An experimental analysis of the driver’s attention during train driving

1.1 Train protection systems

The train protection system is the technical equipment used on the railway tracks. It transmits signals to the traction unit to check the driver for compliance with the train’s limitations (speed). In the event of a breach (over speeding or non-response to the STOP signal), the train automatically stops.

Train protection functions are:

• informing the driver of the signal to which the train is approaching;

• control driver vigilance (vigilance button, dead man’s button).

The train protection system consists of trackside and mobile parts. The mobile part is placed on traction vehicles or driving cars. It can also be used to transmit and display signals to the driver’s station. It includes a vigilance button.

The train protection system is a railway safety system that communicates track status and condition information to the train driver of a locomotive, railcar or multiple units [3]. The data is continually updated, giving an easy to read the display to the train driver.

The most straightforward systems display the track-side signal, while the systems that are more sophisticated also display allowable speed, location of nearby trains, and dynamic information about the track ahead. Cab signals can also be part of a more comprehensive train protection system. They can automatically apply the brakes stopping the train if the operator does not respond appropriately to a dangerous condition.

The primary purpose of a signal system is to enforce a safe separation between trains and to stop or slow trains in advance of a restrictive situation. The cab signal system is an improvement over the wayside signal system, where visual signals beside or above the right-of-way govern the movement of trains. It provides the train operator with a continuous reminder of the last wayside signal or a constant indication of the state of the track ahead.

1.2 Train protection system types

All train protection systems (cab signalling systems) must have a continuous in-cab indication. It informs the driver of track condition ahead; however, these fall into two main categories. Depending on the method of transmission of information from track to train, trackside train protection equipment is divided into continuous and intermittent systems.

Intermittent train protection systems at discrete points along the rail line and between these points the display will reflect information from the last update. Continuous cab signals receive a constant flow of information about the state of the track ahead and can have the cab indication change at any time to reflect any updates. The majority of cab signalling systems, including those that use coded track circuits, are continuous.

Intermittent train protection systems provide constant reminders to drivers of track conditions ahead, but they are only updated at discrete points. That can lead to situations where the information displayed to the driver has become out of date. Intermittent cab signalling systems have functional overlap with many other train protection systems such as trip stops. The difference is that a driver or automatic operating system makes continuous reference to the last received update.

Intermittent train protection systems include:

• Stopper - the forerunner of all point safeguards. It is used, for example, by the Berlin S-Bahn or the Prague metro;

• Crocodile (France, Belgium, Luxembourg);

• INDUSI and similar-PZ80, PZB90 (Germany, Austria, Romania, Canada);

• Integra-Signum (Switzerland);

• ZUB 121 (Switzerland);

• SHP (Poland);

• AWS (UK, Ireland, India);

• TPWS (UK, Northern Ireland, Australia).

Continuous train protection systems have the added benefit of failsafe behaviour in the event a train stops receiving the ongoing incident relied upon by the cab signalling system. Early systems use the rails or loop conductors laid along the track to provide continuous communication between wayside signal systems and the train. These systems provided for the transmission of more information than was possible with contemporary intermittent systems and they enabled the ability to display a miniature signal to the driver “cab signalling”. Continuous systems are also more easily paired with Automatic Train Control technology, which can enforce speed restrictions based on information received through the signalling system. That is because the continuous cab signals can change at any time to be more or less restrictive, providing for more efficient operation than intermittent ATC systems. Continuous train protection systems include:

• LS (former Czechoslovakia, Czech Republic and Slovakia),

• MIREL (Slovakia and Czech Republic),

• EVM 120 (Hungary),

• RS4 Codici (Italy),

• ALSN (Russia),

• ATB (Netherlands).

1.3 Modern train protection

Modern train protection systems combine the advantages of both old systems - continuous information transmission (not always) along with accurate vehicle location. The most important characteristic is that the train control station receives overall information on the length and speed profile of the section ahead of the train for which the next run is permitted. If the speed limit is approaching or the end of the section is approaching, the device displays the maximum safe speed, which continuously decreases according to the brake curves. If this speed is exceeded, emergency braking is applied to prevent the train from passing beyond the leg for which it is allowed to run. These breakers eliminate the possibility of errors caused by improper operation or inattention [4, 5]. The data transmitted to the vehicle can also be used for automatic train control.

They consist of:

• Stationary (tracked) parts (central or distributed) autonomous devices mostly dependent on the state data of SZZ, ensuring the processing of necessary information on the given route and their subsequent transfer to a moving train with a particular transfer guarantee.

• The mobile part, which is installed on individual units (or vehicles) and which receives signals or data from stationary parts, decodes them and it provides information about the next section of the train path to the driver or vehicle computer. It also calculates braking curves and exceeds the maximum safe speed (signalled or calculated according to the braking curve) triggers emergency braking.

