Sie sind auf Seite 1von 18

Cognition, Technology & Work (2002) 4:211228 Ownership and Copyright # 2002 Springer-Verlag London Limited

Cognition Technology & Work

Eects of Airborne Data Link Communication on Demands, Workload and Situation Awareness
T. Muller and H.-G. Giesa
Department of Human-Machine Systems, Institute of Psychology and Ergonomics, Technical University of Berlin, Berlin, Germany
Abstract: An airborne air-to-ground data link communication interface was evaluated in a multi-sector-planning scenario using an Airbus A 340 full ight simulator. In a close-to-reality experimental setting, eight professional crews performed a ight mission in a mixed voice/data link environment. Experimental factors were the medium (voice vs. data link), workload (low vs. high) and the role in the cockpit (pilot ying vs. pilot non-ying). Data link communication and the usability of the newly developed communication interface were rated positively by the pilots, but there is a clear preference for using a data link only in the phase of cruise. Cognitive demands were determined for selected sections of en-route ight. Demands are affected mainly by increased communication needs. In the pilots view, although a data link has no effect on safety or the possibilities of intervention, it causes more problems. The subjective workload, as measured with the NASA Task Load Index, increased moderately under data link conditions. A data link has no general effect on pilots situation awareness although ight plan negotiations with a data link cause a distraction of attention from monitoring tasks. The use of a data link has an impact on air-to-ground as well as intra-crew communication. Under data link conditions the pilot non-ying plays a more active role in the cockpit. Before introducing data link communication, several aspects of crew resource management have to be reconsidered. Keywords: Air-to-ground data link; Aviation; Cognitive demands; Communication; Crew resource management; Flight simulator; Multi-sector-planning; Situation awareness; Workload

1. INTRODUCTION
1.1. Background A considerable increase in European and worldwide civil air trafc is expected in the next 20 years. Since the air trafc management (ATM) system is currently operating at its limits, new models and concepts are already under development. Any future increase in capacity should also mean an increase or at least the maintenance of todays safety standards and economic improvements (Eurocontrol 1998, 1999). The bottleneck in ATM is at least partially due to air trafc control. Given specic safety standards and particular technical equipment, only a limited number of aircraft can be controlled in one control sector at any given time. In the past, this problem was solved by reducing the size of sectors. However, this generally limits the planning perspective of the air trafc controller and/or makes more co-ordination necessary with neighbouring sectors.

1.2. Multi-Sector Planning and Data Link A more promising approach introduces a new planning authority called a multi-sector-planner (MSP) who operates on a longer time perspective and aims at preventing rather than solving conicts. Multi-sector planning is a key feature of an ATM concept developed at the Technical University of Berlin (Fricke et al 2000). The MSP is responsible for several radar sectors (multi-sector area). Aircraft ight plan data is used on the ground for a computer simulation of planned trafc trajectories to predict conicts in the multi-sector-area. Based on these predictions, conict resolutions are calculated as modied ight plans and negotited via a data link with aircraft in the multi-sector area. Air trafc controllers (ATC) can then supervise a greater number of aircraft when conict density (conicts per aircraft) is reduced. The MSP requires permanently updated information about ight plans, current position of aircraft and shortterm interventions by ATC. An efcient data link is needed between the computers of the MSP and ATC on the ground as well as to and from airborne ight manage-

212

T. Muller and H.-G. Giesa

ment systems (FMS). The VHF voice communication now commonly used between pilots and ATC will be replaced by digital data link communication. For a complex exchange of information (except ight plan negotiations) or in an emergency, voice communication will serve as a backup system. In addition to the potential for trafc optimisation with the MSP concept, further benets over conventional voice communications are expected from the use of a data link for ATC communications. Rehmann (1997) mentions a reduction in miscommunication, a reduction in frequency congestion, the potential for direct entry of data into an aircrafts autoight systems and the permanent documentation of clearance messages. Several changes occur for airborne crews: negotiating ight plan data is an additional task, the MSP is a further partner in communication, and the data link interface is a new communication device. The ergonomic design of the data link interface and the dependability of the entire humanmachine system require special attention.

2. STATE OF THE ART


2.1. System Design and Acceptance Numerous functional, procedural and design aspects of the implementation of data link systems have been investigated in empirical and analytical studies (SAE 1994; Rehmann 1997). Particularly careful studies have been done on the location of data link interfaces within the cockpit: the display should be located within the forward eld of view (Knox and Scanlon 1991; Kerns 1994; Rehmann et al 1995; Rehmann and Mogford 1996). For the input of information, the Multi Purpose Control and Display Unit (MCDU) is considered to be well suited (Rehmann et al 1995). Regardless of the ight phase in question, pilots should always be in radio contact with an ATC facility (Rehmann 1997). Van Gent (1995) has found that the rated acceptance of a data link is highly dependent on the displays page layouts and operating procedures. Pilots acceptance of air-to-ground data link communication also strongly depends on the ight phase; the highest acceptability was found for ground operations and cruise. 2.2. Data Link Communication as a Part of the CrewAircraft System The FAA Human Factors Team (1996) states that in modern ight deck systems solutions that involve the entire system are required, not just focused solutions to individual problems. The effect of the use of the new communication system on the behaviour of the entire system has rarely been investigated explicitly in previous studies. In the case of a data link a risk of interference between simultaneous

communications and ying actions is anticipated (Navarro and Sikorski 1999). Since the overall efciency of operations is supposed to increase with the use of a data link and at any rate not decrease (Kerns 1994), it is necessary to evaluate the impact of a data link on the whole situation. Problems experienced by the ight crew and a degraded perception of safety are indicators of bad system integration. Since a lot of problems have occurred as a consequence of too much automation, it is suggested that tasks should be designed in such a way that the operators are better involved in the processes and enjoy a higher degree of freedom while operating the system. For ight guidance, such design characteristics have been discussed for two decades as a measure to address the ironies of automation and to enhance safety (e.g. Wiener and Curry 1980; Bainbridge 1983; Amalberti 1993). Operators working in the system are often confronted with situations which cannot be foreseen by system developers (unknown unknown situation; Sheridan 1983). It can be concluded as in the concept of human-centred automation (Billings 1991, 1997; ICAO 1994) that new systems should be designed in a way that actively involves pilots and allows them the possibility of intervention (e.g. Boy and Cacciabue 1997). In highly automated systems such as modern civil aircraft, traditional performance measures, e.g. ight path accuracy, are hardly attributable to pilots. The focus has to be directed on the pilots potential to maintain overall system performance on a tolerable (i.e. safe) level when confronted by system failures or the unexpected behaviour of technical systems. This potential is discussed in terms of keeping the pilot in the loop (e.g. White 1991; Endsley and Kiris 1995), situation awareness (e.g. Orasanu 1995; Endsley 1995, 1996) or attention allocation (e.g. Sarter 1995). 2.3. Impact of Data Link Communication on Pilots Workload Studies do not agree on the effect of data link communication on pilots workload. Kerns (1994) reports that there is little evidence that the use of a data link impacts on the overall workload, but she emphasizes that there is a shift from an auditory and speech workload to a visual and manual workload. Van Gent (1995) and Rehmann and Mogford (1996) also found that pilots workload is not affected by a data link. In a comparative study of three data link interfaces (with voice as control condition), a signicant impact of the communication medium on workload was identied for the pilot non-ying only, although both pilots tended to perceive workload as minimal under voice communication conditions (Rehmann et al 1995). Lee et al (1997) investigated the usage

