IMPACT OF DELEGATING SPACING TASKS ON FLIGHT CREW ACTIVITIES PRELIMINARY RESULTS

Transcription

1 IMPACT OF DELEGATING SPACING TASKS ON FLIGHT CREW ACTIVITIES PRELIMINARY RESULTS Eric Hoffman, Isabelle Grimaud, Laurence Rognin*, Karim Zeghal Eurocontrol Experimental Centre, BP, 9 Bretigny, France Introduction The delegation of spacing task from the controller to the flight deck is envisaged as one possible option to increase controller availability. In addition, it is expected that flight crew would gain in awareness and anticipation by taking an active part in the management of the situation with respect to the concerned aircraft. For aircraft within an arrival stream, the delegation could consist in tasking the flight crew to determine and perform the necessary speed adjustments so as to maintain a given spacing to a lead aircraft. We conducted controller-in-the-loop simulations to assess the impact on the controller side []. The feedback from the controllers was positive: the delegation is perceived as useful and potentially leading to more anticipation and to an overall workload reduction. The results show with delegation, an overall reduction of manoeuvring instructions. In addition, eye movement analysis suggests that, even with very high traffic, controllers can still anticipate the sequencing of the arrival streams. Pilot-in-the-loop simulations were performed, to assess feasibility and performance for in-trail following [] [] [] and merging situations []. More recently, a pilot-in-the-loop experiment aimed at assessing both human performances and impact in terms of workload and monitoring activities []. Different speed control modes were proposed, from manual to fully coupled, all relying on the provision of a suggested speed on the primary flight display. Achieved spacing performances were satisfying, with a limited impact on workload, but a perceived head-down time higher. We conducted an initial assessment of feasibility in a simplified environment consisting of a fix based cockpit simulator with limited assistance to the flight crew to perform the spacing task []. Even though the focus was more on interface usability, a number of human factor issues has been identified, stressing expected benefits and possible limits in terms of workload, function allocation and safety. In order to go a step beyond and assess the impact of spacing on flight crew activity, we conducted another part-task pilot-in-the-loop simulation. The experimental design took advantage of the previous controller-in-the-loop simulations, and in particular its operational context and the validation objectives. This papers aims at presenting the preliminary results obtained. It is composed of four sections: background, experiment objectives, experiment plan and setup, and results. Background Starting with the analogy of visual separation, the proposed task allocation relies on the delegation of spacing tasks in which the flight deck is tasked to implement a solution defined by the controller. Restricting the delegation to implementation tasks (as opposed to decision making tasks) is expected to preserve controller authority and understanding of the situation ( mental picture ). The delegation of spacing is at controller initiative, who can decide to end it at any time. The flight crew however can only abort it in case of problem onboard such as a technical failure. The delegation applies to pairwise situations: one aircraft is delegated, the other being target. In terms of responsibility, as opposed to visual separation which implies a transfer of separation responsibility, the delegation of spacing can be seen as the use of a new instruction. Thus, the controller is responsible to issue the appropriate instruction to guarantee the spacing (and the separation), and the flight crew is responsible to follow it. CRNA Sud Est, F- Aix en Provence, * Pacte Novation, F-9 Issy les Moulineaux {eric.hoffman, isabelle.grimaud, laurence.rognin, karim.zeghal}@eurocontrol.int

