The effects of clutter on the visual search performance ...

Aviation Human Factors Division Institute of Aviation

AHFD

University of Illinois at Urbana-Champaign 1 Airport Road Savoy, Illinois 61874

3D Navigation and Integrated Hazard Display in Advanced Avionics: Workload, Performance, and Situation Awareness Display Clutter Issues in SVS Design Christopher D. Wickens, Ashley Nunes, Amy L. Alexander, & Kelly Steelman Final Technical Report AHFD-05-19/NASA-05-7 November 2005 Prepared for NASA Langley Research Center Hampton, VA Contract NASA NNL04AA20G

ABSTRACT Two experimental paradigms are examined to establish and model the role of clutter in visual search of aviation displays. The first paradigm involves searching for target aircraft within a cockpit display of traffic information. Clutter imposed by map overlay produces noteworthy costs. Clutter imposed by adding more aircraft imposed less consistent costs. The benefits of the clutter-mitigating solution of airplane highlighting are demonstrated. Only partial support is provided for the most straightforward model of clutter costs, the serial self terminating search model. The second paradigm is a more complex flight task in which the costs of a cluttered synthetic vision system (SVS) display in which traffic and terrain are overlaid, are contrasted with the scanning costs of a more spatially separated SVS display suite. Flight path tracking is equally preserved with both types of display design, and the two forms of costs trade off against each other regarding traffic and terrain awareness. Hence these results are consistent with a weighting model of display formatting that posits equal weights for the two costs. OVERVIEW The design of advanced aviation displays is confronted with two conflicting pressures. On the one hand, there is a desire to provide the pilot with a great deal of integrated hazard information, in which traffic, weather, airspace information and terrain may be displayed on a single map (Kroft & Wickens, 2003). On the other hand, such designs present the enduring problems of clutter. While a good deal is known about the problems of clutter in the search and readout of information (e.g., Yeh & Wickens, 2001; Teichner & Mocharnuk, 1979), less is known about how clutter effects can be precisely modeled, so that their influence on search time can be predicted. Valid quantitative models will have the advantage of enabling precise predictions of “how much clutter is too much” (a value of importance in display certification), and of predicting the benefits of solutions to clutter problems through highlighting or decluttering (Wickens, Alexander, Ambinder, & Martens, 2004). In the following report, we describe the results of two experimental paradigms. The first describes a series of four experiments with relatively simple aircraft traffic displays, to evaluate the viability of a classic computational model of visual search, when that search takes place against the background of a typical electronic map. The second describes an experiment in a more complex synthetic vision system environment designed to assess the influence of clutter. PARADIGM 1: CDTI SEARCH The serial self terminating search model of Neisser, Novick, and Lazar (1964) provides a foundation for such a model, stating that the search time through an array of N elements is predicted by the formula: ST = a + bN/2 when a target is present, and a + bN when the target is absent, where b is the time required to examine each item and determine that it is not the target, and a is the intercept term that characterizes the “readout” of the actual target. Such a model has provided a reasonable approximation of search time through military maps (Yeh & Wickens, 2001), and air traffic control displays (Remington, Johnston, Ruthruff, Gold, & Romera, 2000), as the number of items in the map are varied. The goal of the set of four experiments we describe was to establish the viability of the serial self-terminating model for pilots searching a traffic map over the background of varying types of ground map clutter, to establish the extent to which

1

the latter inhibits the search time/item (increases slope b), or the item readout time (increases intercept a). The first three experiments reported here, which evaluate the nature of the degrading effects of background clutter also provide a foundation for examining the benefits of highlighting in Experiment 4. Experiment 1: Baseline, No-Map Background The purpose of Experiment 1 was to examine the viability of the serial self terminating search model of an uncluttered (by background material) CDTI. Subjects Sixteen pilots from the University of Illinois’ Institute of Aviation participated in the present study. Pilots ranged in age from 18 to 23 years (mean = 19.3) and in experience from 15 – 300 flight hours (mean = 105.6). All pilots were paid $8/hr for their participation in the study. Task The participant’s task was to search an array of aircraft on a traffic display for a target. We first presented the pilot with a question pertaining to the altitude or trajectory of a specified aircraft in the array. The altitude question would vary in terms of whether it asked to locate an aircraft (a) above, (b) at, or (c) below a target altitude. The trajectory question asked to locate an aircraft either converging, diverging, or heading parallel to ownship’s course. Once this question, which appeared at the top of the screen was read, a key had to be pressed, following which the aircraft array appeared below the question box. Participants had to then search the array and click on the aircraft that served as the target for the specified question. On 10% of the trials, the target was not present amongst the aircraft array. In these situations, pilots were instructed to click the ‘NO TARGET’ button that was present at the top of the screen. Display Design The array consisted of multiple aircraft being presented against a blank background (Figure 1.1). The pilot’s own aircraft was always located in the center of the screen and was clearly denoted by the label “OWNSHIP.’ In contrast, the other aircraft had data tags, which conveyed the aircraft’s call-sign, altitude and speed. The heading could be inferred from the orientation of the aircraft icon. A typical data tag is shown in Figure 1.2.

2

Figure 1.1. Screen shot from Experiment 1.

Figure 1.2. Aircraft icon & data tag. Design and Variables A within-subjects design was used, with two independent variables being manipulated, namely target presence and aircraft array load. On 90% of trials (54 trials), the target was present whereas on remaining 10% of trials (6 trials), the target was absent.

3

There were either three, five or seven aircraft present on any array. Aircraft were placed randomly on the screen from trial to trial Procedure Upon arrival, ethical consent was obtained following which pilots were briefed on the experiment. They then engaged in a short practice session (consisting of 10 trials) to familiarize themselves with the task. Following this, the experiment began with each participant being required to conduct the target search across 60 trials. The order in which the trials were presented was randomized across participants to control for order effects. Results Response time data analysis (Figure 1.3) revealed a significant main effect for target presence (F(1,15)=102.72, p