Effects of Airborne Data Link Communication

213

of the ight management system in a way similar to typical data link applications; workload was signicantly increased compared to the other experimental conditions. The authors point out that this is true for the approach phase only and report opposite ndings for en-route conditions. Any change to the work environment will have consequences for the pilots overall work. In aviation the demands are analysed in many cases by investigating their effect on the pilots using the concept of workload. The ndings of workload modications caused by data link are typically based on pilots reports of perceived workload. Up to now there has been a lack of detailed analysis of the actual demands resulting from the use of a data link. In accordance with the general shift from sensory motor to cognitive tasks in modern cockpits (Billings 1997) an appropriate analysis should focus on cognitive demands. 2.4. Attention and Situation Awareness The effects of data link communication on attention and situation awareness are discussed in several terms: distraction of attention from other tasks through the use of the data link interface, reduction in attention demands because it is no longer necessary to monitor the VHF communication of the surrounding trafc (party line), and reduction in situation awareness due to the loss of party line information and the reduced transparency of communication to the pilot ying. In contrast to listening to the voice radio, the handling of incoming and outgoing messages via a data link requires looking at the display and/or keyboard. This distracts the pilots attention from observing the airspace as well as from monitoring other displays. Even if the principles of crew resource management dictate that the pilot non-ying handles communication, incoming data link messages increase the head down time of the pilot ying, too (Rehmann et al 1995; Jorna 1997; Lee et al 1997). In todays air-to-ground communication, all messages within a control sector are transmitted on a single VHF frequency (party line), i.e. pilots have to listen continuously to the call signs which identify the addressed aircraft. On the other hand, data link communication is point to point, i.e. aircrews receive exclusively those messages which are intended for them. Therefore, data link leads to a reduction in attention requirements regarding the necessity to continuously listen to the party line and identify call signs. An analysis of incident reports (Rehmann 1995) as well as simulator studies (Ineld et al 1995) indicate that the confusion of similar sounding call signs is a severe problem in todays air trafc. Finally, with voice transmission problems occur with incorrect readbacks (short-term memory overload) after longer ATC messages and ambiguities due to procedural deviations, e.g. omission of

call signs (Morrow et al 1990). The avoidance of such problems requires increased attention. A data link typically gives support by presenting the message until it is acknowledged, an additional message history functionality, and generally more standardized communication procedures (Navarro and Sikorski 1999). On the other hand, the monitoring of the party line has some advantages. The communication between ATC and crews of other aircraft in the same geographical sector gives information about the trafc situation, weather conditions etc. It is assumed that this information contributes to pilots situation awareness. In the near future where only relatively few aircraft will be equipped with data link capabilities and reduced party line information is still available, no major problems with regard to pilots situation awareness are expected (Ineld et al 1995). In an exclusive data link scenario, the lack of information is assumed to be more dramatic in a terminal environment and close-to-airport manoeuvres than within the en-route environment (Rehmann 1995). Voice communication with ATC, even if performed by the pilot non-ying exclusively, is immediately accessible to both pilots. Keyboard data entries and messages on displays outside the primary vision eld are less transparent to the pilot ying. Therefore, the two pilots mental representations of the ight situation may differ to a larger extent than under voice conditions (Navarro and Sikorski 1999). As a consequence, the pilot non-ying often reads incoming messages aloud to the pilot ying (e.g. Logsdon et al 1995) or synthetic speech (Rehmann and Mogford, 1996) is used in data link scenarios. 2.5. Eects on Communication Behaviour Many studies dealing with an air-to-ground data link focus on the transaction time as a central theme. Some of them report an increase in duration (e.g. Kerns 1994; Logsdon et al 1995; Jorna 1997), while others nd no signicant differences (McGann et al 1998). These ambiguous results are due to different locations of the data link communication interface in the cockpit, different ways to consider the technical transmission times (which can be of considerable duration in mode S radar transmission), additional tasks and nally the ergonomic design of the interface. Jorna (1997) comes to the conclusion that a data link in itself is not critically slower as compared to voice, provided that the communication interface does not require extensive manipulation to create an answer. Finally, increased transaction times can be an indicator that pilots perform tasks with higher priority before answering to ATC (Navarro and Sikorski, 1999). Even if a data link primarily inuences air-to-ground communication, effects on the communication within the cockpit are observable as well. Hrebec et al (1995) report

214

T. Muller and H.-G. Giesa

an increase in the number of communication activities and changes in the structure of intra-crew communication under data link conditions. The rst ofcers (who were responsible for air-to-ground communication in this experiment) became more active in communication, i.e. they performed more initiative speech acts than they did under voice conditions. Crew communication patterns were found to be good predictors of crew performance (Kanki et al 1989, 1991). Supplementary, intra-crew communication seems to be a coping mechanism that helps to lessen and manage both the causes and effects of stress (Sexton and Helmreich 1999).

3. OBJECTIVE OF THE STUDY


For a better understanding of the effects of a data link on the working situation within the cockpit, an experimental investigation was performed in which an experimental prototype of a data link communication interface was compared to conventional VHF voice communication. The aim was to evaluate the use of a data link within a multisector planning scenario. It was necessary to re-examine the acceptability of the data link interface employed which was tested in several usability studies during the design process under this aspect. The experimental investigations were guided by the expectation that data link communication with ground authorities (ATC, MSP) has an inuence on the cognitive demands and as a consequence on pilots perception of the system state and workload. According to the ndings of earlier studies (see above), we expected effects on pilots situation awareness and communication behaviour. In contrast to typical usability studies, the performance of a ight mission under close-to-reality conditions was the object of the study, not the solving of isolated tasks.

4. METHODOLOGY
4.1. Equipment 4.1.1. Flight Simulator For reasons of safety, research strategy and economy, all experimental investigations were carried out in a ight simulator. For the experiments described in detail later, the Airbus A340 training and research full ight simulator of the ZFB (Zentrum fur Flugsimulation) Berlin was available. This ight simulator is equipped with full vision and motion features as well as with a Scientic Research Facility that allows features to be adjusted in the Flight Management System (FMS) and in several display and control devices.

4.1.2. Data Link Communication Interface An experimental prototype of a data link communication interface was developed and implemented in the ight simulator. It uses already existing displays: the Navigation Display (ND) and the Multi-purpose Control and Display Unit (MCDU). The ND is located in the pilots primary vision eld. The MCDU is situated on the centre aisle pedestal; therefore the use of the MCDU requires looking downward. Figure 1 (left) shows the location of the displays in the cockpit. The ND was redesigned to provide space for two text lines and an icon related to the message, e.g. symbolising a climb or descent. In the example in Fig. 1 (upper right) the icon refers to the MCDU, where more details are presented. The MCDU, which is equipped with an alphanumerical keyboard, is used for pilots entries and for the presentation of more detailed messages like ight plans. Several new screen pages for data link communication with ATC and for ight plan negotiations were developed (Muller et al 1999, 2001). Some of the new MCDU pages were based on existing pages, e.g. the page for ight plan negotiation was derived from the secondary ight plan index page. Other data link pages (for ATIS, Pilots Requests, Message History) were designed from scratch. The system design process was a cooperation between ight engineers, human factors engineers and professional pilots. Tasks, operational structures and technical systems were designed simultaneously with a rapid prototyping technique. During this design process, professional pilots were subjected to certain stages of development in usability studies. Two series of usability studies were carried out: one concentrated on ight plan negotiations, the other on short-term (tactical) ATC messages. In the rst series, the new functionality and procedures for ight plan negotiations were considered by 13 pilots in three phases corresponding to different stages of development. In the second series, data link functionality for shortterm communication between pilots and the controller were investigated in two phases with nine pilots (Muller et al 1999). With the integration of multi-sector planning, it is possible to change the actual ight plan during ight for medium-term optimisations. A new feature allows the pilots to initiate ight plan negotiations, e.g. due to weather conditions. In Fig. 2, the sequence of MCDU pages for ight plan negotiations between pilots and multi-sector planner is given as an example of a data link communication procedure. The sequence starts with the Flightplan Negotiation Page (1). A copy of the active ight plan is loaded (COPY ACTV F-PLN) and displayed on the MCDU screen (2). Now it can be modied. The consequences of each modication on estimated time of arrival, distance to destination and estimated fuel on board