2 The delegation covers two classes of application: sequencing operations in terminal areas, and crossing and passing applications in enroute airspace. For the sequencing applications that will be considered in this paper, in-trail and merging situations are proposed along with two levels of delegation (Table ). For remain (resp. merge), the flight crew is tasked to adjust speed to maintain the desired spacing (resp. at the merging point). For heading then merge, the flight crew is tasked to () select the heading, () resume navigation to the merging point when desired spacing is obtained, and then () adjust speed to maintain the spacing. Delegation level Maintain spacing Resume then maintain spacing In-trail Remain behind Heading then remain behind Merging Merge behind Heading then merge behind Table. Sequencing applications. The controller-in-the-loop experiment provided a pseudo pilot perspective. The assumption was that the overall reduction of manoeuvring instructions was equally spread among every aircraft, and not detrimental to some of them. The following figures (Figure ) show the repartition of both heading and speed instructions per aircraft. With delegation, it can be observed that more aircraft get fewer instructions (both heading and speed). Although this is a positive indication, this does not reflect the necessary speed adjustments that shall be performed in the flight deck. Experiment objectives Two steps guided the definition of the experiment objectives. The first one was based on the human factor issues previously evoked, the second one consisted in deriving flight crew metrics from controller ones. Number of instructions per aircaft Number of instructions per aircaft 8 8 Heading Number of aircraft Speed Number of aircraft Without With / aircraft not delegated With / aircraft delegated Figure. Repartition of heading and speed instructions per aircraft (LFPG arrivals, very high traffic, session, November ). Human factor issues Removing vs. adding tasks. Although the delegation introduces additional spacing tasks in the flight deck, it might relieve the flight crew of the time-critical handling of some instructions. This will be partly addressed here through the analysis of speed actions induced by delegations. Maintaining spacing vs. flying aircraft. The delegation may require concurrent tasks, with a risk of cognitive tunneling. We decided to cover this point in a future experiment through eye movement analysis. Monitoring target vs. controlling traffic. Providing traffic information to the flight crew might lead them to feel that they are in a position to play a direct role in traffic control. Since we decided on purpose to restrict the traffic display to the target aircraft (upon controller request), this point could not be evaluated. However, this will be investigated later with the introduction of the display of surrounding traffic (TCAS emulation) along with the party line.

3 Being aware vs. interpreting situations. This is related to the trade-off between sufficient and excessive information. This is captured when assessing the interface usability. Anticipating vs. Reacting. Today, in terminal areas, flight crew have to react to time-critical instructions. The goal-oriented delegation instructions are expected to expand temporal horizon, enabling anticipation of required actions and a smoothing of overall activity. To a certain extend, this will be addressed here through the temporal analysis of flying and spacing actions. Detecting vs. generating errors. Despite new risks of errors associated to a new task, the involvement of the flight crew in the spacing may provide redundant monitoring. We decided at this early stage to focus on nominal situations, and not simulate degraded or abnormal ones. Deriving flight crew metrics The analysis of the impact on controller activity resulted in the identification of metrics aiming to capture the human performances and their consequences in terms of effectiveness and safety. These high level dimensions also applicable to the flight deck guided the derivation of flight crew specific metrics and measures. In particular, the activity assessed through controller instructions are translated into the actions performed by the flight crew (Table ). Similarly, specific measures of quality of control, e.g. conditions of transfer to next sector, were translated to quality of flying, e.g. spacing accuracy (Table ). Experiment plan and setup Two independent variables were considered: the delegation types (remain, merge, heading then merge) and the target attitudes (stable then indescent, in-descent). Regarding the dependant variables, only a subset of the metrics presented previously discussed was analysed: speed actions (speed selection on the flight control unit and possible speed brakes) and spacing accuracy. In addition to objective data collected through system recordings (aircraft states, pilot actions, spacing values), we gathered subjective ones through questionnaires, NASA-TLX, observations, and debriefing. Dimensions Concept acceptability Human perspective Effectiveness Safety Controller metrics Acceptance Usability Individual workload Situation awareness Controller activity (EXC/PLC) Collective activity (TRM, inter-sector) Quality of control Pseudo pilot perspective Flight efficiency Control errors Delegation-related errors Flight crew metrics Acceptance Usability Individual workload Situation awareness Pilot activity (PF/PNF) Collective activity (CRM) Quality of flying Pseudo controller perspective Flight errors Delegation-related errors Table. Controller and flight crew metrics. Dimensions Flight crew metrics Flight crew measures Concept Acceptance acceptability Usability Individual workload Situation awareness Communication (radio occupancy, ) Flight crew Pilot activity perspective (PF/PNF) Flying tasks (actions, ) Monitoring (fixations, scanning patterns, ) Collective activity (CRM) Task distribution PF/PNF Co-ordination and cross check Flight efficiency (fuel, speed profile, ) Effectiveness Quality of flying Respect of scheduled time of arrival Spacing accurracy Pseudo controller perspective Respect of instruction (reaction time, ) Safety Flight errors Alarms, omissions, Delegation-related errors Loss of spacing, execution errors Table. Flight crew metrics and measures. The task of the flight crew consists in a fully automatic flight (including checklist, operational flight plan, automatic terminal information service, briefing) and a manual spacing. A dedicated cockpit interface was proposed, consisting in data input on the control and display unit and display cues on the navigation display (Figure ). The display cues can be divided into three main elements: A data block displaying among others, target ground speed and closure rate. An arc of circle (in magenta) indicating the desired spacing. A broken arrow representing the position on the trajectory where the spacing will be acquired.

