Assessment of an intelligent robotic rollator implementing navigation assistance in frail seniors

Abstract

BACKGROUND:

Assistive technologies are playing vital role in frail seniors.

OBJECTIVE:

The paper presents the methodology and findings from the assessment of an active “intelligent” robotic rollator incorporating navigation assistance in frail seniors.

METHODS:

Thirty seniors, having moderate motor impairment were grouped into two categories according to their cognitive status. A complicated trail was defined to test their performance using a robotic rollator. The main descriptive statistics were given, while inference was performed via the corresponding parametric $t$ tests and the non-parametric Mann-Whitney U or Kruskal-Wallis tests. A principal component analysis was utilized to extract the variable classes that explain the overall behavior of the sample.

RESULTS:

Navigation assistance was proved helpful to the cognitive impaired group. The results between the two groups were comparable in terms of completion time and success rate. The provided directional audio cues led to smoother walking paths and better orientation. There was no correlation between the MMSE values and the total time of the walk ( $p=$ 0.696 $>$ 0.05 for Pearson’s correlation between MMSE and total walk time).

CONCLUSIONS:

The integration of navigation assistance in mobility assistance devices can help the elderly to effectively orient themselves in challenging and unknown environments.

Keywords

Robot rollator assistive technology assessment evaluation rehabilitation human-robot interaction mobility subjective evaluation navigation

1. Introduction

Life expectancy at birth has increased by almost 20 years over the last half century [1]. According to the latest statistics, 962 million people are over 60 years old [2]. By the year 2050, it is estimated that more than 2 billion people will be aged 60 and over, and almost 400 million of them will be aged 80 $+$ [3]. The ageing of population is one of the major problems faced by the society in many aspects (healthcare expenditure, social care, etc.), since the prevalence of disability increases dramatically by age.

Given the growth of the older population, health systems around the world should prepare to address the various health concerns of older persons. The leading causes of disability globally among aged 60 years or over, according to WHO estimates for 2012 are: unipolar depressive disorders, hearing loss, back and neck pain, Alzheimer’s disease and other dementias, and osteoarthritis. Falls is also a major cause of disability among older persons globally [4].

The loss of mobility, which is a very common occurrence, can be devastating to the elderly and seniors. The Census Bureau of the USA reports that mobility problems are the most common disability among the elderly [5]. Due to loss of mobility the falls and injuries are increasing and can be life-threatening. Mobility loss is also highly interconnected with depression. Mobility aids have been used to maintain independent mobility. Several studies have found that use of a mobility aid is a prospective predictor of increased risk of falling in older adults or is associated with falls and related injuries. Other studies have argued that use of the devices may actually increase risk of falling by causing tripping or by disrupting balance control through other mechanisms [6, 7, 8, 9]. According to the same study, although many studies have reported that mobility aids and/or environmental obstacles are associated with falls, the possible link between these two risk factors has not been studied. Furthermore, it appears that frame-walker-related injury can occur as a result of contact and/or “catching” of the mobility aid with environmental objects, such as carpets and doorframes that would not normally be considered obstacles [10].

Nowadays, there are a variety of mobility assistive devices in the market available. Gait assistive device users are involved at a significant percentage of fall-related injuries that end-up for hospital treatment [11]. 60% of all fall injuries concerning gait assistive device users occurred at home and 16.3% of the injuries associated with walkers (walking frames or rollators) occurred at nursing homes. Approximately 8.3% of injuries involving walkers and 12% involving canes happened in public places. Approximately 1.6% of walker injuries involved falls on stairs or steps [11].

Even though mobility aids can help in several issues, it is possible that the act of lifting the mobility aid (referring to walking frames or other similar devices) can lead to instability in the same way that lifting the foot can cause the centre of mass (COM) to fall toward the unsupported side during unassisted gait. By suddenly reducing the base of support (BOS), lifting the device could create a state of imbalance in which the COM lies outside the BOS limits. Moreover, the tendency of the walking frame to tip increases with the magnitude of the applied horizontal force and with the height of the walker [10]. When it comes to gait speed, the use of any walking aid, in particular walkers, results in a slower gait speed and requires considerably more energy and cardiovascular fitness than walking unassisted [12]. Also, as mentioned by [13], with a fixed frame walker, physiological cost index (PCI) was higher in a weakened elderly individual due to heart rate increase and lower walking speed.

There is little evidence to suggest whether the use of the walking aid alone leads to this risk or if it is related to the decreased level of physical function, increased frailty, and poorer general health that users of walking aids may have. However, inappropriate walking aid prescription, inadequate training of the user and un-prescribed use of walking aids are likely to exacerbate the problem [14].