Effects of Airborne Data Link Communication

215

Fig. 1. Location of Navigation Display (ND) and Multi-purpose Control and Display Unit (MCDU) in the cockpit of the Airbus A 340 (captains side; another set of displays is located on rst ofcers side).

are displayed as delta values. The modied section of the trajectory is highlighted to indicate the differences between the active and the modied ight plan. After having nished the modications, the pilots proposal for the new ight plan is sent to the multi-sector planner (3). Now it is indicated that the negotiation is in progress (4). The multisector planners decision regarding the requested ight plan is displayed (5). In the example the ight plan is accepted and can be activated (6). As soon as a ight plan proposal is displayed on the MCDU a graphical representation of the ight plan is also available on the Navigation Display. For the purpose of a factorial experimental design, a simplied procedure for ight plan negotiations under voice conditions was also developed. The core is an oral transmission of a list of waypoints. Such a procedure might also be of practical value in the case of a temporal breakdown of data link transmission facilities and should be available as a backup procedure. Figure 3 gives an overview of the handling of ATC short-term messages. It starts with the incoming messages TURN RIGHT HEADING 230. (Note: Same message is displayed on the Navigation Display). Pilots have the choice to accept (WILCO) or to reject (UNABLE). In the example, the pilots accept by pressing WILCO. The

transmission to ground requires activation of the SEND key afterwards. 4.2. Experimental Design 4.2.1. Participants The participants in the experiments were 16 professional pilots (eight crews) from various airlines with Airbus type rating. Half of them were captains and half were rst ofcers. The mean professional experience was 13.5 years as a commercial pilot. Table 1 gives further information. 4.2.2. Procedure The participating pilots received a short manual attached to the invitation to the experiment in which the new data link communication interface and the modied communication procedures were described. Before starting with the experiment, the pilots rst had to handle typical communication tasks (e.g. ATC requests, ight plan negotiations) on a PC-based training version of the data link communication interface to ensure that each participant had basic experience in using the system. After this training session and the preparation of the physiological measurement equipment, the pilots received a brieng which included an introduction to the underlying air trafc

216

T. Muller and H.-G. Giesa

Fig. 2. Example of the sequence of MCDU pages for ight plan negotiations initiated by the pilots.

Fig. 3. Example of the reception of a short term ATC message and compliance by aircrew.

Effects of Airborne Data Link Communication


Table 1. Test subjects: age and professional experience (mean and standard deviation) All pilots (n = 16) Age/years Experience as pilot/years Total ight time/hours 38.7 (SD = 7.8) 13.5 (SD = 9.1) 7706 (SD = 5140) Captains (n = 8) 42.5 19.9 11,288 (SD = 7.9) (SD = 8.9) (SD = 5115) First ofcers (n = 8) 34.7 7.1 4125 (SD = 5.9) (SD = 2.4) (SD = 1067)

217

management system with the MSP, the distribution of the data link areas and the ight mission. Required documents like ight plans, information about technical parameters and maps were handed over to the pilots. Although the aircraft was already positioned at the runway when the pilots entered the ight simulator, the simulator session started with the manual input of the ight plan data into the ight management system and preparations for the ight (which normally takes place at the gate) to bring the pilots in the loop. The properties of the simulated ight and the methods used will be described later. 4.2.3. Flight Scenarios To evaluate the ATM conception under controlled as well as close-to-reality conditions, ight mission scenarios were worked out. The scenarios were detailed scripts covering messages from ATC and MSP, certain events which occurred during ight at scheduled time or location and general ight conditions (e.g. daytime, weather). The pilots mission was a ight from Copenhagen to Frankfurt and vice versa. During this ight, the aircraft passed through four radar sectors. In different versions of the scenario, a data link was provided in the two northern

sectors only or in the southern sectors respectively. Under both conditions (voice and data link) the initial call after entering a new sector was performed via voice. During the round trip of a duration of approximately 2 hours, eight test sections of special interest (approximately 3 minutes each) were dened where scheduled events (e.g. technical failres, weather fronts) produce different degrees of workload (Fig. 4). The pilots were not informed about the beginning and end of the test sections before or during the ight. Standard ATC messages are typically allocations of ight levels or headings and identication procedures when entering a new radar sector. There were two to four of these standard messages per test section. In four sections, ight plan negotiations took place. Flight plan negotiations are not a standard procedure for pilots yet. Under voice conditions, a simple procedure was used which started with an ATC message like There is a ight plan change report when ready to copy. Under data link conditions, the respective messages were displayed on the ND/MCDU with the newly designed data link pages beginning with ATC constraint list received. In three sections, ight plan negotiations where initiated by ground authorities. Since pilots were briefed to re-route by means of ight plan negotiations if

Fig. 4. Denition of test sections.

218

T. Muller and H.-G. Giesa

the weather conditions worsened, pilot-initiated negotiations (with the pilots own ight plan proposal) took place in sector D. Technical faults were in one case the occurrence of engine vibrations which required activity by the pilots, including shutting down the affected engine for a while. In another case, a failure in the heating of a standby pitot probe tube was displayed. The fault had to be recognised, but did not require any further action. In two sections, faulty ATC commands were sent, e.g. maintain on ight level 220 when the aircraft was actually on ight level 250. 4.2.4. Experimental Factors Two levels of workload were implemented with the denition of the test sections. Since a data link is provided in the northern or southern sectors only, two levels of the communication medium were available. Furthermore, pilots were briefed to take their roles as pilot ying or pilot non-ying according to a schedule and to change roles only one time at the stopover. Finally, a balanced factorial design of the type 2 6 2 6 2 with the (within-subect) factors communication medium (voice communication vs. data link), workload (low vs. high) and role (pilot ying vs. pilot non-ying) was obtained (Fig. 5).

Fig. 5. Experimental factors (VC = voice communication; DL = data link; WL = workload).

4.3. Methods of Measurement 4.3.1. Questioning After nishing the simulator ight, the pilots were confronted with video recordings from the cockpit taken during the test sections. To support the recognition of the situation, each section began and ended with an ATC message. After watching the video sequence of a ight section, pilots answered a questionnaire. The questionnaire for each section was split into two parts: a German translation of the NASA Task Load Index (TLX; Hart and Staveland 1988) and a questionnaire for situation judgment. The NASA TLX contains six items to measure subjects perceived workload on scales ranging from 0 to 100. These items are Mental demand, Physical demand, Temporal demand, Performance, Effort and Frustration. In addition, an overall workload score is computed from a weighted combination of the six scores. The weights were determined from a paired comparison of relevance by the participants according to Hart and Staveland (1988). The

