Keywords

1 Introduction

A recent major development in air traffic control is the change from the conventional tower to remote tower services. The conventional tower workplace is characterized through a direct view on the airport operational surfaces, runway and the airspace surrounding. In remote tower operations air traffic control services for a single airport are performed from a physically remote located control center workplace (single-mode). Since there is no direct view of the airport surface, real-time video cameras are positioned at the airfield, which provide an image on screens at the remote tower workplace in the tower [1].

The interest to serve multiple airports from a limited number of workplaces in the remote tower center is increasing. Here, the upcoming step is providing multi remote airport service, meaning, that more than one airport is controlled from one air traffic controller (ATCO) at one workplace at the same time (multi-mode). While technological feasibility is already given, the impact on mental models of the ATCO is not yet fully understood. The aim of this study was to investigate the effects of updating of and shifting among mental models in multi remote tower. The chosen approach comprises a Human-In-The-Loop simulation that shall allow a comparison of human performance measures in single and multi-mode.

2 Theoretical Background

2.1 Multi Remote Tower as a Safety Challenge?

Multi remote tower is considered as a fundamental change for ATCOs, since it influences how, which and when tasks are done. Presumably, visual scanning and task patterns require modifications [2]. From a cognitive perspective, the impact on ATCO workload in multi-mode must be considered. A relevant aspect that influences workload is the increased traffic volume in multi-mode [3]. In a simulation study by Moehlenbrink et al. this influence was investigated and a significant increase in workload in multi-mode in comparison to single-mode was shown [4]. A further aspect is the constant updating of the respective mental models for all controlled airports by the ATCO [2, 3]. Mental models contain mental representations of airspace, aircraft as well as air traffic control (ATC) procedures and technical systems. These models support the development of situational awareness [5]. Since multi-mode is characterized by a higher degree of dynamic changes, it is considered to be more complex than single-mode, which seems to be associated with reduced situational awareness [6]. Updating of mental models is accompanied by the monitoring and coding of external information and requires replacing of outdated information, which is affecting workload [7]. In order to update all mental models, a frequent cognitive shifting among the models is also necessary, affecting workload as well [8]. For shifting, attention is practically relocated from on to the other airport. Updating and shifting are executive functions that are considered as causing cognitive costs [9]. This influence of shifting among and updating of mental models on workload and situational awareness in multi-mode has not yet been investigated. Increased workload and reduced situational awareness pose a potential challenge for multi-mode, since safety could be decreased.

2.2 Multi Remote Tower as a Safety Chance?

Multi remote tower is of special interest for airports that feature low traffic volumes. In the conventional tower, a low traffic volume can cause working conditions that promote low workload (underload). Underload can lead to monotony and decreased vigilance, which could impair performance, alertness and safety in air traffic control [10]. Moreover, Straussberger and Schaefer [11] showed a link between monotony and increased sleepiness in ATCOs. In the multi remote tower, a safety benefit is expected by the lower exposure of the ATCO to monotony which is, combined with a demand for high alertness, a known stressor [12].

To compare, the radar workplace in air traffic control allows changes of the controlled area (sector) size, based on traffic load and time of day. The idea behind this procedure is, to avoid overload within peak hours (decreased sector size) and underload during nighttime (increased sector size). The multi remote tower enables a similar flexibility of balancing workload, which could be a key to adapt to human demands. Hence, adjusting the workload by increasing or decreasing the number of airports and aircrafts provides the scope to respond to monotonous working conditions.

For multi-mode, it could be interesting if shifting among and updating of several mental models is workload sensitive. Besides the higher amount of traffic in multi-mode, the increase in workload could support ATCO alertness. Increased alertness shows a potential chance for the multi-mode, since this could improve safety and working conditions.

2.3 Research Aim and Hypotheses

Research activities undertaken so far show an ambiguous picture on the expected effects of multi-mode to ATCO performance and related safety. The central aim of this study is to gain a better understanding of mental models and executive functions in multi remote tower.

