their level of fidelity, this does not help an organisation decide whether it is suitable for the kind of training they want to use the simulator for.

Transfer of Training Effectiveness

Many simulator users make the mistake of assuming that a driving simulator is a valid measure of driving behaviour based on validation studies on other driving simulators. Studies are not routinely conducted on a new system to assess validity, and they should be.  A simulator can be validated simply by evaluating the differences between several performance measures observed during real-life driving and during simulated driving. A practical way is to assess ‘relative validity’ considering whether driving performance data in both the simulator and the real vehicle (i.e. preferred speed on a particular type of road under similar kinds of conditions) is relatively similar.

The most important criteria of the effectiveness of the simulator for training though is whether transfer of training from the simulator to operational driving takes place. This is particularly important for training purposes and is often referred to as functional validity or behavioural validity. Major problems with establishing behavioural validity are often attributed to the realism of the visual display or motion and complex visual patterns, in other words, the fidelity of the simulator. However, behavioural validity does not always correspond to fidelity as, according to the flight simulator literature, departures from reality can actually improve training success. To our knowledge, there is only one publication that demonstrates a transfer of training effect of a driving simulator in terms of reducing accident rates. A group of 553 learner drivers in the US drove different configurations of a simulator (Allen, Park and Cook, 2008). The study set out to investigate whether simulator fidelity affected transfer of training effects by comparing different configurations on accident rates (see Figure 2). The findings showed that drivers taking part in the wider field of view condition had a significantly decreased accident rate post-licensure compared with those that were trained on a single monitor presentation. The implications are that full size wide aspect ratio images may provide an affordable answer to simulator-based training and can be provided by flat panel and projected displays in the future.

Conclusion

Currently, the academic literature offers little guidance about the optimum content and delivery of simulator-based driver training programs so companies are often guided by simulator manufacturers about what their training should be. Companies are also often restricted in the kinds of scenarios these manufacturers offer with little flexibility in making changes themselves. Companies should choose a system that is not only flexible for their current and future needs, but also allows data capture so that the benefits of using a simulator for their training purposes can be evaluated. Very often, systems with excellent visual systems sacrifice their computational power so that there is little left over to record performance data in any meaningful way.

In answer to the question, can simulators improve bus driver behaviour then – the Cranfield studies demonstrate that the ABS can and does predict those drivers that are likely to go on and be involved in a crash. Data is still being analysed to see whether simulators can improve crash risk over the longer term. What we can say is that companies investing in a system must ensure that validation studies are conducted prior to driver training. There also needs to be careful collection of data post-training to ensure the benefits outweigh the costs. Currently, manufacturers present their own data as evidence, but this has not been subject to peer-review.

Driving simulators offer distinct advantages, but until the research has been conducted to ascertain the context under which they are most effective, companies are simply taking a leap of faith.

Contact us for a demonstration of our behavioural risk assessment – Bus Driver Risk Index™.