text
stringlengths 4
179k
| predicted_class
stringclasses 5
values | confidence
float64 0
100
|
|---|---|---|
Some service professionals saw the lack of guidelines as somewhat advantageous in allowing for individualized and ‘creative’ care (HC001, HC003, HC004, HC010). This most commonly meant that service professionals could choose comparable non-transgender specific ICD or DSM codes that might better facilitate medical scheme payments for a service (HC007, HC010, HC011). It was also suggested by interviewees that the absence of guidelines allows for flexibility in their provision of care. For example: | study | 30.77 |
[…] I do treat them as guidelines every so often. I mean, I haven’t done, I haven’t said anyone must go for surgery early. But the gender affirming hormones and that, sometimes I think, you know, really, this kid’s ready. And I know [the plastic surgeon] has come and said to me ‘why must we wait until whenever’. Because it’s so distressing to the kid concerned (HC001). | other | 30.66 |
The public health sector in South Africa is extremely resource constrained , making pragmatic adaption of international guidelines important and necessary. This is also in line with the WPATH’s move towards context-specific guidelines in their most recent version: | study | 28 |
Another reason these professionals gave for not concerning themselves with advocating for national guidelines was that to do so would detract from the GAC that they are already struggling to provide, often in unsupportive institutional contexts (HC008-010, HC012). Those HCPs who have committed themselves to serving transgender and gender diverse clients do so in a severely constrained environment, and, with their time at a premium, saw spending time on the development of guidelines as beyond their capacity and the scope of their jobs (HC001). | other | 28.22 |
Some respondents were ambivalent about the role of national guidelines. In relation to the constraints imposed on reimbursement of care by medical aids, one psychologist noted, “[t]he care is the best you can give at the time” (HC007). At the same time, some service professionals wondered whether the absence of national guidelines was impeding access to funding by medical schemes (HC003, HC004). That is, having nationally derived guidelines that account for the contextually appropriate and available resources and services and which acknowledge the necessity of GAC could be useful in advocating for financial coverage (HC001, HC004). | other | 30.55 |
The absence of national guidelines for HCPs has a number of additional implications. First, while the HCPs we interviewed had consulted a variety of documents, both within and across the specific sectors of GAC, these guidelines are not necessarily aligned. And, because of the constantly evolving standards of care for GAC, some HCPs knew of colleagues who used outdated guidelines, or used outdated guidelines themselves (HC004, HC007, HC009, HC010). As a result, different HCPs have different ideas of what constitutes appropriate GAC (HC001, HC004, HC010, HC012). Because GAC is an interdisciplinary process, this can create significant problems as clients move from one HCP to the next, as the following quote illustrates with an example about the revised requirements (from SOC-6 to SOC-7) in order to initiate hormone therapy: | study | 30.08 |
We [the psychiatrists] have got a lot of [patients] coming back, with endocrines [the endocrinologist] saying the person is not coming dressed [aligned with their identified gender], and we say, “but SOC-7 doesn’t say that so…’' But it is an issue because at one stage it was that SOC-6 did have it [as a requirement] (HC007). | other | 30.34 |
Moreover, interviewees noted an unwillingness and/or inability of HCPs in the provision of appropriate GAC as another hindrance to access for clients (HC005, HC007, HC008, HC010, HC011). That is, psychologists and psychiatrists that make a diagnosis of GID/GD and who refer their client to an endocrinologist must first find an endocrinologist who is willing and able to provide gender affirming hormone therapy. A number of interviewees (HC001, HC005, HC009, HC010, HC012) felt that the absence of guidelines meant that they spent a significant amount of time advocating for access to GAC for their clients, both with other health professionals and health institutions: | other | 29.08 |
[…] the transgender services aren’t necessarily completely formalised. They’re still in the process of developing […] in the public healthcare system. I think across the country. So, in that sense, I think you need to be quite sensitive to how you actually deal with stuff and how you deal with the professional relationships (HC012). | other | 30.34 |
Lastly, as discussed above, the absence of national guidelines or protocol for the provision of GAC often required that service professionals be ‘creative’ in order to get their client the service they needed (HC003, HC007, HC010, HC011). As one interviewee pointed out, although this could facilitate access to GAC for some lucky clients, for other clients it implies a barrier to care which needs to be addressed: | other | 28.61 |
The part that bothers me around coding creatively is, are you coding creatively because you have to get around obstacles? So that your client can get the care they need? And I have a problem with that; because often the obstacles that you are trying to get around are obstacles that shouldn’t be there. Because they are obstacles informed by not up-to-date reasons and knowledge around the issue, or it’s informed by prejudice. So my natural activist is, no I don’t want to have to go around things to get for my client I want you to remove what is in the way […] of my client being able to get the care that they deserve (HC003). | other | 29.34 |
A natural consequence of the lack of structure and guidelines for providing GAC in South Africa is the limited number of informed and willing GAC professionals (HC009, HC011). As has previously been discussed, the vast majority of known GAC professionals and all of those interviewed for this study are located in urban centres of the Gauteng, KwaZulu-Natal, and Western Cape provinces. Thus, clients seeking GAC who reside in rural settings, or in one of the other six provinces, will often be required to travel to access care (HC002, HC009, HC012). But this has implications for the quality of services they receive over the long term. One service professional noted: | other | 27.64 |
[…] we’ve had patients from mostly Gauteng but also more rural areas, even though that makes it quite hard actually to follow up on services and things like that. Gauteng or [Johannesburg] is a bit better. But yeah, we do see people from the more rural governances and so on, even though it’s not ideal […]. Because, for instance, it’s hard to find professionals in the rural areas that are sensitized to transgender issues or who are sensitized to gender dysphoria (HC012). | other | 32.6 |
Even within those urban centres, though, it can be difficult to locate a HCP who specializes in or is even comfortable providing GAC. Many of the interviewees noted the importance of transgender advocacy and LGBTIQ community organizations in providing information and referrals to GAC service professionals (HC003, HC005, HC007, HC010, HC012). Based upon our interviewee’s responses, it would appear that clients’ ability to access GAC can rely heavily on their initial point of contact. That is, if someone is fortunate enough to have access to the LGBTIQ community or a GP with knowledge of transgender health, they are likely able to refer them to a doctor who provides GAC. And once that individual has seen that doctor, they often are part of a community of professionals—or at least know of other service professionals—providing complimentary components of GAC (HC001, HC005, HC007). As the following quote illustrates, though, it is rare to find a GP who has adequate knowledge, and if GPs are knowledgeable about referral pathways for GAC, they know so because of their initiative to seek out more information: | other | 30 |
