1. Introduction
From beginning as a military option, in a short time, uncrewed aircraft have been taken on as a useful tool in many applications within civilian operations. Those applications are becoming very diverse and will continue to be used in ever more industries as new technologies are developed. It is predicted that non-military users of RPA will make up nearly 30% of drone activity [
1]. Remotely Piloted Aircraft Systems (RPASs) are a sub-set of unmanned aircraft systems. The RPAS consists primarily of three components, the remotely piloted aircraft (RPA), of which there is a wide range of designs, the control station (remote pilot station) and the command-and-control link (C2). CASA [
2] in Australia differentiates between an RPA and a remotely piloted aircraft. The former includes remotely piloted aircraft but not kites, balloons or model aircraft.
Within the United States there are more than 1.6 million registered unmanned aircraft with over 10% of the registrations being for aircraft used for commercial operations [
3]. The rate of growth of the civil RPA sector over the next decade is forecast to be greater than that of the military sector with an annual growth rate of over 12%. The revenues for the civil sector up to 2028 are projected to be more than USD 88 billion [
1]. This increased activity will require a large increase in the number of pilots required to fly the burgeoning fleet.
With the ever-increasing range of Remotely Piloted Aircraft System (RPAS) operations and flight hours there has been an attendant increase in incidents and accidents involving RPA. While in modern conventionally crewed flight the most dominant factor in accident and incident causation has been HFs, recent research analyzing RPAS accidents has indicated mechanical and equipment issues being the most dominant factor in RPAS accident causation [
4,
5]. With RPASs being an emergent technology it is not surprising that technical issues are a large contributor to RPAS accidents. However, Cooke and Pedersen [
6] note that, despite the importance of human factors in unmanned flight, there is a surprising lack of attention given to HFs in this sphere of aviation. Elbanhawi et al. [
7] expressed surprise at the prevalence of technical issues over human factors in Micro Air Vehicle (MAV) operations. It was expected that, owing to the challenging physical settings within which MAVs operated, along with human limitations, there would be a greater HF influence on the causes of accidents.
RPAS technical issues that are currently the leading cause of accidents need attention and solutions. However, RPAS HFs should also be given attention if the lessons learnt in conventionally crewed flight can be extrapolated across to the RPAS sphere. At the beginning of conventionally crewed powered flight, 80% of accidents were caused by mechanical or technical issues. This was not surprising as the aviation industry was in its infancy, the early aircraft were fragile and underpowered and technological advancements were slow in being developed. Even after the developments in World War II, there were major technical failures such as the de Havilland Comet. These technical causes of accidents receded during the last decades of the twentieth century as improved technology made aircraft progressively safer and HFs became the main cause of 60–80% of accidents [
8].
As the technology of aircraft improved, allied with the strengthening of regulations devoted to technological improvements, the accident rates, while trending downwards, did not fall to zero. Pilot error was increasingly identified as the cause of aviation accidents [
9]. To meet this challenge of human failures, the ICAO, in 1994, issued their Human Factors Digest No. 7 [
10] to address the need for aviation to investigate HFs in accident causation.
Despite the human operator not being co-located with the aircraft during RPAS operations it has been found that, just as in conventionally manned flight, uncrewed flight is also negatively influenced by HFs. The remote pilot does not receive tactile feedback from the aircraft. The viewing of the visual field and the relationship of the RPA to items within the visual field are uniquely different for the remote pilot. Whilst Wild et al. [
4] identified technology issues as a major contributor to RPAS incidents and accidents, further research by Wild et al. [
5] identified the growing importance of human factors in the contribution to RPAS accidents. HFs were identified as the second largest contributor. There has also been both a lag in relevant regulations for RPASs and inconsistent national development and application of RPAS rules and regulations [
11]. This has applied to regulations for both the technical development of RPAS hardware and the licensing of people who operate RPASs.