4 A colour coding was used to highlight drifting situations. No speed advisory was provided during the maintaining phase. Figure. Navigation display with the representation of the target aircraft. The target aircraft is represented by the white chevron. The merging configuration is highlighted by a double dashed line linking target and delegated (own-ship) via the converging point. Five crews with full Airbus rating participated ( test pilots and 8 airline pilots), one day each. Five runs per day were carried out, with pilots alternating function (pilot flying, pilot not flying). The flights were Paris Orly arrivals, from cruise to final approach and lasted approximately minutes. The target aircraft was under conventional control and received a speed reduction. Only a limited training could be provided, thus leading to gradually introduce the different applications: remain, merge, heading then merge. The limited realism of the simulation environment should be noted: it is a part-task experiment with rather short flights, the cockpit simulator only features automatic related instruments with no external view, no party line and no TCAS display. Results During each individual crew debriefing, pilots stressed the positive aspect of taking an active part in the management of the spacing (be in the loop) and gaining a better understanding of the situation (with respect to the preceding aircraft). In addition, they mentioned that delegation would probably enable more anticipation (be less in a reactive position) and an optimised management of their flight. However, they clearly stressed that a new task is added in the flight deck, with a potential risk of workload increase. In terms of assistance, they acknowledge the usability of the existing interface, but requested additional guidance cues such as spacing trend and speed bug, and eventually a fully managed (coupled) mode. In terms of simulation realism, the display of TCAS traffic and the party line was felt as necessary. In terms of objective results, as discussed previously, two aspects will be considered here: speed actions and achieved spacing. For illustration purposes, two typical examples of speed and spacing profiles for heading then merge are given in (Figure and Figure ). It can be observed in the first one, although the resume action was initiated slightly too early (thus inducing a speed reduction then acceleration), the spacing was achieved within ±.NM with regular speed reduction during the descent. In the second example, although the resume action took place on time, the spacing varied within ±.8NM and there were switches between descent modes, selected and managed. speed (kt) spacing (NM) 8 Dir-To OP OP DES DIR TO speed actions 9 Figure. Example of heading then merge speed and spacing profile. Dark line on the speed profile graph represents the selected speed and the light one the actual airspeed. The x axis is a time line.

5 The resume action is indicated by a DIR TO label on the speed profile graph. speed (kt) spacing (NM) 8 DIR Dir-To TO OP DES OP DES 9 DES DES OP DES OP speed actions Figure. Examples of heading then merge speed and spacing profile. As an indication of the performance of spacing, the figure below (Figure ) shows the spacing deviation achieved during the maintaining phase ( seconds after the acquisition phase). It always remains below the tolerance margins (set to ±NM), despite the basic assistance and the noncoupled mode Spacing deviation (NM) after sec Spacing (maximum) Spacing (9%) Spacing (average) Figure. Distribution of spacing deviation (maximum, at 9% and average) during maintaining phase (for all applications). To get an insight on the flight crew activity, we analysed the speed actions necessary to achieve such spacing. The figure below shows the average number of speed actions per minute against average spacing (Figure ). One assumption was that more actions (i.e. speed adjustments) would lead to a more accurate spacing. This is not confirmed by the results, and no clear trend emerges. No correlation can be made neither when considering delegation type nor target attitude. It is not clear at this stage if this results from the experiment setup (e.g. small sample size, sensitivity to initial conditions) or from the measures being simply not appropriate (e.g. use of average or maximum spacing, number of actions). Average nb of actions per minute,,,,, HMB MEL HMB OKR RB ATN MB MEL MB OKR,,,,,,8, Average deviation (NM) after sec RB ATN MB OKR MB MEL HMB OKR HMB MEL AVERAGE Figure. Average number of speed actions against average spacing deviation (for all applications and for each delegation type). RM, MB, and HMB denote respectively remain, merge and heading then merge applications, while OKR(IX) and MEL are the converging point (for OKRIX, the target aircraft is in descent). In terms of impact on flight crew activity and workload, we considered the temporal distribution of speed actions. Not surprisingly, there are peaks of speed actions following changes of the target aircraft (descent and speed reduction). In some cases, up to five actions per minute were observed, which raises the issue of workload and compatibility with other flying tasks. By the way, defining an action is also an issue. Typically, there is no difference between a speed modification and a succession of speed adjustments, except the time interval between each increment of speed (here, a time interval of seconds was chosen). Beyond number and distribution of actions, we also investigated the magnitude of speed variations (Figure ). In most cases, actions correspond to speed adjustments (within and +knots) rather than large changes. However, some large changes can be observed (more than knots). This needs to