NASA TLX has been used in numerous studies in the aviation eld and has turned out to be a valid and reliable instrument for workload measurement (Gawron 2000). High correlations between the total workload measured with the NASA TLX and task difculty have been found (Hancock et al 1995). There are recommendations about interpreting the scores on the individual scales of the NASA TLX as well as about the use of the weighed summary score (Hart and Staveland 1988; Byers et al 1989; Moroney et al 1992). Moderate delays of 15 minutes between task performance and lling out the questionnaire are reported to have no impact on the results (Moroney et al 1995). In a second questionnaire, pilots rated the situation in different dimensions. For the analysis reported here, ratings on the following scales were used: Safety, Sufcient possibilities of intervention, Problems perceived and three scales on the knowledge of ight parameters (primary ight parameters, engine parameters and navigation parameters). All situational dimensions were rated with a ve-point scale. The reported knowledge of ight parameters is perceived to give information about pilots situation awareness. The technique of measurement is similar to the China Lake Situation Awareness Rating Scale (described in Gawron 2000). Since subjective measurements of situation awareness are of limited reliability, e.g. because they can reect only such deciencies in situation awareness which are conscious to the rater, they are supplemented by related performance aspects according to the recommendations of Jones (2000). After the assessment of the test sections, pilots rated the system usability with a usability scale adapted from Brooke (1996). Finally pilots attitude towards data link communication was measured with another seven-item scale. For both ratings, a ve-point scale of agreement (like above) was used. All questionnaires used were in German. 4.3.2. Video Analysis The same video recordings as mentioned above were analysed under three aspects: cognitive demands, communication activities and attention-related performance indicators. The assessment of cognitive demands employed a method closely related to a part of the Cognitive Reliability and Error Analysis Method (CREAM) introduced by Hollnagel (1998). The analysis was oriented on the Extended Method for performance prediction, which starts with a task analysis following a given list of 15 cognitive activities. From these cognitive activities, four cognitive functions or cognitive demands have been derived according to Hollnagel (1998): observation, interpretation, planning and execution. The data used in the analysis reects the ideal course of actions. It is based on

Effects of Airborne Data Link Communication

219
Table 2. Mean (M) and standard deviations (SD) of agreement with statements regarding data link communication (n = 16 pilots); scales ranging from 0 (I totally disagree) to 4 (I totally agree); some scales inverted, so higher values indicate positive attitude towards a data link Item 1. 2. 3. 4. 5. 6. Data link contributes to ight safety Cutting down the voice contact with ground is a disadvantage [score inverted] Data link facilitates the work in the cockpit With data link the work in the cockpit becomes less interesting [score inverted] Data link is less error tolerant than VHF communication [score inverted] I dont really care whether the communication between controller and pilot is done via data link or not Data link will completely replace VHF communication in the future M (SD) 2.62 (.89) 2.06 (.68) 2.06 (.77) 3.06 (1.06) 3.00 (.63) 1.81 (.83)

the analysis of the test sections of the ights of two out of the eight crews and is corrected with perfect performance as a control. A communication analysis was carried out because a large portion of all cognitive activities are related to communication. Units of analysis are so-called communication acts. A communication act is the ow of communication until there is a change in speaker/sender, a change in topic or a break of 5 seconds or more. The classication of communication acts into initiative and response communication acts follows a scheme introduced by Kanki et al (1989). In addition, communication acts were classied by topic. Topics are combined into nine groups: (1) procedural communication (new procedures designed for the simulator ights), (2) data link system (statements, questions, answers or self-instructions related to the features and the use of the date link interface), (3) event-related elements (related to the scheduled events weather and technical faults, see Fig. 4), (4) other cockpit systems (e.g. autopilot), (5) about air-to-ground communication (e.g. speculations about ATCs intentions), (6) ight parameters (e.g. speed, altitude, heading), (7) ightplan, (8) ight (general) and (9) others. Some attention-related performance indicators were extracted from video recordings. According to the Crew Co-ordination Concept, pilots are obliged to inform the other crew member(s) about any deviations from the regular process. Video recordings were analysed as to whether, after which response time, and by which of the two pilots critical events were detected. Critical events were faulty ATC messages and technical failures (see above). An indicator for detection was the rst statement related to the critical event.

7.

1.19 (.98)

The pilots ratings on the usability scale (Table 3) for the concrete system are more homogeneous than the ratings for a data link in general. Mean ratings for the ten sub-scales as well as the total usability score tend to be about 3 on the scale ranging from 0 to 4. A total of 96.9% of all 160 (10 items 6 16 pilots) ratings are non-negative (a rating of 2 or higher) and 79.4 % are clearly positive (rating of 3 or 4), but only 14.4 % are extremely positive (rating of 4). Even if there is a clear tendency towards a positive rating of the system, it is not ideal yet. None of the two scales makes any restrictions on the scope of application. Table 4 shows that the data link interface is rated worse for take off and approach, mainly positive for climb and descent, and clearly positive for the cruise (100 % of all ratings are 3 or 4).

5. RESULTS
5.1. Acceptance of Data Link Communication The acceptance of data link communication comprises two aspects: the acceptance of a data link in general and of the communication interface used in the experiment in particular. The general attitude towards data link communication was measured as the degree of agreement with seven statements listed in Table 2. The right column of the table shows the mean scores of a ve-point scale where 0 corresponds to total disagreement and 4 to total agreement. Three of the scales are inverted, so higher scores correspond to a positive attitude towards a data link. The results indicate that pilots do not expect general problems regarding safety and errors (items 1 and 5), but they do not want to do without voice communication (items 2, 6 and 7). Furthermore, they do not expect that their job will become either less interesting (item 4) or easier (item 3) when using a data link.
Table 3. Mean (M) and standard deviations (SD) of agreement with usability items and overall usability score (n = 16 pilots); scales ranging from 0 (I totally disagree) to 4 (I totally agree); some scales inverted, so higher values indicate higher usability Item I think that I would like to use this system frequently The system was unnecessarily complex [score inverted] The system was easy to use The various functions of the system were integrated well There was to much inconsistency in this system [score inverted] Most pilots would learn to use the system very quickly The system was very cumbersome to use [score inverted] I felt very condent using the system I had to get used to a lot of things before I could get going with this system [score inverted] The technical changes (ND, MCDU pages) and its use t well into the overall cockpit design (Airbus philosophy) Usability score M (SD) 3.00 3.13 2.94 3.00 2.88 3.13 3.00 2.62 2.44 (.63) (.34) (.57) (.63) (.72) (.81) (.63) (.62) (.81)

2.94 (.57) 2.91 (.35)

220
Table 4. Ratings of the suitability of the data link interface by ight phase Suitability for . . . Take-off Climb Cruise Descent Approach M (SD) 1.50 2.37 3.50 2.69 1.50 (1.21) (1.09) (0.52) (0.79) (1.03)

T. Muller and H.-G. Giesa

5.2. Demands The demands produced by the tasks within the ight scenario were compared according to the eight cells of the experimental schedule. Each cell contains data from four test sections of the ight (according to Fig. 4 high workload conditions exist in sections A, C, F and H, low workload conditions in sections B, D, E and G). The cognitive functions proles are shown in Fig. 6. The gure indicates that the overall load (count of coded cognitive functions during one test section) increases for both of the pilots with higher workload as well as with the switch from voice communication to data link. The overall amount of employed cognitive functions tends to be slightly higher for the pilot non-ying. The pilot yings task employs the functions of interpretation and observation more often than the pilot non-yings task. On the other hand, for the pilot non-ying there are more codings of execution. Note that a large portion (approximately half) of the execution encodings are due to the activity of communication. The function planning is found only under high workload conditions and is closely related to ight plan negotiations. With the exception of planning, the counts for the

cognitive functions correlate positively with each other (Spearman r = 0.33 . . . 0.71). Since the duration of the test sections was not xed, but depended on reaching certain waypoints or solving tasks, duration is, among other effects, a result of the crews performance. Therefore it is treated as an dependent variable, too (Fig. 7). Mean time consumption correlates positively with the counts of all four cognitive functions (r = 0.39 . . . 0.50). For an estimation of workload derived from the pilots cognitive functions, employed during ight, the temporal density of the employment of these functions is of special interest. Figure 8 shows the employment of the four cognitive functions per time. As opposed to the absolute

Fig. 7. Mean duration of ight sections (in seconds) for the eight factor combinations.