Firstly, we hypothesize, that the ATCO develops a separate mental model for each airport in multi-mode. As the monitoring of multiple airports requires the shifting among and updating of the various mental models, we secondly hypothesize an increased ATCO workload in multi-mode. Thirdly, we hypothesize, that this increased workload causes reduced situational awareness, which can be regarded as ‘shift costs’. Finally, we hypothesize that higher workload increases alertness, as there is less exposure of the ATCO to monotony.

3 Methods

Our approach bases on a Human-In-The-Loop simulation study that was chosen for gaining empirical evidence by observations and subjective questionnaires. A pivotal element of the simulation is to keep the traffic volume at an equal load across the single and multi-scenarios. This is to exclude any secondary variance induced by an added traffic load on workload and situational awareness.

3.1 Sample Characteristics

In the study, n = 8 licensed Swedish ATCOs (1 female) with a mean age of 48,8 years (SD = 8,5) and a mean work experience of 24,2 years (SD = 8,1) participated.

3.2 Apparatus

We conducted the study in the ATC research lab at the Area Control Centre in Malmö. The platform based on the NARSIM simulator that featured a high-fidelity simulation of a remote tower workplace. Video presentation of 14 simulated cameras was shown on six 55’’ wide-screen screens. A smaller, pan-tilt-zoom camera was shown in a picture in picture view for each airport. The pan-tilt-zoom camera could be operated by means of a control stick and buttons for pre-defined zoom levels and camera directions (Fig. 1).

Fig. 1.
figure 1

Workplace during multi-mode, including left-right-split for outside view and screens.

Full size image

3.3 Test Design and Scenarios

Within the study, participants controlled air traffic across two scenarios in four simulation runs; each run had a duration of 90 min. All runs followed a crossover-design for counter balancing confounding factors and to provide a design for within-subject comparison. The scenario set consisted of two runs in single- and two in multi-mode as independent variables. All scenarios shared the same amount of flights, weather figures and ground movements. Simulated flights compromised scheduled Instrument Flights (IFR) and non-scheduled Visual Flights (VFR). VFR-Flights performed touch-and-go and circling exercises. In multi-mode, traffic movements were split fairly over both airports, while in single-mode the traffic was handled at one airport. Calls of the adjoining control sector and from the meteorological department were simulated as well. Two pseudo pilots and one simulation operator ensured a realistic radio communication and aircraft behaviour during every run.

Prior to the first simulation run, a 90 min training run was conducted to ensure sufficient system knowledge and confidence. Before and after each experimental run, participants completed questionnaires.

3.4 Measured Variables

Situational Awareness was measured by means of the Situation Assessment Rating Technique (SART) [13], applied after each run. Workload was measured during and after every run. During simulation, the Instantaneous Self-Assessment (ISA) [14] was administered. It had to be completed every three minutes (indicated by a tone) and represents a workload self-rating. After the run, participants completed the NASA Task Load Index (NASA-TLX) [15]. Alertness was measured using the Karolinska Sleepiness Scale (KSS) [16] before and after each run.

4 Results

Pre-analysis of the data characteristics revealed a lack of normality (significant Kolmogorov-Smirnov-Test) and variance equality (significant Levene-Test). Mann-Whitney U-Tests were applied to analyse the data obtained from SART, ISA, NASA-TLX Raw and KSS.

4.1 Situational Awareness

Analysis of SART data revealed no differences between single- and multi-mode (global score, U = 93.0, Z = −1.07, p = .283; U = 103.5, Z = −.658, p = .510; supply, U = 73.0, Z = −1.88, p = .060; understanding, U = 111.0, Z = −.362, p = .718).

4.2 Workload

During each run, workload was measured every 3 min through ISA queries. Figure 2 shows the mean values with respect to both modes. The analysis revealed significant (p < .05) higher ratings in multi-mode in comparison to single-mode for ISAMinute 48 (U = 72.0, Z = −2.14, p = .033) and ISAMinute 51 (U = 75.0, Z = −2.18, p = .029).