3. Methods
A post-incident analysis methodology has been used by Wild et al. [
4] as an effective means of identifying factors that could lead to improving safety. Drupsteen and Guldenmund [
34] describe the process an organization can undertake in learning from incidents (LFI) as, “detecting events, by reflecting on them, by learning lessons from them and by putting these lessons into practice to prevent future incidents” (p. 81). Although this process has been described as tombstone safety [
35] it has a long history in aviation studies. Flanagan [
36] sought to understand the critical requirements of the job of an airline pilot using data that included “critical pilot behaviors reported in accident records, and critical incidents reported anonymously in interviews by the pilots themselves” (p. 330). The use of occurrence reports for LFI and the enhancement of safety outcomes emphasizes the importance of all participants in aviation operations, including RPAS operators, reporting incidents and accidents [
37,
38]. Once the data were gathered they were analyzed and classified into the critical requirements of the airline pilot. It is the learning from the commonalities of the incidents that provided the patterns that lead to understanding and further development of relevant and critical tasks for a job [
39].
For this study, the critical incident reports came from the database of RPAS incidents and accidents of the Australian Transport Safety Bureau (ATSB), the Australian government’s independent agency charged with investigating transport incidents and accidents, including those involving aviation occurrences. The ATSB reports are publicly available.
Reports that had HF issues as the seminal cause of the incidents and accidents were examined. The period of reporting was across eleven years from 28 May 2008 to 31 December 2019. The database contained 290 occurrence reports concerning RPAS operations, although only 10 of the reports had been investigated (3.4%). The remainder were pilot self-reports.
Information that was provided in the ATSB reports included the location of the occurrence, aircraft manufacturer, operation type, airspace type and class and a report of the occurrence. All reports were assessed for HF causation and, if there was an identified HF cause of the incident or accident, it was then codified using Wiegmann and Shappell’s 2003 version of the HFACS taxonomy [
18]. Of the 290 reports in the database there were found to be 34.5% or 100 reports with HFs as a leading cause. After coding to one of the four levels of the HFACS, a second more fine-grained analysis utilizing sub-levels [
18] within each of the main levels of the HFACS was carried out.
After identifying the frequency of occurrences within the HFACS a further analysis of the occurrences was made with a Pareto analysis. This provides an 80/20 analysis whereby 80% of the RPAS incidents and occurrences can be identified as arising from 20% of the HFACS levels and sub-levels. This allows for the identification of areas that, when priority is afforded to them, will lead to the largest improvements in RPAS safety outcomes.
Coding was performed by the first author of the study. His experience includes teaching human factors and accident causation over a nearly 25-year career in universities in multiple countries along with previous experience with coding accidents using a taxonomy. The quantitative data analysis including the Pareto analysis was conducted using Microsoft Excel.
To highlight the unique issues facing RPAS HFs, the results from the RPAS HFACS Level 1—Unsafe Acts—findings were compared to those Unsafe Acts for Australian powered aircraft as published by the ATSB [
24] and Australian general aviation as published by Lenne et al. [
20].
4. Results
The database of Australian RPA incidents and accidents was examined for HF causes of RPAS incidents and accidents and coded in the HFACS taxonomy. Of the 290 reports in the database, 100 reports (34.5%) were found to have human factors as a leading cause when codified against the different levels of the HFACS. The frequency of occurrences for HFACS levels and sub-levels is indicated in
Table 1.
All the RPAS occurrences that were found to have an HF causation were coded in the lower two levels of the HFACS, Unsafe Acts and Preconditions. These lower two levels of the taxonomy relate to the direct actions of the operator and the conditions surrounding the operation. There were no coded occurrences in the two higher levels that relate to the organizational issues that surround operations.
A chi-squared test, using the data presented in
Figure 1, for HFACS Levels 1 and 2 relative to a uniform expected distribution gives X
2 (5) = 55.19,
p < 0.001. As such, the null hypothesis can be rejected and the alternative hypothesis that the failure sources are not equal can be accepted. The significant contributions to this are the much greater than expected skill-based errors and the below-expectation violations and technical environment.
A Pareto analysis of the HFACS coded occurrences in Levels 1 and 2 of the taxonomy, shown in
Figure 2, indicates that skill levels of RPAS operators should be prioritized for lowering the rates of occurrences. The physical environment in which the RPAS operation takes place is also a leading cause of trouble for RPAS operators.