What was clear from the interviewees was the importance of professional networks—the connection and sharing of information between GAC service professionals (HC001, HC004, HC007). The informal nature of GAC in South Africa both requires and fosters a collaborative approach to educating, providing, and advocating for GAC (HC009, HC012). This is seen in the formation of transgender units within one of the tertiary hospitals in South Africa, as well as in the tight-knit relations between professionals in the GAC community (HC001, HC003, HC004, HC007, HC008, HC010, HC012). A Cape Town-based HCP felt that, in addition to referrals from GPs, ten percent of their referrals came from Gender DynamiX, ten to twenty percent from other physicians at Groote Schuur Hospital, and sixty percent from a single clinical social worker with the Triangle Project (HC007). | other | 28.55 |
Gender DynamiX also hosts a biennial conference for the transgender community, which has been a significant venue for connecting GAC service professionals across the country and educating them on best practices and updated international guidelines. As one interviewee noted: | study | 30 |
[…] we understand that there are people like that doing work and when Gender DynamiX held a couple of conferences, a lot of people pitched up. So it was quite, it was nice that people in the business, we thought we were quite alone or we were feeling alone and we suddenly find other people also saying that there are a lot of people there and it’s hard to connect them (HC007). | other | 31 |
There are a number of challenges that HCPs face in their efforts to deliver GAC in South Africa. The first of these starts with the training they receive in their respective specializations, as transgender health-related topics are largely absent from health sciences curricula (HC005, HC009, HC011, HC012) . Thus, for the HCPs interviewed in this study, the education they have acquired regarding the provision of GAC is the result of individual interest and initiative. However, not all healthcare professionals will have the ability to take this on: | other | 30.12 |
So as to find someone willing or interested to learn. Because what I found is people are so overwhelmed with their work. There is not a lot of space. So […] I can see the patient load. To bring in a very minority group that needs healthcare services has created quite a challenge (HC009). | other | 31.6 |
For many HCPs, their gateway to GAC was a personal connection to the transgender and gender diverse community (HC009, HC011). For others, it was a specific patient or another HCP that initiated their pursuit of information (HC003, HC005, HC010). Either way, seeking information on GAC occurs out of HCPs’ own volition, if and when that volition is present. For example, most HCPs said they use the WPATH, or other preferred guidelines, because of referral by other HCPs, NGOs, or because of their own research initiative. | study | 29.44 |
The rarity of that engagement with transgender health-related topics and GAC has implications for both individuals seeking and professionals trying to provide GAC in South Africa. Because of the interdisciplinary nature of GAC, HCPs and clients rely on a number of health professionals, from GPs to psychologists, endocrinologists, pharmacists, and surgeons. Thus, one professional’s ability to provide a service often depends upon another professional’s ability and willingness to fulfil that service (HC005, HC007, HC008, HC010), which is where the previously discussed professional networks become so integral. To this end, a number of HCPs spoke of recruiting and advocating for GAC within institutions (HC004, HC007, HC008-010). And, when they engaged people and institutions for the first time on an issue with which they lack familiarity and comprehension, it could be a prolonged effort: | other | 29.05 |
[…] we are also trying to convince Greys [Hospital] to do [surgery]; that is our most important thing that we’re trying to push for and we’re still in process, a long educational route. Last year was Gender DynamiX—we did a kind of awareness program at Greys, as well, where we had a doctor come in and where, well, I presented from this side, then we had to, I think it was last, no, it was the previous year, 2014, we had that as well. So it’s basically education (HC009). | other | 31.94 |
Beyond the availability and accessibility of GAC professionals, professionals’ ability to provide services, and clients’ ability to fulfil these are often impeded by South Africa’s funding mechanisms (HC002-005). In the public sector, this is impacted by resource constraints. As one interviewee noted, the public sector hospital they work at is allocated a budget—based purely upon the number of patients seen—which is only ever adjusted for inflation (HC007). And because that funding is from the provincial health department, only residents of that province can access those services (HC001). The limited funding then has implications on how much of a service can be provided. For example, the surgeon in the Multidisciplinary Transgender Unit at Groote Schuur Hospital can only perform two to three gender affirming genital surgeries per annum based upon the theatre time the hospital will allocate towards such procedures (HC004, HC007) [26, 28]. | study | 30.64 |
Privately, access to funding for GAC services is equally challenging. Access to coverage of services is highly dependent upon the type of service, with psychology and psychiatry services being more commonly covered than gender affirming hormone therapy and gender affirming surgeries (HC004, HC007). Inherent to this challenge is the role that coding plays in billing for medical services, where the use of ICD’s F64 codes for diagnoses of GID–which may be specified under diagnoses for Transsexualism (F64.0), Duel role transvestism (F64.1), GID of childhood (F64.2), Other GIDs (F64.8), and GID, unspecified (F64.9) –or DSM’s codes for diagnoses of GD–which may be specified under diagnoses for Gender Dysphoria in Children (302.6) or Gender Dysphoria in Adolescents and Adults (302.85) —can result in non-coverage. Resultantly, health professionals resort to ‘creative coding’ to help their clients’ achieve coverage for the required services (HC003, HC007, HC010). In this context of medical aids refusing coverage, HCPs also suggested their provision of pro bono services (HC002, HC005, HC010), but this is not a viable approach for all GAC in South Africa. In addition, it does not address the problem of discrimination experienced by transgender South Africans in accessing health services, another issue raised by HCPs and noted in the literature (HC002, HC003) [23–25]]. | study | 30.81 |
Such discrimination poses a number of additional challenges to the provision of GAC. It is often difficult to find a willing HCP to provide GAC and sometimes even HCPs who are open to providing GAC do not have the necessary knowledge (HC005, HC010). This highlights the need for training on transgender health and GAC. And, as one HCP asserts, this is tied to the absence of national guidelines: | study | 27.52 |
[…] you know if you see a consultant or you see someone that is against transitioning or hormone replacement therapy, then [the client will] be blocked. So, that’s our biggest problem and barrier—trying to get a uniform system with uniform rules that everybody in our country would follow, which would make it so much easier (HC008). | other | 30.9 |