Fig. 2.
figure 2

Mean results of ISA values. Error bars represent standard deviation. *p < .05

Full size image

Analysis of NASA-TLX Raw data revealed no significant difference between single- and multi-mode (U = 95.5, Z = −.969, p = .338).

4.3 Alertness

The analysis of KSS data indicated no differences between both groups in pre- and post-measurements (pre, U = 103.5, Z = −.987, p = .324; post, U = 126.0, Z = −.059, p = .953). In order to analyse for any changes over the time of the scenarios, a Wilcoxon-Signal Rank test was conducted. Analysis within every group revealed for single- and multi-mode no differences between pre and post measurements (single, Z = −1.35, p = .177; multi, Z = −.302, p = .763).

5 Discussion

The present study investigated the effects of shifting among and updating of mental models in single and multi-remote tower. Simulations with experienced ATCOs were conducted. Measures for situational awareness, workload and alertness were applied.

The results revealed no differences for workload and situational awareness between single and multi-mode. Shifting among and updating of two mental models in multi-mode did not lead to reduced situational awareness and increased workload. While this result is contrary to our hypothesis, this means nevertheless no additional risk for multi-mode. ATCOs were able to control a high traffic volume across both modes with good performance. The initial assumption, that multiple mental models are developed, needs to be rejected. Instead of developing two independent mental models for two airports, the ATCO is perhaps developing a common, hybrid mental model for both airports at a time, including a clear distinction between both airports. Since ATCOs are trained to work in high traffic situations, the participants had probably sufficient remaining cognitive capacity to handle traffic on both airports parallel and update a single mental model accordingly. Furthermore, as multi-mode featured the same traffic amount across two airports, less of control advices are necessary to keep traffic safely separated. This concerns especially the runway and airspace whose capacity was doubled in multi-mode on the expense of time separation between simultaneous landings. The added separation may lead to a reduced demand for situational awareness in multi-mode. A probable aspect to consider is behavioural adaption of the participants that may feel uncertain in multi-mode due to its novelty beyond the familiar working environment. This may hide any effects, which perhaps become evident on the long-run. Through a possible subjective feeling of a higher situational instability in multi-mode, participants may have checked more the surrounding area and concentrated more.

The hypothesis, that multi-mode increased workload which in turn leads to higher alertness cannot be confirmed. The shift costs per se was concludingly not a pivotal factor for workload. However, traffic amount and characteristics are dominant factors as demonstrated by the workload diagram during the snowstorm.

5.1 Conclusion: Safety Challenge or Safety Chance?

Based on the data gathered, shifting among und updating of mental models in multi remote tower should not be considered as a safety challenge. Shifting among mental models is, if existing, not as workload demanding, and hence situational awareness reducing and alertness increasing, as expected.

The indicated, workload is rather associated with traffic characteristics and volume. Multi remote tower will feature increased traffic volumes, that will have an effect on workload subsequently. While overload could lead to reduced situational awareness, slightly increased workload could act as a countermeasure for monotony, supporting alertness in turn [17]. Additionally, we did not measure long-term effects of working in both modes. This is from a great interest, since the hypothetical optimum of ATCO workload is dynamic and changes over time due to workload and shift work factors [18]. Moreover, a long-lasting strain due to constant shifting is associated with cognitive fatigue [19]. Taking together, a well-balanced mixture between under- and overload seems to be a key for safety and for operational implementation of multi remote tower. Here, a follow-up study to compare realistic traffic volumes in single- and multi-mode should be striven for.

Finally, we conclude, that multi remote tower still accommodates ‘challenges’ and further research should be performed to answer remaining questions. The ‘chance’ to overcome monotony and promote alertness by multi-mode is evident, however, it requires a careful workload management, that takes margins for individual variations and overload peaks due to emergencies into account.