A fine-grained examination was conducted on Level 1 Unsafe Acts only using a Pareto analysis, shown in
Figure 3. This confirmed the predominance of operator skill level followed by perceptual errors as leading causes of RPAS incidents.
In their formulation of the HFACS, Wiegmann and Shappell [
18] describe possible causes of skill-based errors. These were used to try to further understand the underlying causes of skill-based errors. The occurrences originally coded to skill-based errors were further coded according to this list of possible causes of skill-based errors, shown in
Table 2.
The greatest number of skill-based errors arose from the poor technique of the operator when flying the remotely piloted aircraft. The second most frequent cause of skill-based errors was poor airmanship which can be closely aligned to technique. An example of skill error is seen in ATSB reference 201407488, “While conducting aerial survey operations, the remotely piloted aircraft (RPA) over-banked during a turn”.
The Pareto analysis of causes of skill-based errors, shown in
Figure 4, indicates these two underlying causes—poor technique and poor airmanship—are the two causes that should be focused on to reduce the largest number of occurrences.
The second largest cause of unsafe acts was perceptual error which accounted for 22 of the unsafe acts, shown in
Table 1. These perceptual errors were almost solely the misjudging of distance (
n = 21) with one report describing the drone pilot experiencing spatial disorientation. The misjudging of distance arises from a lack of depth perception. An example of this type of error is seen in ATSB reference 201808386, “While manoeuvring, the propeller struck a tree and the remotely piloted aircraft collided with terrain”.
The third largest cause of unsafe acts was decision error. While this has been seen in conventionally crewed flights to have a large impact on incidents and accidents, the number for RPAS operations is still small with only 9% of HF-caused occurrences being coded to decision errors, shown in
Table 1.
Using Wiegmann and Shappell’s [
18] possible causes of decision errors, the two largest causes of these decision errors are not following correct procedures and the lack of systems knowledge, shown in
Table 3. An incorrect RPAS procedure was illustrated in an incident during a flight off a beach in NSW, Australia. During the planning for the flight, co-ordinates for a northern hemisphere location were erroneously selected. During the flight there was a breakdown in the C2 link, leading to the aircraft automatically flying to the incorrect location. The aircraft was lost [
40].
From the second level of HFACS, the preconditions of the unsafe acts are environmental factors. While there was only one report that described the technical environment as leading to an unsafe act, the physical environment of wind (n = 9) and bird strikes (n = 15) contributed to 24 occasions that led to an unsafe act.
A comparison of the unsafe acts between the RPAS sector and conventionally crewed aircraft was made. The conventionally crewed sector was broken down to all powered aircraft from an analysis conducted by [
24] and specifically general aviation activity [
20], shown in
Table 4.
Skill-based errors were the single largest cause of incidents and accidents for all three sectors of the aviation industry. General aviation had the largest number of accidents attributable to skill errors whilst RPASs had the smallest number. However, the gap between RPASs and all powered aircraft was relatively small compared to the large gap between RPASs and all aircraft and the higher total of general aviation as seen in
Figure 5.
Decision errors were similar for both segments of conventionally crewed activity and these were more common than occurrences caused by decision errors for RPAS operations. This is not surprising as the influence of decision-making factors in accidents has long been recognized in conventionally crewed flight, e.g., O’Hare et al. [
8] identified the importance and influence of decisional factors in fatal aviation accidents in New Zealand.
Perceptual errors as a cause of occurrences were higher for RPASs than conventionally crewed aircraft. With perceptual errors being the second largest cause of skill-based errors, this finding further indicates the difficulties of depth perception and related obstacle avoidance for pilots not co-located with the aircraft. That all powered aircraft had lower perceptual errors than general aviation aircraft and RPAS operations could indicate that piloting experience levels have an influence on accidents caused by perceptual errors. With RPAS operations still a relatively new segment of the aviation sector, many drone operators may have yet to build substantial experience levels from which they can draw upon to avoid the occurrences arising from perceptual errors.