Moreover, HCPs felt that a state-sanctioned resource for patients and HCPs, which recognized GAC as appropriate and necessary might help combat some of the prejudice that transgender and gender diverse people face in the health system and beyond (HC002-004). | other | 31.69 |
The current provision of GAC in South Africa is inconsistent at best and non-existent at worst. While, technically, both the public and private health systems provide these services, there are disadvantages to both. As Klein recognizes, given that “income is still intimately connected with gender and skin colour, the situation is especially difficult for persons who did not grow up categorized as white males—but private care is much too expensive even for many who have been categorized this way” . Thus, while GAC is theoretically available to South Africans, practical access is unequal, can be extremely protracted, and is limited for all but the wealthiest. | other | 32.72 |
Though the South African health professionals that we spoke to have been resourceful in seeking out guidance in their provision of GAC, the absence of formal guidelines has a number of consequences for both clients and HCPs. Significantly, whether GAC is provided at all, and how, is often discretionary, and dependent on individual interest, effort and networks. As willing professionals may refer to any of various international and sector-specific guides, when it is provided, care is highly varied. This is compounded by the interdisciplinary and sometimes highly specialized nature of GAC, where one client—depending on the extent of services they require to help fulfil their gender affirmation—may rely on the services of a mental health professional, endocrinologist, pharmacist, and surgeon. Resultantly, without a central process to inform all HCPs, clients’ access to care can be exceedingly heterogeneous. | other | 31.28 |
Yet, the consequences of withholding GAC from transgender individuals who wish to access it are severe: a study in Ontario, Canada found that the status of individuals’ medical gender affirmation had a significant positive correlation with suicidality, where participants who were planning a medical gender affirmation had either seriously considered (55%) or had attempted (27%) suicide . This is in contrast to those who had seriously considered (23%) and who had attempted (4%) suicide amongst participants who had either completed their gender affirmation or were not planning to medically affirm their gender at all . Likewise, informed by evidence put forth by WPATH in its SOC-7 and an expansive body of research (such as [38–41]), the American Psychiatry Association has recognized that the “lack of access to care adversely impacts on the mental health of transgender and gender variant people, and both hormonal and surgical treatments have been shown to be efficacious in these individuals” . In light of this evidence, and in addition to its constitutional provisions on the right to access to healthcare, South Africa’s health system also has an ethical obligation to increase access to GAC. | other | 29.03 |
Within the community of GAC-providing health professionals, the knowledge of what colleagues were willing to provide and believed to be appropriate gender affirming care begins to act as a guideline. That is, when one HCP knows how another HCP approaches GAC, they begin to shape their service provision accordingly. For example, while some endocrinology guidelines suggest six months of gender affirming hormone therapy prior to certain surgeries, one endocrinologist noted their awareness of other professionals’ flexibility in this regard. Thus, in the absence of national guidelines, the various actors that come together to provide GAC create a self-regulating system of sorts. While this can have positive outcomes for some clients where health professionals seek to facilitate access to GAC, this can equally mean that others are constrained in their efforts to access care. For example, evidence from healthcare service provision to other stigmatized groups in South Africa, including sexual minorities and teenagers seeking sexual and reproductive health services, show that HCP discretion significantly impedes access to services . | other | 28.98 |
Further, where access to services becomes dependent on certain HCPs, it cannot be considered sustainable—those few HCPs who are currently providing GAC won’t always be available, as they retire, experience occupational fatigue, or incur other obstacles to their continued practice. Currently, given the absence of GAC from health education curricula [37, 43], there is nothing in place to ensure their replacement. Moreover, there is little long-term viability to accessing the necessary services when these are only affordable due the pro bono GAC that some HCPs elect to provide. And, because trans people are even more stigmatised and information is comparatively less readily available , this means that personal discretion plays an even more central role in in the provision of GAC, with the effect that transgender individuals experience routine discrimination and exclusion in South Africa’s public health facilities [23, 24]. | other | 29.55 |
All of which is predicated, however, on the assumption that HCPs are comfortable engaging in GAC. While we have been able to speak with a variety of knowledgeable and engaged providers of GAC, these individuals are in the minority in their respective professions. Many of the interviewees discussed a variety of scenarios where access to care for transgender clients was impeded by individuals in the health sector who had no or limited awareness or training regarding gender affirming care. This is not surprising, given that transgender health concerns in general, and GAC in particular, are rarely, if ever, covered in health professions education in South Africa [37, 43]. However, HCP education is a crucial step towards increasing access to GAC, as education around sexual and gender minority health increases HCPs’ willingness to provide services [23, 45], and also equips health professionals with the tools to become advocates for their patients. Once again, national policy guidelines are key, as without them it is unlikely that many HCPs will access education on trans health and GAC. These should take into account the suggested reclassification of ‘gender incongruence’ in the ICD-11 , as well as contemporary debates around depathologisation and access to care | study | 30.08 |
We believe that our findings are not unique to South Africa. Research, even if scarce, shows that the obstacles for GAC that our findings highlight are global: health professions education does not adequately cover GAC [37, 43, 46]; national health policies often do not acknowledge the existence of transgender individuals beyond their needs for HIV prevention and treatment ; and with the exception of some Canadian provinces and some insurances in the US, health insurance schemes do not cover GAC-related procedures . Studying the ways in which HCPs provide GAC in the absence of formal policies, guidelines, or health professions education is therefore relevant to all contexts where GAC is marginalised in health policy and planning. | other | 30.66 |
We believe that nationally approved guidelines, acknowledging the necessity of GAC, are an important step toward broadening and improving access to GAC in South Africa. It would provide a state-sanctioned mandate for the provision of GAC and clear clinical guidelines to standardize care, homogenizing and expanding the provision of GAC in the public sector beyond existing urban hubs. Additionally, it would provide an impetus for financial coverage for concomitant GAC within the private sector, improving access for those able to procure private medical insurance. | other | 30.47 |