Fig. 6. Cognitive function proles for the eight factor combinations.

Effects of Airborne Data Link Communication

221

Fig. 8. Cognitive function per time for the eight factor combinations.

values in Fig. 6, the count of cognitive functions per time shows an inverted tendency: a decrease with higher workload and the switch to the new communication medium. 5.3. Pilots Subjective Workload The perceived workload was surveyed using the NASA Task Load Index (TLX). As depicted in Fig. 9, the group means of ratings were rather similar for all the six scales. The ratings on all the scales correlated positively with each

other (r = 0.41 . . . 0.81). As expected, mental and temporal demands were rated higher than physical demands. Typically there was an increase with higher workload (WL) and with the switch from voice communication (VC) to a data link (DL). Ratings were almost identical for both of the pilots. All the scales range from 0 to 100 (good to poor for the performance scale, low to high for all other scales). Since all group means were within a range from 10 to 50, the perceived workload might be characterised as low to moderate for all experimental conditions.

Fig. 9. Results from subjective workload ratings on the six scales of the NASA Task Load Index for the eight experimental conditions (group means).

222

T. Muller and H.-G. Giesa

effort and even frustration in an equal way: there is no signicant difference in any of the scales between pilot ying and pilot non-ying. In contrast to all other scales of the TLX, there are no signicant main effects of the three experimental factors on the performance scale. However, there is a signicant interaction effect of medium and workload (F = 4.99; p < 0.05). For none of the other scales were interaction effects found, i.e. the effects are additive. 5.4. Pilots Situation Judgment Pilots assessed the ight situation for the individual test sections in several dimensions retrospectively. Figure 11 shows the mean ratings in the dimensions safety, possibility of intervention and problems perceived for the eight factor combinations. The mean ratings of safety and possibilities of intervention are on a high level for all eight conditions. The safety of the ight sections is rated worst under high workload conditions with no respect to the communication medium. The possibilities of intervention are estimated worst under high workload combined

Fig. 10. Combined NASA TLX workload score for the eight experimental conditions (group means).

The combined score of the six scales of the NASA TLX (Fig. 10) shows the tendencies more clearly. It shows an additional impact of workload and communication medium on perceived workload where both of the experimental factors have effects of approximately the same size. Differences between the pilot ying and pilot non-ying are negligible. The TLX ratings (Fig. 10) reect the overall demands (Fig. 6) and duration of the test sections (Fig. 7) to a higher extent than the temporal density of the demands (Fig. 8). Mean ratings correlate positively with the duration (r = 0.58) as well as with the counts of cognitive functions (r = 0.25 . . . 0.66). There is an even stronger correlation (r = 0.62) between the additive count of all cognitive functions (as indicated by the total height of the columns in Fig. 6), but no correlation with the cognitive functions per time (r = 0.02). In Table 5, the results of an analysis of variance of the six TLX scores and the combined score are shown. The effects of the medium on the individual scales are not as evident as they are in the combined scale. The analysis of variance conrms that both of the pilots perceive demands,
Table 5. Results of an analysis of variance (F-scores) of six scales of the NASA TXL and the combined score by role (pilot ying vs. pilot nonying), medium (voice vs. data link), and workload (low vs. high) Variable Mental demand Physical demand Temporal demand Performance Effort Frustration Combined score Role 0.02 0.01 0.33 2.14 0.71 0.20 0.77 Medium 2.48 3.57 8.14* 4.54 7.72* 10.95** 22.09*** Workload 24.94*** 4.91* 17.37** 2.63 18.45** 14.03** 22.46***

* p < 0.05; ** p < 0.01; *** p < 0.001.

Fig. 11. Pilots ratings of safety of the ight situation, their possibilities of intervention, and the problems they perceived by experimental situation.

Effects of Airborne Data Link Communication


Table 6. Results of an analysis of variance (F-scores) of the variables safety, possibilities of intervention and problems by role (pilot ying vs. pilot non-ying), medium (voice vs. data link) and workload (low vs. high) Variable Safety Intervention Problems * p < 0.05. Role 0.21 0.07 0.81 Medium 0.06 0.43 8.45* Workload 8.44* 1.50 5.45*

223

with data link conditions. That is also the situation where both of the pilots perceive the most problems. In Table 6, an overview of the results of an analysis of variance of the same three variables is shown. The safety of the ight situation as it was perceived by the pilots was affected signicantly by the factor workload, even if the effect was rather weak in absolute terms. The effects of the experimental factors on the perceived possibilities of intervention remained within the range of chance. The perception of problems was affected by the medium as well as by the workload level, whereas the effect of the medium was slightly stronger. In the ratings on all three scales, there were no signicant differences between pilot ying and pilot non-ying. 5.5. Situation Awareness 5.5.1. Reported Knowledge of Flight Parameters In highly automated systems like the cockpit of modern commercial aircraft, the permanent knowledge of ight and system parameters is of primary importance. The pilots were asked to classify their knowledge of three groups of parameters retrospectively:
. primary ight parameters: ight level (altitude), speed, heading, rate of climb/descend; . system/engine parameters: fuel, thrust, vibrations, rotation speed; . navigation parameters: way points, actual position.
Fig. 12. Reported knowledge of ight parameters by experimental condition. Table 7. Results of an analysis of variance (F-scores) of the reported knowledge of ight parameters by role (pilot ying vs. pilot non-ying), medium (voice vs. data link) and workload (low vs. high) Variable Primary ight param. System parameters Navigation parameters Role 0.09 4.64* 4.31(*) Medium 0.00 0.74 0.43 Workload 0.32 0.63 9.00**

(*)p < 0.10; * p < 0.05; ** p < 0.01.

Figure 12 gives an overview of pilots ratings. For the primary ight parameters, both of the pilots reported a knowledge on a high level (note that the scale ranges from 0 to 4, where 4 represents perfect knowledge of the parameters). There are no signicant variations between the experimental conditions (see also Table 7). The overall knowledge of system parameters is on a lower level than the knowledge of the primary ight parameters and there are greater differences between the experimental conditions. The pilots non-ying report a signicantly weaker knowledge of these parameters. For the navigation parameters, the pilots ying report marginally better knowledge and both of the pilots report a signicantly better knowledge of these parameters under low workload conditions.