A chi-squared test, for the data presented in
Figure 5 (
Table 4), comparing RPASs to the expected distribution of powered aircraft, gives X
2 (3) = 105.7,
p < 0.001; hence, the null hypothesis can be rejected and the alternative hypothesis that the distribution of unsafe acts for RPASs is different to that of all powered aircraft can be accepted. Specifically, we note that decision errors are significantly less common while perceptual errors are significantly more common, with skill errors and violations being similar. Noting that, for powered aircraft, the percentage of perceptual errors is 6.1, this suggests that if RPASs were similar, we should expect 3.5 cases out of 75, and we observe 22 cases, which is 6.3 times greater.
5. Discussion
From the ATSB database of reports and investigations of civilian RPAS incidents and accidents across an eleven-year period from 28 May 2008 to 31 December 2019, most occurrences arose from technical issues. However, slightly more than a third of the occurrences (34.5%) were identified as having HF causes. If RPAS operations are to follow a similar historical trajectory to conventionally crewed operations, the number of HF accidents for RPAS operations will grow.
The occurrences identified as having an HF cause were codified using the HFACS, an established and well-used taxonomy that has been shown to have validity across a wide range of aviation sectors, other industries and different cultures. Most of the occurrences that had HF causes were coded to Unsafe Acts (75%), the lowest level of the taxonomy. This level of the HFACS relates to the actions of the individual pilots. A large number of occurrences being attributable to individual error was also identified in a review of U.S. Army UAS accidents where 93% of accidents arose from individual errors [
39].
The remaining 25% of the HF occurrences were coded to the physical environment and technical environments in which the RPAS flight took place, which comes from the second level of the HFACS [
41].
There were no HF-aligned occurrences from this study of civilian RPAS operations coded to the upper two levels of the taxonomy. These levels identify the causation antecedents of accidents as being in the areas of managerial oversight, planning and organizational influences such as allocation of resources and operating culture. For military RPAS accidents, however, it has been identified that the leading cause of accidents originated in these higher levels of the taxonomy [
33]. These researchers were able to establish links between unsafe acts at Level 1 with higher levels of the HFACS. Having a relationship between both the individual operator and the organizational and managerial spheres is the intention of the SCM, from which the HFACS has been developed.
That there was no such relationship for the studied civilian RPAS occurrences is not surprising as RPAS operations are an immature sector of the aviation industry and still are dominated by small companies using an owner-operator model. CASA lists over 2500 companies and individuals holding a Remotely Piloted Aircraft Operator’s Certificate (ReOC), the organizational certificate required for the commercial operation of RPASs in Australia. This is in comparison to 15 holders of an Unmanned Aircraft System Operator Certificate (UOC)—the forerunner of the ReOC—in 2012 [
42].
Further, there is a large difference in organizational structures surrounding civilian and military types of flying operations. In large organizations such as the different arms of the military, there are strong organization structures guiding operations. In the civilian RPAS sector of the industry, the search for latent conditions contributing to accidents may not yet be beneficial for ongoing safety enhancements.
The Pareto analysis of the first and second levels of the HFACS indicates that the areas that should be attended to for increases in safety outcomes are the skill levels of the pilots, the environment in which they fly and depth perception issues. These three areas of skill, environment and perception can be indicative of the current training systems not fully preparing the pilots for the demands of the tasks they are conducting. Although small in overall number, decision errors are caused by poor procedures and lack of knowledge is further confirmation that the RPAS pilots may not be fully and adequately prepared for operational flying.
The number of accidents attributable to skill level—or lack of—across most sectors of the aviation industry including RPASs, general aviation and all powered aircraft is worryingly high and must be of concern. Whilst RPASs have the lowest level of skill error accidents amongst the three groupings it is at a level that should be considered unacceptable, with efforts made to lower this cause of safety occurrences. The breakdown of the causes of the skill errors indicates that the poor technique and handling skills of RPAS pilots are by far the biggest cause of the unsafe acts, with nearly two-thirds of these unsafe acts being caused by a lack of handling ability. This reflects the training environment and the preparedness of student RPAS pilots for the operational challenges awaiting them. As the database covers the earliest days of drone operations, the lack of flying skills could have arisen from an earlier “park-flying” mentality in RPAS flying. The accessibility and affordability of drones made purchasing one easy and simple [
42]. The ab initio RPAS pilot could take the newly acquired drone to the local park and engage in self-instruction, resulting in drone pilots who were less than optimally prepared for operations. As Barsch [
42] comments, there was a lack of civilian operating experiences that were able to be drawn upon to shape RPAS pilot qualifications.