Modern train protection systems can be well sorted by country of origin:

• SCMT (Italy);

• LZB (mainly Germany and Spain);

• ETCS (Europe);

• KLUB (Russia and former USSR countries).

The newer systems use cab signalling, where the trains regularly receive information regarding their relative positions to other trains. The computer shows the driver how fast they may drive, instead of them relying on exterior signals. These types of systems are in everyday use in France, Germany and Japan. The high speeds of these trains made it impossible for the train driver to read exterior signals, and distances between distant and home signals are too short for the train to brake.

These systems are usually far more than automatic train protection systems; not only do they prevent accidents, but also actively support the train driver [6, 7, 8]. Some of these systems being nearly able to drive the train automatically.

1.4 Human factor as part of railway safety

Exist many safety systems used in railway transport, but the human factor, especially in train driving is still significant. As we can see in Table 1, the numbers of safety accidents (data from Czech Rail Safety Inspection Office) are significant, and the trend is not very positive [9].

The data in Table 1 included only accidents (passing the main/shunting stop signal), not all railway accidents. 99% of this type of accident is caused by carrier - train driver. It is because the driver’s inadequate attention mainly causes this type of accidents during train/ shunting operation. The main reason for not good driver’s attention is incorrect driving skills and experiences or overworking. In railway transport, do not exist any specific safety system, which can analyse driver attention, driving skills or good/ bad practise.

3.1 Aim and Methodology

The main objective of this article is to research driver’s attention during train operation on selected railway infrastructure. Based on this research, we will able to achieve the secondary objective – to understand the driver’s way of work and to contribute to increasing railway safety on the selected line [16]. To be able to solve the objectives of this article, we used the following techniques and tools.

The SMI (SensoMotoric Instruments) Eye Tracking Glasses were used to collect data regarding the drivers’ visual behaviour in real traffic conditions. This eye-tracking glasses with the lightweight are designed to record a person’s natural gaze behaviour in real-time in a broad range of applications, see Figure 3. Glasses have on the bottom of the frame and also in the front frame placed three high-speed cameras. Two of them record the movement and position of the pupils in the eye (in the infrared spectrum), and the third camera records the surroundings (in the typical visual spectrum). The mobile phone is connected to the glasses, which visualises the surroundings on the screen and shows the point of view by the cursor. These eye-tracking glasses provide native, binocular tracking at 60 Hz sampling rate over the whole trackable field of view. Combined with a high definition scene camera with resolution 1280×960p @24 fps eventually 960×720p @30 fps and HDR (high dynamic range) mode with high sensitivity for low light and automatic parallax compensation this ensures robust and accurate data with tracking range 80^∘ horizontal, 60^∘ vertical and gaze tracking accuracy 0.5^∘ over all distances.

Figure 3

SMI eye-tracking glasses and mobile device (Source: hmi-lab.uniza.sk)

The SMI BeGaze software for eye-tracking provides analysing and structuring information on experiments and displaying the eye-tracking data as graphs, all in one sophisticated application. This application offers different types of analysis and results from experiments. One of the most useful for our purposes is Scan Path, or also called Gaze plot. It is a map which shows gaze fixations of the driver and in which order they occurred, based on numbered circles. The next exciting tool included in BeGaze software is AOI Editor (Areas of Interest). The AOIs can be defined for video or still images stimuli. In the function “Move&Morph” the AOIs change their position and size during the sequence of single video frames.

All pieces of equipment (software and hardware) are parts of the Human-Machine Interaction Laboratory (HMI-LAB) that is located at University Science Park of University of Zilina. The HMI-LAB is oriented to provide the research and testing for the human-machine interaction in various conditions.

To fulfil the objectives mentioned above in the article, we had to set up the research procedure correctly – to use the appropriate research methodology [17, 18]. The methodology of the research is shown in the following Figure 4, and it is divided into 3 phases: preparatory, implementation and analytical phases.

Figure 4

Methodology of research

In the preparatory phase of the research, we had to identify the place where the experiment will be realised – the line of the railway infrastructure. We chose the railway track Žilina – Púchov. At this line, we localised two situations: drive between the two stations and drive through the station (with the stop).

For identification of the driver operations in the train cabin, we need to formulate the zones that are important for train control and operation. The name of these zones is the area of interests (AOI). During the experiment, we measured where (position), how long (duration), and how many times (frequency) the gaze of the driver is focused on those areas. This set of metrics we will call the key performance indicators (KPI) [19, 20, 21]. Mainly, we will detect for each AOI in the cabin these KPI (see Figure 5):

Figure 5

KPI analysis with BeGaze software

• dwell time – the total amount of time a participant fixates (fixations) or glances within an AOI (saccades). It is represented by absolute format – time in ms or relative format - % of total time,

• fixation count – the number of gaze fixation on the selected AOI,

• average fixation – the average fixation duration informs how long (in ms) the average fixation lasted for,

• revisits - the number of revisits provides information about how many times a participant returned his gaze to an AOI. The revisits allow examining which areas repeatedly attracted the participant (for better or worse), and which were seen.