None of the three variables shows any evidence of an effect of the communication medium on the knowledge of ight parameters. There are no signicant interaction effects between the experimental factors. 5.5.2. Performance-Related Indicators of Situation Awareness In two of the eight test sections, the pilots received faulty messages from ATC. Video recordings from the cockpit were analysed with respect of the recognition of incorrect information. Under both of the conditions, voice as well as data link, the faults were identied immediately. Since the identication of these faults is closely related to the knowledge of primary ight parameters, these results are in accordance with the reported high level of knowledge of the primary ight parameters under all experimental conditions. Under voice conditions, a corresponding statement was mostly given rst by the pilot ying (six out of eight cases),

224

T. Muller and H.-G. Giesa

Table 8. Mean reported knowledge of primary ight parameters in ight sections B and E by role in cockpit (pilot ying vs. pilot non-ying) and response to faulty ATC messages First response Pilot ying Pilot non-ying 3.25(n = 8) 3.67(n = 6) Not rst response 3.16(n = 6) 3.12(n = 8)

5.6. Communication 5.6.1. Air-to-Ground Communication The introduction of a data link is assumed to primarily affect air-to-ground communication. On one hand, the data link interface used in the experiments offers several predened requests and reasons for an application or a refusal. These extended facilities were employed by the pilots. On the other hand, air-to-ground communication is simplied because it is point to point (no display or readback of a call sign is necessary) and no disturbances of transmission occur (no readback of messages). Finally, the average number of communication acts between cockpit and ground authorities is slightly reduced from 12.0 (voice) to 10.5 (data link) communication acts (uplink and downlink) per crew and ight section (F = 5.20; p = 0.057: marginally signicant). 5.6.2. Intra-Crew Communication The number of communication acts within the crew increased signicantly under data link conditions: 22.1 (voice) vs. 47.1 (data link) communication acts per crew and test section (F = 75.1; p < 0.001). This is partly due to the reading aloud of incoming messages by the pilot nonying with a data link: in data link sections, this was true for 3.6 messages on average. (Note that not all pilots followed the requested procedure consistently.) This part of intra-crew communication replaced only a part of air-toground communication. Corrected for these communication acts the difference remained signicant (F = 59.8; p < 0.001). Looking at the content of communication, we can see in Table 10 that the differences in quantity were mainly due to two types of communication: procedural communication (e.g. statements like Wilco and the reading of incoming messages mentioned above) and communication about the data link system (e.g. questions, answers, and self-instruction). Furthermore, the pilots talked more about the air-toground communication (e.g. speculations about the conTable 10. Number of communication acts on various topics per ight section and crew under voice and data link conditions (n = 8 crews) Content of communication Procedural communication Data link system About air-to-ground communication Flight parameters Flight plan Flight (general) Other cockpit systems Event related elements Others Total intra-crew communication Voice 0.34 1.13 0.97 2.44 4.00 4.00 2.37 3.91 2.91 22.06 Data link 7.53 8.81 1.97 4.91 4.22 6.16 2.50 6.63 4.34 47.06 ANOVA (F-score) 120.7*** 38.0*** 7.1* 3.1 0.1 1.2 0.0 1.9 11.2* 75.1***

but under data link by the pilot non-ying (in four out of six reliably analysed cases). This might be due to the fact that under data link the pilots non-ying were briefed to read aloud the incoming messages. But there is some correspondence between the individually reported knowledge of the primary ight parameters and the rst response to faulty ATC messages (Table 8): the respective pilot who made the rst statement about faulty information also reported better knowledge of the primary ight parameters. Furthermore, in two test sections technical faults occurred. The two faults (standby pitot probe heating and engine vibrations) were indicated in the regular way on a visual display: the display of the message of the defective pitot tube heating was accompanied by an acoustic alarm. The mean time from signal occurrence to the pilots response was 2 seconds (voice) vs. 3 seconds (data link) respectively for the pitot tube heating alarm. Mean response times for the engine vibration alarm were 3 seconds (voice) vs. 50 seconds (data link) respectively. Under voice conditions, the rst response came in seven cases from the pilot ying and in one case from the pilot non-ying. Under data link conditions, the rst response came three times from the pilot ying and four times from the pilot non-ying (for one case data are not available). The detection of signals indicating technical faults is related to attention toward system parameters. Table 9 gives an overview of ight section G (pitot probe heating failure): the pilots who rst respond to the signal tend to also report better knowledge of system parameters. The delay in the detection of the engine vibration signal under data link conditions was obviously caused by a distraction of the pilots attention: in the case where the fault was detected rst by the pilot non-ying, the pilot ying was occupied with an MSP ight plan proposal. The longest delay (94 seconds) occurred in a crew where both of the pilots attention was drawn by a new ight plan. Flight plan negotiations under voice conditions did not have such an effect on pilots attention.
Table 9. Mean reported knowledge of system parameters in ight section G by role in cockpit (pilot ying vs. pilot non-ying) and response to signal (pitot probe heating failure) First response Pilot ying Pilot non-ying 2.00(n = 4) 2.25(n = 4) Not rst response 1.75(n = 4) 1.50(n = 4)

Effects of Airborne Data Link Communication

225

trollers intention). Since the ight sections last about 50% longer under data link conditions, there was a general tendency to communicate more on most of the topics, although communication about ight plans and other cockpit systems (e.g. autopilot) was almost constant. 5.6.3. Structure of Communication Communication acts may be divided into two types: initiative acts and response acts. Initiative act like statements, questions or commands may start a sequence of communication acts on a certain topic. They do not require a previous communication act. On the other hand, response acts like answers or conrmations follow a previous initiative act. Since not every initiative act is followed by a response act, there are more initiative than response acts (65% vs. 35%). This ratio does not depend on the communication media, i.e. this ratio is found under voice conditions as well as under data link conditions. But there is a remarkable distinction between pilot ying and pilot non-ying: under voice conditions, initiative communication acts comprise 72% of the pilot yings speech acts directed to the pilot non-ying and 60% of the pilot non-yings speech acts directed to the pilot ying. Under data link conditions this ratio is inverted: only 57% of the pilot yings speech acts are initiative as opposed to 73% of the pilot non-yings. The interaction effect of role 6 medium is statistically signicant (F = 30.1; p < 0.001). This effect becomes weaker after removing the reading of incoming messages by the pilot non-ying from calculation (F = 18.8; p < 0.01).

6. DISCUSSION
The results for the general attitude towards data link communication indicate that pilots do not expect general problems with the medium. A clear tendency towards a positive rating of system usability (79% ratings are clearly positive) was found in the study. This conrms the results from our previous usability studies. But even though recommendations from literature were taken into account and the interface was tested in previous usability studies, the system was not yet regarded as ideal. Two restrictions to the application of a data link found in other studies were conrmed. First, the acceptance of a data link depended on ight phases. It was rated best for cruise and worst for approach and take-off. Second, a data link should not replace VHF communication completely. Voice communication should be available at least as a backup system. A complementary concept based on data link and VHF communication would probably lead to the highest acceptance. An analysis of the demands produced by the tasks shows that the frequencies of all cognitive functions increase with higher workload and with the switch from voice to data

link communication. This indicates that a data link as well as higher workload leads to more or more complex activities within the cockpit. An increase of activities is in line with the results from a communication analysis in that a growing frequency of communication acts was observed for data link communication. The consideration of time used for each test section shows that the pilots extended the time for fullling tasks when demands increase. Although demands increased slightly with the use of a data link and an increase in workload, time-related scores decreased, as duration grew faster than demands. The temporal density of cognitive functions was lowest under those conditions where pilots perceived the most problems. This points to an effective personal resource management which caused the pilots to slow down activities for the purpose of keeping their workload with certain limits, even if demands increased. The decrease of the number of demands per time might be interpreted as an overcompensation. Obviously the increased use of time was not just an effect of increased transaction times or a complex handling of the interface. It was rather a result of more complex changes in work habits as the example with the delayed detection of the engine vibration signal shows: the involvement of both pilots with an ATC ight plan proposal indicates inappropriate crew resource management. The results of the subjective workload ratings show an additional impact of increasing demands and communication workload on the perceived workload. The increase of subjective workload ratings for a data link is in the same range as the increase induced by higher workload. The most disadvantageous combination of experimental factors was high workload (which also represents high demands) in combination with data link communication. Since these factors do not enhance each other (no effect of interaction was found in ANOVA), the total increase of subjective workload ratings may not be very critical. It can be assumed that at least a part of the higher subjective workload can be attributed to the unfamiliarity of the pilots with the system and will decrease with time and practical experience. But, the increase in subjective workload is in line with the results for the acceptance of a data link. In ight phases with high demand such as during approach, a data link seems to be less suited than for ight phases with lower demands. Remarkably, the TLX scores reect the demands in the test sections in an integrative way, i.e. they might be interpreted as something like momentary effort times duration. Therefore, the results are compatible with the hypothesized compensation strategy. Since no negative effects of a data link on the evaluation of safety and on possibilities of intervention were observed during cruise, a data link seems to be applicable for this ight phase. The compensation strategy mentioned above can completely offset only the effects of a data link.