In this new digital era, where location-based services are essential in many applications and the number of devices with Internet connectivity has increased dramatically, research on the topic of indoor positioning is flourishing. In contrast to outdoor positioning and navigation, where Global Navigation Satellite System (GNSS) is a mature technology and might be considered a de facto standard, indoor positioning is still in its infancy. The main reason is that GNSS is widespread and is able to provide robust solutions for many applications in a wide range of performance/price ratios, while indoor positioning solutions are few, each specialised for a given scenario and generally expensive. | other | 27.25 |
Many different technologies are being considered for indoor positioning, including Inertial Sensors , Radio Frequency using RFID, BLE, WiFi or UWB ranging , Ultrasound , Computer Vision , and Light . This diversity in technologies has a significant consequence: there is no generally accepted solution for indoor positioning and navigation. In other words, there is not an indoor equivalent to the GNSS-based portable navigation devices commonly used outdoors. Nevertheless, there are already many companies offering indoor positioning solutions and the forecasts estimate an indoor positioning market of more than USD 20 billion by 2021 . | other | 29.39 |
Although the expected commercial impact of indoor location solutions is very high, there is still a lack of standardization in the design, application areas, or evaluation metrics for a fair comparison and assessment of the most promising proposals, both in the literature and at the commercial level. As an example, most of the existing works in the literature show the results of the experiments that were carried out in the authors’ own research facilities, typically department corridors and small research laboratories, which were selected as representative real-world scenarios. Those facilities are not commonly open to other researchers to reproduce their experiments in them. In general, comparing existing working systems is not a trivial task if they have used different technologies, have been evaluated in different scenarios or their performance metrics are not equivalent. | other | 30.2 |
Despite the evaluation heterogeneity found in the literature, some surveys can be found which try to give a consistent overview in this field . However, most of them show the results as provided by developers, which often do not correspond to a common metric, meaning that an exhaustive comparison cannot be performed. | other | 28.66 |
The research community feels the need for accepted methods of measuring the performance of the proposed indoor positioning systems, devices and technologies . In fact, one can observe rising interest in providing open-access datasets, proposing benchmarking tools and, even, organizing competitions in order to facilitate comparative studies. | other | 31.44 |
The EvAAL framework has already been shown to be a consistent way for measuring the performance of personal navigation-oriented positioning systems in indoor environments with known characteristics and, therefore, for rigorously comparing indoor localization systems . These studies back its ambition of becoming a standard in this field. | other | 33.16 |
Although this framework originally aimed at establishing benchmarks and evaluation metrics for comparing Active and Assisted Living solutions, positioning accuracy was an important evaluation criterion. The 2016 Indoor Positioning and Indoor Navigation (IPIN) conference integrated the EvAAL community and framework in the IPIN Competition to compare indoor positioning systems due to their solid experience in organizing competitions. Specifically, three different scenarios were considered in the IPIN 2016 Competition: positioning of people in real-time, positioning of people off-line and robotic positioning. Despite the fact that the robotic positioning might not be related to AAL, it is of interest for the indoor positioning navigation community and it was integrated into the IPIN competition. Therefore, the competition considered different dimensions in the scenario space: pedestrian versus robotic navigation; on-site versus off-site evaluation; usage of smartphones versus customised sensing platforms; infrastructure-based versus infrastructure-free environments. The IPIN community used a well-known evaluation framework in its competition, whereas the EvAAL community had the opportunity to apply their framework in a challenging competition and in other contexts (e.g, the robotic track). | other | 31.73 |
The main objective of this paper is twofold: on the one hand, this paper studies the diversity and heterogeneity of location solutions (different environments, sizes, sensor data and databases, used hardware, etc.) and a review of the different approaches used to evaluate them as proposed by the authors, or by international competitions. On the other hand, this paper aims at demonstrating how the EvAAL benchmarking framework can be used for multiple scenarios and data streams under the large repertory of evaluation tracks at the 2016 IPIN Competition (variety in person/robots, smartphone/custom hardware, multi-storey building, on-line/off-line processing, etc.). We want to demonstrate that a unique evaluation framework, such as EvAAL, can be fitted to many different indoor location use cases. | other | 30.77 |
The remainder of this paper is organized as follows. Section 2 introduces the evaluation heterogeneity found in the literature and the related work aimed at fair benchmarking and comparisons. Section 3 generally describes how the EvAAL criteria were applied to the IPIN 2016 Competition. Section 4 describes the features of the individual competition tracks and shows the results obtained by competitors. Section 5 introduces some lessons learned from the IPIN 2016 competition and the consequent conclusions. | other | 29.3 |
This section is devoted to showing the different problems that arise when trying to compare different results from a large diversity of scenarios, different sensor data streams and non-standardized evaluation methods. The next subsections will present the diversity or heterogeneity problem when presenting research location solutions, public databases for testing new algorithms, and the problem related to the diversity of competition conditions and set-ups. | other | 30.16 |
Some criteria to evaluate Indoor Positioning Systems (IPSs) for personal networks were proposed in , for example, privacy, cost, performance and robustness. The authors of observed that the two main performance parameters are the accuracy and the precision, where the former is related to the geometric error and the latter was defined as the success of the position estimations with respect to a predefined accuracy (e.g., the space-based location or the percentage of error in positioning below a threshold). In general, the evaluation metrics depend on the author. | other | 29.1 |
RADAR , the first Wi-Fi based indoor positioning system, used the quartile values of the error in positioning (defined as the Euclidean distance between the actual and estimated positions) in order to compare the proposed method to other naïve solutions in a basic analysis. The experimental testbed was located on the second floor of a 3-storey building, with an area of 43.5 m by 22.5 m, more than 50 rooms and three Wi-Fi Access Points (APs). HORUS , another well-known IPS developed in 2003 by Youseff et al., provided the median error in positioning and it was tested in two environments: an area of 68 m by 26 m with 172 locations and 21 APs, and an area of 36 m by 12 m with 110 locations and six APs. The two testing approaches are quite different because the density of APs was 0.003 APs/m2 in RADAR, whereas HORUS provided two scenarios with densities of 0.012 APs/m2 and 0.14 APs/m2 respectively, i.e., the scenarios where HORUS was tested had approximately four times the AP density of the evaluation testbed of RADAR. Therefore, comparing the results provided in the original references might not be fair. | other | 34.12 |