6 be carefully analysed since this may impact both flight crew and controller workload (by possibly requiring more monitoring) and flight efficiency. This may also have an impact on the following aircraft (if any), and eventually create oscillatory effects on the whole sequence of delegated aircraft (if any). [...;-] ]-;-] ]-;-] ]-;-] ]-;-] ]-;-] ]-;-] ]-;[ ^ [;[ [;[ [;[ [;[ [;[ [;[ [;[ [;8[ [8; ] Distribution of speed variations in knots (merge behind OKRIX) Figure. Distribution of speed variation (merge behind OKRIX scenario). Conclusion The pilot-in-the-loop experiment enables to gain some initial understanding of the impact of the delegation of spacing tasks on flight crew. Pilot stressed positive benefits, mainly being in the loop, but also a potential risk of workload increase. Initial results show that spacing could be maintained within tolerance margins, with globally relatively few and rather small adjustments. Some assumptions could not be confirmed however, such as the correlation between number of actions and spacing accuracy. The impact of descent mode was not assumed and needs to be considered for future experiment, typically is it a dependent or independent variable? The next experiment will feature TCAS display and party line through the full replay of a controller-in-the-loop experiment, along with improved assistance. Two main points will be addressed. Firstly, the relation between the required spacing accuracy and the activity (workload). Secondly, the impact of delegation on the monitoring and the scanning patterns. Acknowledgements As part of the Air Ground Co-operative ATS Programme, the Cospace project is sponsored by Eurocontrol Experimental Centre (EEC) and European Air Traffic Management Programme (EATMP) of Eurocontrol. The authors would like to thank test and airline pilots for their participation in the experimentation and the technical team for their support. Special thanks to Nayen Pene and Carol Sheehan (preparation and analysis), Anne Papin, Renato Sanhueza and David Antoine (cockpit simulator). References [] Grimaud I., Hoffman E., Rognin L., Zeghal K. (). Assessing the impact on the flight deck of delegation of separation tasks. Digital Avionics Systems Conference, Florida, USA. [] Grimaud I., Hoffman E., Rognin L., Zeghal K. (). Limited delegation with arrival streams: more insight on its impact on controller activity. AIAA Conference on Guidance, Navigation and Control. Monterey, USA. [] Kelly J.R., Abott T.S. (98). In-trail spacing dynamics of multiple CDTI-equipped aircraft queues. NASA TM-899. [] Oseguera-Lohr R.M., Lohr G.W., Abbott T.S., Eischeid T.M. (). Evaluation of operational procedures for using a time-based airborne interarrival spacing tool. In Digital Avionics Systems Conference, Irvine, CA. [] Pritchett A.R., Yankosky L.J. (998). Simultaneous design of cockpit display of traffic information and air traffic management procedures. SAE Transactions, Journal of Aerospace. [] Sorensen J.A., Goka T. (98). Analysis of intrail following dynamics of CDTI-equipped aircraft. Journal of Guidance, Control and Dynamics, vol.. [] Williams D.H. (98). Self-separation in terminal areas using CDTI. Proceedings of the Human Factors Society Annual Meeting.