226

T. Muller and H.-G. Giesa

The reported knowledge of ight parameters indicates no negative effects of data link communication on the related aspects of situation awareness. Furthermore, there is no obvious inconsistency with performance data regarding the detection of ATC faults or signals. On the other hand, there are (at least in individual cases) distractions of attention which are precarious. These problems are attributable to ight plan negotiations (and not the communication medium per se) and they are therefore crucial for the entire concept of multi-sector planning. On the ight deck, the problem seems to be a matter of crew resource management as mentioned above not the optimisation of the data link. Regarding communication aspects, many ndings of prior studies were veried. Even if the measurement of pure transaction times seems not to be very meaningful in a close-to-reality setting, the more global indicator of the duration of test sections conrms the hypothesis of increased transaction times in terms of practical consequences. The reasons are manifold and not restricted to aspects of the handling of the interface. They are rather also a result of changes in the personal and the crew resource management as discussed above. The effects of a data link on air-to-ground communication seem to be similar to the substitution of telephone calls by electronic mail messages in an ofce environment: the receiver is not impelled to respond immediately but can nish other tasks or consider the answer more thoroughly. In addition, in the cockpit the receiver of the message is not an individual person, but a crew, i.e., they get the opportunity to discuss the topic. Some specic features of our data link interface support this tendency. For instance, the graphical representation of a ight plan proposal received from the multi-sector planner on the Navigation Display is a good basis for a discussion. This was used by the pilots. Under voice conditions, none of the crews discussed the list of waypoints they received via VHF voice radio in the equivalent procedure. All this should improve the quality of decisions on the ight deck, although it may become crucial in situations where quick responds are necessary. To some extent a possible increase in transaction times will be offset by a reduction and simplication of airto-ground communication in terms of the number of communication acts. The increase of intra-crew communication is partly due to procedures adapted to a data link and replaces communication with the ground to some extent. It can be speculated that the overall increase in communication needs forms a temporal bottleneck which contributes to the increased duration of the test sections by slowing the solving of primary tasks. The analysis of the content of communication shows that a considerable part of the increase under data link conditions was due to communication about the data link system. It is plausible that

communication about this topic will decrease when pilots are more familiar with the system. The ndings on intra-crew communication patterns validate the study of Hrebec et al (1995): the pilot who was responsible for communication with ground authorities took a more active part in the intra-crew communication. In contrast to the ndings of that study, the changes do not inuence the relationship between captain and rst ofcer, but the relationship between pilot ying and pilot nonying. Generally an air-to-ground data link appears to have more and more severe consequences for intra-crew communication than for air-to-ground communication. Even when the analysis focused on the cruise phase, the pilots experienced data link communication from runway to runway. A clear preference for the use of the system in cruise is in accordance with the ndings of many other studies and seems to be independent from the concrete design of the interface. Since most communication with ground authorities in this ight phase is less time critical than in the approach or landing phase, increased transaction times seem acceptable. The few time-critical communication acts in cruise can be handled via voice. Alternatively, an additional feature for urgency notication could be advantageous. For practical reasons, mandatory maximum response times for both parties, aircrews and ground authorities, should be xed. Crew resource management under data link conditions needs some reconsideration. The use of a data link for airto-ground communication obviously supports a more active role of the pilot non-ying. The pilot non-ying shows better performance in detecting critical signals and governs intra-crew communication. In principle, these ndings also appear to be true regardless of the concrete design of the interface. The division of labour between the two pilots is also affected by another aspect. Dependent on the degree of functional integration of a data link, acts of air-to-ground communication impact on the control of the aircraft, e.g. when the acceptance and the activation of a ight plan are performed within the same procedure. We did not focus on this problem in our experimental scenario, but some pilots were apprehensive about this point in the debrieng interviews and discussions. There might be a contradiction between the responsibility for air-to-ground communication (and in a wider sense the handling of the data link interface) by the pilot non-ying and the responsibility for controlling the aircraft by the pilot ying. The problem described above, i.e., that both pilots were concerned with a ight plan proposal and neither of them monitored displays, is also the result of ambiguous role allocation in the cockpit. The sequence of reception, decision about acceptance, formal acknowledgement and the nal activation of a ight plan proposal require the

Effects of Airborne Data Link Communication

227
Gawron, V. J. (2000). Human performance measures handbook. Erlbaum, Mahwah, NJ. Hancock PA, Williams G, Manning CM, Miyake S (1995). Inuence of task demand characteristics on workload and performance. International Journal of Aviation Psychology 5:6386. Hart SG, Staveland LE (1988). Development of a multi-dimensional workload rating scale: results of empirical and theoretical research. In Hancock PA, Meshkati N (eds). Human mental workload. Elsevier, Amsterdam, pp 139183. Hollnagel E (1998). Cognitive reliability and error analysis method: CREAM. Elsevier, Oxford, UK. Hrebec DG, Ineld SE, Fiedler FE (1995). The effects of datalink on ight deck intra-crew communication patterns. In Jensen RS, Rakovan LA (eds). Proceedings of the eighth international symposium on aviation psychology, Vol 1 2427 April 1995, Ohio State University, Columbus, OH, pp 700705. ICAO (1994). Human Factor Digest No. 11. Human factors in CNS/ATM systems (Circular 249-AN/149). ICAO, Montreal, Canada. Ineld S, Logan A, Palen L, Hofer E, Smith D, Corker K, Lozito S (1995). The effects of reduced partyline information in a datalink environment. In Fuller R, Johnston N, McDonald N (eds). Human factors in aviation operations: Proceedings of the 21st conference of the European Association for Aviation Psychology (EAAP), Vol 3. Avebury Aviation, Hants, UK, pp. 5156. Jones DG (2000). Subjective measures of situation awareness. In Endsley MR, Garland DJ (eds). Situation awareness analysis and measurement. Erlbaum, Mahwah, NJ, pp. 113128. Jorna PGAM (1997). Human machine interfaces for ATM: objective and subjective measurement on human interactions with future ight deck and Air Trafc Control systems. In Federal aviation administration and Eurocontrol, 1st USA/Europe air trafc management R&D Seminar, Saclay, France, 1729 June 1997 (http://atm-seminar-97.eurocontrol.fr/ jorna.htm). Kanki BG, Lozito S, Foushee HC (1989). Communication indices of crew co-ordination. Aviation, Space, and Environmental Medicine 60:5660. Kanki BG, Folk VG, Irwin CM (1991). Communication variations and aircrew performance. International Journal of Aviation Psychology 1:149162. Kerns K (1994). Human factors in ATC/ight deck integration implications of data link simulation research (MP 94W0000098). MITRE, McLean, VA. Knox CE, Scanlon CH (1991). Flight tests with a data link used for air trafc control information exchange (NASA TP 3135). NASA Ames Research Center, Moffet Field, CA. Lee KK, Sanford BD, Slattery RA (1997). The human factors of FMS usage in the terminal area. American Institute of Aeronautics and Astronautics, Reston, VA. Logsdon EW, Ineld SE, Lozito S, Mackintosh M-A, McGann A, Possolo A (1995). Cockpit data-link technology and ight crew communication procedures. In Jensen RS, Rakovan LA (eds). Proceedings of the eighth international symposium on aviation psychology, Vol 1, 2427 April 1995, Ohio State University, Columbus, OH, pp. 324329. McGann A, Morrow D, Rodvold M, Mackintosch M-A (1998). Mixedmedia communication on the ight deck: a comparison of voice, data link, and mixed ATC environments. International Journal of Aviation Psychology 8:137156. Moroney WF, Biers DW, Eggemeier FT, Mitchell JA (1992). A comparison of two scoring procedures with the NASA task load index in a simulated ight task. In Proceedings of the national aerospace and electronics conference, Vol 2, May 1992. IEEE, Dayton, OH, pp 734 740. Moroney WF, Biers DW, Eggemeier FT (1995). Some measurement and methodological considerations in the application of subjective workload measurement techniques. International Journal of Aviation Psychology 5:87106. Morrow DG, Lee A, Rodvold M (1990). Analysis of routine pilot controller communication. In Society of Automotive Engineers (ed). Managing the modern cockpit: third human error avoidance techniques