These four essential KPI will play a significant role in the process of analysing train driver behaviour in real conditions.

4.1 Example 1: Drive between two stations Žilina – Horný Hričov

The total time of drive in this section was 255 seconds. During the drive, the train driver fixated his gaze on window section 57% of the total time, on panel section 14.5% of the total time and at the railway signal section 2.9% of the total time. The number of fixations at the sections correlates with dwell time metric. More fixations on a particular AOI may indicate that the AOIs are more significant (or the AOIs are changing) and more noticeable to the driver than the others. In summary, the number of fixations on the window section was 508, on the panel section was 181 and at the railway signal section only 12. It is logical because the scene on the frontal window is still changing.

The cognitive process of the driver is represented by the duration of average fixation at the section. We take a closer look at the particular section. The average fixation is varying from 186.5 ms (panel E) to 452.3 ms (railway signal section). It can be explained, that at the panel E are information that is easy to understand, and that information is displayed on the panel (bars, flashing buttons, ..). However, at the railway signal section are information that is moving in the space in front of the train and the drivers must identify the meanings of the signals.

Last KPI is the revisits. It represents how many times the gaze of train driver visited particular AOI. It shows us the real importance of the selected AOI. In this part of the train line, the most revisits have a front window part (68), and the least revisits have a panel K (0).

4.2 Example 2: Drive through the station Dolný Hričov

Another type of train driver operation is drive through the station (with stop). We analyse the drive through the station Dolný Hričov. Total time of drive in this section was 250 seconds. During the drive, the train driver concentrated his gaze on window section 71.6% of the total time, on panel section 26.7% of the total time and at the railway signal section 1.7% of the total time. These values of dwell time show that the driver was more focused on the operation as it was during the drive between the stations (security of passengers on the platform during stopping and acceleration process). When we compare the summary of fixations, on the window section was 502, on the panel section was 265 and at the railway signal section only 14. The increased number of fixations on the panel section show needs to use, control and operate the panels during this type of operation.

When we take a look at the average fixation during this operation, we can find that the lowest average fixation is 157.8 ms (panel L) and the highest average fixation is at the panel I 478.2 ms. The necessary information about the train line is located at the panel L. The control buttons that are important for controlling the train, when it is arriving and departing the station (automatic speed control buttons) are at the panel I. Notes at the panel L have only information function, then the average fixation was very short. However, controls at the Panel I have an essential function for the train control, so the average fixations were the longest (there we can identify the cognition process of the train driver during manipulations with them).

Revisits at this particular operations show the real importance of the particular AOI. In this part of the train line, the most revisits have a front window part (63), and the least revisits have a panel I (1) and K (1).

We analysed two different situations in real railway traffic. Now we can compare the obtained data from these two situations. Heatmaps can visualise the overall attention of the train driver. Heatmaps show the general distribution of gaze points. They are typically displayed as a colour gradient overlay on the presented image (view from the train cabin). The red, yellow, and green colours represent in descending order the number of gaze points that were directed towards parts of the image.

During the driving between two stations – example 1, the train driver was mainly focused on the front window (57% of total time and 508 fixations). Significant attention is devoted to panel E (9.6% of total time and 112 fixations). It is due to the central part of the driver’s work is checking the situation on the track, reading the signals, control the speed and train protection system. At Figure 6, we can find the heatmap of this driving situation. The primary attention of the train driver is focused on the front window and the panels E, D and F. These parts of the train cabin are most important for the train operations – for the driving between the stations.

Figure 6

Heatmap of the train driver attention at the section Žilina - Horný Hričov (example 1)

During the driving through the stations (with the stop) – example 2, the train driver was mainly focused on the front window (32.2% of total time and 254 fixations). The attention (dwell time) and number fixations at this part of the train cabin is only half of the attention and fixations in the previous situation. Significant attention is devoted to the right window and right mirror (summary 36.7% of total time and 248 fixations). It is due to the stop at the station and visual control of passengers at the platform during stop and departing from the station. Figure 7 represents the heatmap of this driving situation. The critical attention of the train driver is focused on the front window, the right window, the right mirror and the panels E and D. Identified parts of the train cabin are very important for securing the safety of arriving, stopping and departing the train from the station.

Figure 7

Heatmap of the train driver attention during driving through station Dolný Hričov