In order to provide a fair comparison of RADAR (the deterministic method developed by Microsoft) and HORUS (the probabilistic technique developed by the University of Maryland), they were both implemented and evaluated using the same testbeds . According to the data provided in , the density of access points in the testbed used in , a university department corridor, was higher than in . This change might be due to an improvement of the Wi-Fi network. Despite this minor change in one of the evaluation testbeds, the work done in showed that a fair comparison requires the use of the same testbed, or testbeds, rather than reusing the results provided in the literature. | other | 34.53 |
Apart from the diversity of environments, there is also diversity in the hardware elements used for localization and in the metrics to evaluate an IPS. A brief resume of a few works presented in the 2016 Indoor Positioning and Indoor Navigation conference (October 2016, Alcalá de Henares, Spain) is shown in Table 1, where the conference session, base technology, evaluation set-up and evaluation metric are also shown. | review | 27.3 |
According to Table 1, each research work uses specific hardware even when they use the same base technology for positioning. A Huawei Mate smartphone was used in , six different devices were used in , and a simulation was carried out in . Using multiple devices in the set-up might be more challenging than using a single device or simulated data. Therefore, comparing the results of those works might not be easy at all. | other | 33.3 |
Although the vast majority of papers agree that the error in positioning is defined as the shortest distance between the estimate and current position, consensus in using a particular metric to evaluate the IPS has not been reached yet, as shown in Table 1. This table shows that many metrics have been used to evaluate a IPS, but there is no clear winner metric used in all papers. This conclusion is in line with the study where 195 papers of the first edition of the Indoor Positioning and Indoor Navigation (IPIN 2010) Conference were analysed . | other | 29.9 |
For instance, the results of are based on the final drift (the geometric error at the end of the track) and visual analysis of the track. The work presented in reports the average error per axis and angle, and the positioning error is not computed. The work presented in shows some plots with measured probability and cumulative distribution instead of providing simple evaluation numbers. Estimated versus real trajectory are the results shown in . | other | 30.25 |
Moreover, the scenarios found in the literature are very diverse: from a small 1 m2 area or a simulated 3 m by 3 m scenario to a large university hospital or a set of tracks with about 60 accumulated kilometres. Even for the same topic and base technology, the evaluation metrics and the environment depend on the author. | other | 29.17 |
Out of Sight is a toolkit for tracking occluded human joint positions based on Kinect cameras. Some in-room test were run for evaluating different contexts (stationary, stepping, walking, presence of obstacle and oclusion). The mean error on the three axes (x, y, z) and the mean positioning error were provided. | other | 34.34 |
A Kalman filtering-based localization and tracking for the IoT paradigm was proposed in , where a simulation in a 1000 m by 1000 m sensor field was performed. The results and comparisons were based on the trajectory plots, and the location and velocity errors in the x and y axes over time (track). | other | 32.9 |
This diversity of evaluation set-ups is not only attached to the IPIN conference; it is a generalized problem in the field of indoor positioning:GETA Sandals have been evaluated in a few scenarios (straight track, rectangular loop, and climbing stairs) and the results have been provided as the average drift (positioning error at the end of the track) of many users and by plotting the walking distance vs. the positioning error.MoteTrack evaluation was focused on understanding how it performed under various parameter conditions and the authors carried out different tests. The results were provided as CDFs and plots based on the error distance and a few percentile values.Out of Sight is a toolkit for tracking occluded human joint positions based on Kinect cameras. Some in-room test were run for evaluating different contexts (stationary, stepping, walking, presence of obstacle and oclusion). The mean error on the three axes (x, y, z) and the mean positioning error were provided.A Kalman filtering-based localization and tracking for the IoT paradigm was proposed in , where a simulation in a 1000 m by 1000 m sensor field was performed. The results and comparisons were based on the trajectory plots, and the location and velocity errors in the x and y axes over time (track).A smartphone-based tracking system using Hidden Markov Model pattern recognition was developed in . They carried out their experiments in the New Library of Wuhan University using three different device models with over 50 subjects walking over an aggregate distance of over 40 km. The CDF and mean accuracy were used to report the results. | other | 29.77 |
GETA Sandals have been evaluated in a few scenarios (straight track, rectangular loop, and climbing stairs) and the results have been provided as the average drift (positioning error at the end of the track) of many users and by plotting the walking distance vs. the positioning error. | other | 36.6 |
Many surveys show comparative studies about indoor positioning technologies (e.g., ). The reported accuracy is often generic and sometimes is reported just as, e.g., ’room level accuracy’, or 50% within around 2.5 m or as a simple range (2 to 5 m). In general, surveys discuss the general accuracy of positioning technologies and the future trends rather than trying to rank the analysed solutions. | review | 27.6 |
Liu et al. introduced a survey about indoor positioning algorithms (angulation, lateration, scene analysis and proximity), performance metrics (accuracy, precision, complexity, robustness, scalability and cost) and technology solutions (GPS, RFID, WLAN, Bluetooth, UWB, cellular). | other | 31.11 |
A smartphone-based tracking system using Hidden Markov Model pattern recognition was developed in . They carried out their experiments in the New Library of Wuhan University using three different device models with over 50 subjects walking over an aggregate distance of over 40 km. The CDF and mean accuracy were used to report the results. | other | 35.78 |
The evaluation heterogeneity present in the literature does not mean that the results shown are not valid; it just makes it more difficult to directly compare the proposed methods with just the information and experiments provided by the authors. The rest of this section will show some initiatives and frameworks for fair evaluation of IPS. | other | 30.11 |
Gu et al. gave a comprehensive survey of some commercial solutions and research prototypes, which were compared by using many different features: security and privacy, cost, performance, robustness, complexity, user preferences. Trade-offs among the features and the viewpoint of a user in a personal network were also outlined. | other | 33.53 |