cooperation of both pilots in terms of their role as pilot ying and pilot non-ying as well as consideration of the roles as captain and rst ofcer. Further development of crew resource management and corresponding procedures which ensure a clear allocation of activities as well as of responsibilities is necessary. Acknowledgements This research was supported by a grant from the German Research Foundation (Deutsche Forschungsgemeinschaft DFG) under Fr 375/48-3. The authors are grateful to their colleague Geerd Anders, who was responsible for the technical implementation of the experiments in the ight simulator and without whose commitment and expertise in ight guidance the experiments could never have been performed in such a realistic way. We would also like to thank Dagmar Ernst and Tanja Kohler, who carried out considerable portions of the communication analysis and task analysis. References
Amalberti R (1993). Safety in ight operations. In Wilpert B, Qvale T (eds). Reliability and safety in hazardous work systems. Erlbaum, Hove, UK, pp 171194. Bainbridge L (1983). Ironies of automation. Automatica 19:775779. Billings CE (1991). Human-centered aircraft automation: a concept and guidelines (NASA TM-103885). National Aeronautics and Space Administration, Washington, DC. Billings CE (1997). Aviation automation: the search for a human-centered approach. Erlbaum, Mahwah, NJ. Boy GA, Cacciabue PC (1997). Optimisation of automation on the civil ight deck. In Harris D (ed). Human factors for ight deck certication (Proceedings of the European workshop to develop human factors guidelines for ight deck certication, London, 2526 March 1977). Craneld University Press, Craneld, UK. Brooke J (1996). SUS: a quick and dirty usability scale. In Jordan PW, Thomas B, Weerdmeester BA, McClelland IL (eds). Usability evaluation in industry. Taylor & Francis, London, pp 189194. Byers JC, Bittner AC, Hill SG (1989). Traditional and raw task load index (TLX) correlations: are paired comparisons necessary? In Mital A (ed). Advances in industrial ergonomics and safety. Taylor & Francis, London, pp 481485. Endsley MR (1995). Toward a theory of situation awareness in dynamic systems. Human Factors 37:3264. Endsley MR (1996). Situation awareness in aircraft. In Hayward BJ, Lowe AR (eds). Applied aviation psychology: achievement, change and challenge. Avebury, Aldershot, UK, pp 403417. Endsley MR, Kiris EO (1995). The out-of-the-loop performance problem and level of control in automation. Human Factors 37:381394. Eurocontrol (1998). ATM strategy for 2000+ (Vol 1). Eurocontrol, Brussels, Belgium. Eurocontrol (1999). European Air Trafc Control Harmonisation and Integration Programme (EATCHIP): Operational Concept Document (OCD) (FCO.ET1.ST07.DEL01) (Edition 1.1; 4 January 1999). Eurocontrol, Brussels, Belgium. FAA Human Factors Team (1996). The interfaces between ightcrews and modern ight deck systems. Fricke M, Dehn D, Muller T (2000). Co-Operative Air Trafc Manage ment (Fortschritt-Berichte VDI Reihe 22, Heft 4). VDI, Duesseldorf, Germany.

228
conference, Dallas, TX, 45 December 1990. Society of Automotive Engineers, Warrendale, PA, pp 3540. Muller T, Giesa H.-G, Anders G (1999). Entwicklung und Evaluierung einer bordseitigen Benutzungsoberaeche fuer Data-Link-Kommunikation zwischen Piloten und Fluglotsen [Design and evaluation of an airborne interface for data link communication between pilots and air trafc controllers]. MMI-Interaktiv, No. 2, October 1999 (http:// www.mmi-interaktiv.de/ausgaben/10_99/). Muller T, Giesa H-G, Anders G (2001). Evaluation of airborne data link communication. Aerosp. Sci. Technol. 5:521527. Navarro C, Sikorski S (1999). Datalink communication in ight deck operations: a synthesis of recent studies. International Journal of Aviation Psychology 9:361367. Orasanu J (1995). Situation awareness: its role in ight crew decision making. In Jensen R, Rakovan L (eds). Proceedings of the eighth international symposium on aviation psychology, Vol 2, April 1995, Ohio State University, Columbus, OH, pp 734739. Rehmann AJ (1995). Flightdeck party line issues: an aviation safety reporting system analysis (DOT/FAA/CT-TN95/12). National Technical Information Service, Springeld, VA. Rehmann AJ (1997). Human factors recommendations for airborne controllerpilot data link communications (CPDLC) systems: a synthesis of research results and literature (DOT/FAA/CT-TN97/6). National Technical Information Service, Springeld, VA. Rehmann AJ, Mogford RH (1996). Airborne data link study report (DOT/ FAA/CT-N95/62). US Department of Transportation, Federal Aviation Administration, Washington, DC. Rehmann AJ, van Gent R, Bohnen H, Jorna P (1995). Flight simulator evaluation of baseline crew performance with three datalink interfaces

T. Muller and H.-G. Giesa


(DOT/FAA/CT-TN95/19). US Department of Transportation, Federal Aviation Administration, Washington, DC. SAE (1994). Human engineering recommendations for data link systems (SAE ARP4791). Society of Automotive Engineers, Warrendale, PA. Sarter ND (1995). Knowing when to look where: attention allocation on advanced automated ight decks. In Jensen RS, Rakovan LA (eds). Proceedings of the eighth international symposium on aviation psychology, Vol 1, 2427 April 1995, Ohio State University, Columbus, OH, pp 239241. Sexton JB, Helmreich RL (1999). Analyzing cockpit communication: the links between language, performance, error, and workload. University of Texas Team Research Project, Austin, TX. Sheridan TB (1983). Measuring, modelling, and augmenting reliability of manmachine systems. Automatica 19:637645. van Gent RNHW (1995). Human factors issues with airborne data link: towards increased crew acceptance for both en-route and terminal ight operations (NLR TP 95666). National Aerospace Laboratory NLR, Amsterdam, The Netherlands. White RG (1991). Keeping the pilot in the loop. In Royal Aeronautical Society (ed). Human factors in advanced ight decks. Royal Aeronautical Society, London, pp 8.18.13. Wiener E, Curry R (1980). Flight deck automation: promises and problems. Ergonomics 23:9951011.

Correspondence and offprint requests to: T. Muller, Technical University of Berlin, Institute of Psychology and Ergonomics, Department of Human Machine Systems, Jebensstrasse 1, 10623 Berlin, Germany.

Das könnte Ihnen auch gefallen