The influence of perceptual errors on RPAS operations is seen in these accidents, being the second largest cause of unsafe acts. The number of depth perception errors being so high indicates the difficulty of flying an aircraft when not co-located with it. A comparison with the causes of accidents for powered aircraft and specifically general aviation aircraft confirms the perceptual difficulties pilots face when not co-located with their aircraft, with nearly 30% of RPAS occurrences being linked to perceptual issues compared to 16% for GA aircraft and 6% for all powered aircraft. Depth perception is a challenging task and becomes exponentially more difficult the further away the aircraft is from the pilot and/or in cluttered flying environments such as operating in and around buildings. If the full potential of remotely piloted aircraft is to be fully realized, then the aircraft is to be flown at large distances from the pilot for VLOS operations and even further for EVLOS and BVLOS operations. For these operations, judging distance from obstacles becomes more difficult for the pilot.
The large impact of skill error and perception-caused accidents would indicate that the training and testing regimes of ab initio RPAS pilots warrant continuing examination to ensure graduating RPAS operators are prepared for the ever-increasing challenges of operations they will face in the future. With the maturing and developing of the RPAS sector of Australian aviation, CASA has introduced a comprehensive RPAS training syllabus with both experiential and practical competence requirements. The findings of this LFI study suggest that continued emphasis on the development of practical flying skills will assist in improving RPAS safety outcomes. A further priority given to depth perception training would be further warranted to develop safe RPA flying practices.
6. Reporting Systems
Developing a culture of reporting is an important step for the future safety of RPAS operations. Sieberichs and Kluge [
43] identified that too few incidents were being reported in the aviation sphere. Barriers to reporting incidents have included personal motivation, lack of recognition of the benefits of LFI and potential negative consequences for the reporting person [
44]. The type of incident also influenced whether a pilot would report it [
43] or the perceived inconsequential nature of the incident [
45]. Within Australian conventionally crewed aviation, it was found that nearly one-third of responding commercial airline pilots did not report safety occurrences to their respective airline voluntary reporting system. The reasons provided were fear of the report being used against the pilot by either the company or the regulator and the effort required to make a report being greater than the perceived benefits. Underreporting the occurrence—that is, only providing partial or minimal information—was also an issue raised in the responses from the commercial pilots [
37]. Walton and Henderson [
38] identified that over 85% of RPAS incidents in New Zealand in the period from 2015–2022 were not reported on any reporting system. The largest reason for the non-reporting was the RPAS operator not thinking the incident was serious enough to warrant a report. The lack of regulatory requirements was also identified as contributing to poor reporting. The researchers identified education regarding the importance of reporting as a way forward to improving low RPAS reporting rates.
This study utilized a postincident analysis methodology with the goal of LFI. For this type of approach to be successful there needs to be the reporting of incidents by RPAS operators. Of the 290 reports contained in the ATSB database, the vast majority of the reports—96.6%—are self-reports from the RPAS pilots involved in the occurrence. Up until the end of 2019 only 10 of the reports arise from an independent investigation carried out by the investigatory authority in Australia. Self-reporting has its limitations as it is a natural response for people not to be overly damning of their actions and nor to believe that these actions could not have led to an incident or accident. A previous study [
46], examining causes of accidents within New Zealand aviation, found limitations in pilot self-reports. As Zotov [
46] comments:
In some cases, the pilots may have been genuine in their belief of what happened, but (whether from incomplete perception or lack of knowledge) that belief was at variance with the facts found by the official investigators. In other cases, the investigators openly expressed their disbelief in the pilots’ veracity.
(page 73)
The value of using confidential incident reports might be reduced because of the limitations of self-reporting by the pilots involved. As the RPAS sector of aviation grows and matures it is hoped that funding will be made available to investigatory bodies to be able to complete independent investigations of not only accidents but also seemingly minor incidents to enhance the lessons that can be taken from these unfortunate occurrences. This, however, should not deter RPAS operators from making self-reports of safety occurrences. The importance of continuing to report RPAS incidents and accidents should be an important consideration for RPAS operators.