He and Chan introduced a survey of Wi-Fi fingerprinting methods, which described advances in two major areas: advanced localization techniques and efficient system deployment. Different methods exploiting spatial and temporal signal patterns were compared according to indoor site availability, additional information for localization estimation, limitations, and reported mean accuracy. Similarly, recent approaches for motion-assisted Wi-Fi localization and typical schemes of collaborative localization were also compared. | other | 31.52 |
Hassan et al. performed a survey on indoor positioning using visible LED light. In a first comparison, they provided general information about other positioning technologies (Wi-Fi, BLE, GSM, among others): accuracy, robustness, complexity, cost and infrastructure reusability. Even for the surveyed LED solutions, they provided an accuracy range in most of the cases. | other | 30.64 |
In general, surveys have one common feature, that is, the IPS are compared using the performance as provided by the authors themselves. Moreover, the metric used to compare accuracy might differ, e.g., the performance of RADAR is 3–5 m in , 4 m with 50% probability in , and 3–5 m (2.5 with 50% of probability and 5.9 with 90% of probability) in . Results shown in survey works are useful to show the differences between different indoor location technologies, but comparing methods relying on the same technology might not be easy in some cases. | other | 32.9 |
A Database related to a Wi-Fi-based positioning system on the second floor of an office building on the campus of the University of Mannheim, released in . The operation area was about 57 m × 32 m but only 221 square meters were covered. The test environment had twelve APs. Seven of them were administered by the university technicians, whereas the others were installed in the nearby buildings and offices. Thirteen additional access points were added for localization purposes. | other | 33.1 |
One data set to aid the development and evaluation of indoor location in complex indoor environments using round-trip time-of-flight (RToF) and magnetometer measurements . It contains RToF and magnetometer measures taken in the 26 m × 24 m New Wing Yuan supermarket in Sunnyvale, CA, USA. The data was collected during working hours over a period of 15 days. | other | 33.53 |
In the machine learning and pattern recognition community, it is common to test new algorithms over a set of well-known datasets. Researchers tend to upload their collected datasets, and the employed software in some cases, to a database platform or to their own website. | other | 30.89 |
The Community Resource for Archiving Wireless Data At Dartmouth (CRAWDAD) platform hosts data sets about wireless data including data sets for indoor positioning and mobility. Some illustrative data sets follow:A Database related to a Wi-Fi-based positioning system on the second floor of an office building on the campus of the University of Mannheim, released in . The operation area was about 57 m × 32 m but only 221 square meters were covered. The test environment had twelve APs. Seven of them were administered by the university technicians, whereas the others were installed in the nearby buildings and offices. Thirteen additional access points were added for localization purposes.Mobility traces at five different sites (NCSU university campus, KAIST university campus, New York City, Disney World -Orlando- and North Carolina state fair) .Two trace files (one for Wi-Fi and one for Bluetooth) collected by the University of Illinois Movement (UIM) framework using Google Android phones .One data set to aid the development and evaluation of indoor location in complex indoor environments using round-trip time-of-flight (RToF) and magnetometer measurements . It contains RToF and magnetometer measures taken in the 26 m × 24 m New Wing Yuan supermarket in Sunnyvale, CA, USA. The data was collected during working hours over a period of 15 days.The database donated in contains the RSS (Radio Signal Strength) data collected with a mobile robot in two environments: indoor (KTH) and outdoor (Dortmund). The RSS metric was used to collect the RSS data in terms of dBm. The mobile robot location was recorded using odometry (dead reckoning). | other | 31.73 |
The database donated in contains the RSS (Radio Signal Strength) data collected with a mobile robot in two environments: indoor (KTH) and outdoor (Dortmund). The RSS metric was used to collect the RSS data in terms of dBm. The mobile robot location was recorded using odometry (dead reckoning). | other | 35.44 |
The server at the University of California Irvine is a well-known repository of databases for machine learning and pattern recognition problems, the UCI Machine Learning Repository , where a few datasets for indoor positioning can be found:The UJIIndoorLoc database is a Wi-Fi fingerprinting database collected at three buildings of the Jaume I university (UJI) campus for indoor navigation purposes. It was collected by means of more than 20 devices and 20 people and was used for the off-site track of the 2015 EvAAL-ETRI Competition .The UJIIndoorLoc-Mag database was a magnetic field-based database collected at one laboratory at the Jaume I university (UJI) campus for indoor navigation purposes .The Indoor User Movement Prediction from RSS Data Set represents a real-life benchmark in the area of Active and Assisted Living applications. The database introduces a binary classification task, which consists in predicting the pattern of user movements in real-world office environments from time-series generated by a Wireless Sensor Network (WSN) .The Geo-Magnetic field and WLAN dataset for indoor localisation from wristband and smartphone data set contains Wi-Fi and magnetic filed fingerprints, together with inertial sensor data during two campaigns performed in the same environment . | other | 31.38 |
The UJIIndoorLoc database is a Wi-Fi fingerprinting database collected at three buildings of the Jaume I university (UJI) campus for indoor navigation purposes. It was collected by means of more than 20 devices and 20 people and was used for the off-site track of the 2015 EvAAL-ETRI Competition . | other | 33.38 |
The Indoor User Movement Prediction from RSS Data Set represents a real-life benchmark in the area of Active and Assisted Living applications. The database introduces a binary classification task, which consists in predicting the pattern of user movements in real-world office environments from time-series generated by a Wireless Sensor Network (WSN) . | other | 33.12 |
On the other hand, a data set has a few disadvantages: it commonly covers one or a few positioning technologies; external researchers (users) do not have control over data collection; some extended information (maps, obstacles, among other useful information from the environment) might not be present; and the data format and the training/validation/test separation depends on the creator/donor. | other | 27.9 |
In this subsection, we present the problem related to the heterogeneity or diversity of different competition tracks proposed in the last few years for indoor positioning. We mainly review the Microsoft and IPIN competition, describing the diversity of scenarios, data streams and hardware used, and the need for a standardization framework. | other | 29.86 |
The Signal processing for the wireless positioning group (Tampere University of Technology, Finland), provides an open repository of source software and measurement data on their website (http://www.cs.tut.fi/tlt/pos/Software.htm). Indoor WLAN measurement data in two four-floor buildings for indoor positioning studies are provided. The data contains the collected RSS values, the Access Points ID (mapped to integer indices) and the coordinates; both the training data and several tracks for the estimation part are provided as indicated in . Other datasets with GSM, UMTS and GNSS data are also provided. | other | 32.2 |
PerfLoc is also running a competition about developing the best possible indoor localization and tracking applications for Android smartphones. The competition participants are required to use a huge database collected by The National Institute of Standard and Technology (NIST) by means of four different Android devices. | other | 32.56 |
The Microsoft Indoor Localization Competition at the International Conference on Information Processing in Sensor Networks (IPSN) is one of the most popular indoor localization competitions. It held its first and second editions in Berlin and Seattle, in 2014 and 2015. The third edition was held in Vienna in 2016, where the evaluation site was a hall of dimensions 30 by 15 m (450 m2). | other | 31.69 |
In the 2016 edition , two different tracks were held, called “2D” and “3D”, where position data (XY or XYZ, respectively) had to be generated by competitors at each of 15 different evaluation points. A total of 49 teams showed interest in the competition, of which 35 were registered and 28 were evaluated, 12 in the 2D track and 16 in the 3D one. | other | 30.03 |
In the 2D track, competitors were not allowed to install any devices in the testing hall. Three Wi-Fi APs where available for use. Competitors could use Wi-Fi-based location systems or self-contained solutions using an inertial measurement unit or a magnetometer on the competitor’s body (e.g., a smartphone alone or with some external IMUs). | other | 32.16 |
All teams had a whole day prior to the competition for calibration. During the competition, some teams were excluded from the awards since some of them (those based on dead-reckoning such as IMU-based approaches) needed an initial known position. Other teams were also excluded because their system was not real-time, as they needed post-processing. | other | 32.16 |
The entry barrier to IPIN is high, mostly because high stability and robustness are required of the competing systems, which must be able to work reliably for a longer time in tougher conditions. Overall, the EvAAL criteria impose rigorous and realistic test conditions which are appropriate to deal with state-of-the art working systems. | other | 30.55 |
The IPIN competition refers to a typical agent indoor navigation scenario, according to which an agent, roaming within an unknown environment, needs to know its current position. The agent moving within the environment can be either a person or a robot. In both cases, we assume that the agent carries an indoor positioning device, e.g., a smart phone or a custom device. For human agents, we refer to a wide indoor environment, such as a mall, a hospital or a big office in which navigation is not a simple task. For robotic agents, we refer to a high precision system. Due to heterogeneity of technologies and target applications, the IPIN 2016 competition includes three different scenarios:positioning of people in real time;positioning of people off-line;robotic positioning in real time. | other | 28.38 |
The main characteristics of the Microsoft competition were kept the same during its first three editions: lack of predefined track formats at the time of subscriptions as formats were precisely defined later, based on competitor’s interest declared during the subscription; the testing environment was about the size of a big hall, with few or no walls, on a single floor; measurements were done in sequence at a series (about a dozen) of fixed points, where the competitors bring their equipment in turn, let the system stabilise, and one organiser reads coordinate estimates on the screen of the competitor’s PC; the metric was the mean of errors, defined as the Euclidean distance between the reference points and the competitor’s estimates. | other | 35.4 |
The International Conference on Indoor Positioning and Indoor Navigation (IPIN) is the oldest conference in this field. IPIN decided to launch its first competition in 2014 and it adopted the EvAAL framework. IPIN formally adopted the EvAAL framework in 2015 and 2016. The main features of the past IPIN Competitions were:IPIN competitions take measurements while naturally moving through a predefined trajectory unknown to competitors, instead of evaluating the competitors by standing still at evaluation points;IPIN competitions are generally done in big challenging multi-floor environments, possibly on multiple buildings, with significant path lengths and duration, instead of evaluating competitors in small single-floor environments;IPIN competitions generally require the competitors to interface with an independent real-time measurement application and test on an independent actor;The final score metrics is the third quartile of the positioning error in IPIN, which makes the accuracy results less prone to the influence of outliers and more in line with demanded accuracy for commercial systems. | other | 30.33 |
The goal of this scenario is to measure the performance of working systems with people moving in realistic conditions. The setting was the Polytechnic School of the University of Alcalá (EPS-UAH), which is divided into four sectors (red, blue, green and yellow) of which two contiguous ones (the red and blue sectors) were used in the competitions for positioning people in real time. Figure 1 shows the floor plans of this scenario, where the usable areas are marked with an orange pattern. | other | 36.88 |
Like the previous one, the purpose of this scenario is to measure the performance of working systems with people moving in realistic conditions. The difference is that the competition is performed off-line. The calibration and evaluation data to set up and check the indoor positioning systems are provided in advance to competitors by organizers. Therefore, all competitors have the same data and information to create and evaluate their systems. | other | 33.1 |
Since evaluation time is not a tough restriction, the setting was composed by many different tracks which were pre-recorded in four heterogeneous buildings: the green sector of the Polytechnic School of the University of Alcalá (EPS-UAH), the CAR building (CSIC Arganda, Madrid), and UB & TI buildings (Universitat Jaume I, Castellon). The setting at EPS-UAH building was different from Tracks 1 and 2, in order to avoid interfering with them. Figure 2 shows the base floor plans (only ground floor for simplification) of the four buildings used for this scenario. The databases of the off-site track are available online at http://indoorloc.uji.es/ipin2016track3/. | other | 33.25 |
The purpose of the last scenario is to continuously log the trajectory followed by an industrial robot. Robots are generally programmed to accomplish specific tasks, e.g., delivering postal mails among offices or delivering medical supplies to laboratories in a hospital. In the competition scenario, the industrial robot followed a closed path, including circular and straight sections alternately arranged. The path was unknown to participants in advance, as shown in Figure 3. Competitors provided both their own custom hardware, that had to be deployed and calibrated in a constrained time, and their algorithm approaches for tracking the robot trajectory. | other | 33.97 |
Robotics competitions have a large tradition, such as the Robot World Cup (Robocup) , the Urban Search and Rescue (USAR) or the AUVSI SUAS Competition for Unmanned Aerial Systems (UAS) . These competitions play an important role in the development of artificial intelligence (AI) and robotics, but none of them is focused on indoor robot positioning. | other | 28.44 |
In this section, we first review some of the most prominent benchmarking metrics proposed to date (EVARILOS and EvAAL), and then we show how the EvAAL framework has been integrated into the IPIN competition to evaluate diverse positioning systems under multiple scenarios (multi-floor, multi-building, multi-sensor, multi user persons/robots, etc.) in order to create a de facto standard for evaluation. | other | 30.64 |
The EVARILOS benchmarking platform , with a competition held in 2014, had the objective of automating the evaluation process, although the final error metrics were calculated with a manual process from the ground truth. The EVARILOS benchmarking proposal defines a taxonomy of metrics (deployment, functionality and performance related) integrated in a weighted sum to finally generate the system scores. Although it targets radio frequency-based systems, some of their ideas could be generalized to broader scenarios. | other | 28.86 |
The first public international localization competitions (2011–2013, Valencia and Madrid) were part of the EvAAL initiative . The EvAAL framework is characterised by several core criteria in order to evaluate the teams:Natural movement of an actor: the agent testing a localization system is an actor walking with a regular pace along the path. The actor can rest in a few points and walk again until the end of the path.Realistic environment: the path the actor walks is in a realistic setting: the first EvAAL competitions were done in living labs.Realistic measurement resolution: the minimum time and space error considered are relative to people’s movement. The space resolution for a person is defined by the diameter of the body projection on the ground, which we set to 50 cm. The time resolution is defined by the time a person takes to walk a distance equal to the space resolution. In an indoor environment, considering a maximum speed of 1 m/s, the time resolution is 0.5 s. These numbers are used to define the accuracy of measurements. When the actor walks, measurements are taken when he/she steps over a set of predefined points. The actor puts his/her feet on marks made on the floor when a bell chimes, once per second. As long as he/she does not make time and space errors greater than the measurement resolutions, which is easy for a trained agent, the test is considered adequate.Third quartile of point Euclidean error: the accuracy score is based on the third quartile of the error, which is defined as the 2-D Euclidean distance between the measurement points and the estimated points. More discussion on this is in the next section. Using the third quartile of point errors, the system under measure provides an error below the declared one in three cases out of four, which is in line with the perceived usefulness of the experimental IPS. Reference argues that using linear metrics such as the mean may lead to strange and unwanted behaviour if they are not properly checked because of, for instance, the presence of outliers. A clear example of this behaviour can be seen in , where a competitor of the off-site track provided severe errors in very few cases and its average error and root mean squared error were negatively highly affected, however the competition metric, the third quartile, showed that the solution proposed was not as bad as the averaged error had shown. More detailed discussion is found in . | other | 29.39 |
The core criteria are the distinguishing features of the EvAAL framework. The first EvAAL competitions additionally adopted the following extended criteria:Secret path: the final path is disclosed immediately before the test starts, and only to the competitor whose system is under test. This prevents competitors from designing systems exploiting specific features of the path.Independent actor: the actor is an agent not trained to use the localization system.Independent logging system: the competitor system estimates the position at a rate of twice per second, and sends the estimates on a radio network provided by the EvAAL committee. This prevents any malicious actions from the competitors. The source code of the logging system is publicly available.Identical path and timing: the actor walks along the same identical path with the same identical timing for all competitors, within time and space errors within the above defined resolutions. | other | 27.95 |
Independent logging system: the competitor system estimates the position at a rate of twice per second, and sends the estimates on a radio network provided by the EvAAL committee. This prevents any malicious actions from the competitors. The source code of the logging system is publicly available. | other | 27.27 |
Realistic measurement resolution: the minimum time and space error considered are relative to people’s movement. The space resolution for a person is defined by the diameter of the body projection on the ground, which we set to 50 cm. The time resolution is defined by the time a person takes to walk a distance equal to the space resolution. In an indoor environment, considering a maximum speed of 1 m/s, the time resolution is 0.5 s. These numbers are used to define the accuracy of measurements. When the actor walks, measurements are taken when he/she steps over a set of predefined points. The actor puts his/her feet on marks made on the floor when a bell chimes, once per second. As long as he/she does not make time and space errors greater than the measurement resolutions, which is easy for a trained agent, the test is considered adequate. | other | 34 |
Third quartile of point Euclidean error: the accuracy score is based on the third quartile of the error, which is defined as the 2-D Euclidean distance between the measurement points and the estimated points. More discussion on this is in the next section. Using the third quartile of point errors, the system under measure provides an error below the declared one in three cases out of four, which is in line with the perceived usefulness of the experimental IPS. Reference argues that using linear metrics such as the mean may lead to strange and unwanted behaviour if they are not properly checked because of, for instance, the presence of outliers. A clear example of this behaviour can be seen in , where a competitor of the off-site track provided severe errors in very few cases and its average error and root mean squared error were negatively highly affected, however the competition metric, the third quartile, showed that the solution proposed was not as bad as the averaged error had shown. More detailed discussion is found in . | other | 30.22 |
During IPIN 2016, four competition tracks were run in parallel. The competitors of each track were evaluated according to the third quartile of the positioning error. This error is measured based on xy coordinates (longitude and latitude). Also, a penaltyP=15 m is added for each floor error. For example, if the xy error is 4 m and the estimated floor is 2 while it should be 0, the computed error for that estimate will be 4+2P=34 m. This is in line with real users’ point of view, because some movements in the environment are restricted by physics and the architectural elements . | other | 34.16 |
The two first core EvAAL criteria are followed closely: in Tracks 1–3 the actor moves naturally in a realistic and complex environment spanning several floors of one (for Tracks 1 and 2) or few (Track 3) big buildings; in Track 4, the robot moves at the best of its capabilities in a complex single-floor track. | other | 30.22 |
The same holds for the third core criterion: the space-time error resolution for Tracks 1–3, where the agent is a person, are 0.5 m and 0.5 s, while space-time resolution for Track 4, where the agent is a robot, are ≈1 mm and 0.1 s. In Track 4, only the adherence to the trajectory is considered given the overwhelming importance of space accuracy with respect to time accuracy as far as robots are concerned. | other | 30.83 |
Subsets and Splits