The use of gait assistive devices, although can provide precious help for the elderly [13] or for the mobility-impaired, can also have limitations or adverse consequences. Studies show that 30% to 50% of people prescribed with a gait assistive device abandon their device soon after receiving it [15]. Furthermore, surveys indicate that almost half of the reported problems associated with cane or walker use fall under the category of difficult and/or risky to use [10]. 58.3% of knee osteoarthritis patients that possess a walking aid are non-users of the aid. At this case, nonuse is associated with less need, negative outcome, and negative evaluation of the walking aid [16]. Another example indicating the difficulty of finding and using an all-circumstances suitable gait assistive device is multiple sclerosis patients. According to [17] the vast majority of persons with MS who use a mobility aid own more than one type and about half own three or more different types of mobility aids. Persons with MS appear to “mix and match” different devices to suit their specific mobility needs.

Age is also the greatest risk factor for cognitive impairment. Due to the ageing of population, the number of people living with cognitive impairment is expected to jump dramatically. The imminent growth in the number of people living with cognitive impairment is expected to place significantly greater demands on the health care systems. New technologies are urgently needed for patients with any level of cognitive impairment in order to assist them in normal daily activities (i.e. walking and navigating in known or unknown places).

Several robotic intelligent mobility aids have been presented in the literature, largely divided into two categories: robotic wheelchairs and robotic walkers [18]. Robotic walkers constitute a mechanized version of the standard walking frame and are further divided into three main groups [19]: the Standard Walking Frame (aka Zimmer Frame), designed to provide support to a person with a lower limb weakness, the Rollators which is a standard frame attached with wheels used where balance – rather than weight bearing – is the major problem, and the Reciprocal Frames, similar to the Standard Frames except that the frame is hinged on either side allowing the sides of the frame to be moved alternately. They are designed to accommodate a normal walking pattern with opposite arm and leg moving together. All robotic walkers are a robotized variation of the Rollator frame, ultimately identified with mobile robots.

Many robotic walkers have been developed, generally presenting the following functionalities: (i) Physical support; (ii) Sensorial assistance; (iii) Cognitive assistance; (iv) Health monitoring and (v) Advanced human-machine interface [20]. The platforms largely fall into two categories: passive and active devices [21]. While passive mobility aids either steer or brake, but cannot move forward without the human applying forces on them, active devices are equipped with actuators and thus, their motion and interaction behavior can be actively controlled. Some examples of intelligent mobility aids are the iWalker robot [22], the GUIDO system (Haptica Inc., Dublin, Ireland), the Personal Aids for Mobility and Monitoring “PAMM” SmartWalker [23], another system also called SmartWalker [24], the JAIST Active RObotic Walker (JARow) [25], Fraunhofer’s Care-O-bot and Care-O-bot II [26], the LEA robot (Robot Care Systems bv, Den Haag, Zuid-Holland) and other.

The acceptance and use of innovative assistive technologies among people with cognitive impairment and their caregivers is still challenging [27]. The performance of daily activities by device-assisted walkers is partially restrained by a series of factors, having to do with the assistive device limitations or/and the limitations emerged from environmental causes. Specifically oriented to walking-aids, valid and reliable assessment tools to classify and clarify their limitations or users’ satisfaction were not tracked. A recent thorough review revealed the most valid and reliable instruments for the subjective measurement of assistive devices [28].

The current paper presents the assessment of an active “intelligent” rollator in order to evaluate the positive effect of a robotic mechanism incorporating navigation assistance in frail seniors. The system is initially planned to be used in institutional spaces.

2. Materials and methods

2.1 Intelligent mobility assistance robot

The robotic system used in this work, is a collaborative effort of several European partners under the FP7 EU funded project “MOBOT” (Intelligent Active MObility Aid RoBOT integrating Multimodal Sensory Processing, Proactive Autonomy and Adaptive Interaction). Focus of the project was the development of robotic mobility aids that provide intelligent and active mobility assistance to the elderly, by supporting safe autonomous proactive control of the physical user-robot interaction, and enabling multimodal sensory processing and natural human-robot communication [29]. MOBOT (Fig. 1) integrates novel functionalities such as Sit-To-Stand assistance [30], indoor localization [31], navigation assistance [32], user following [33], obstacle detection and avoidance [34] and other, into a context-aware mobility assistance robot. For a more thorough presentation of MOBOT’s functionalities, the reader is referred to [35].

Figure 1.

The MOBOT intelligent mobility assistant robot, used in the experimental sessions.

2.2 Navigation assistance system

MOBOT’s intelligent functionalities have been developed using the Robot Operating System (ROS) [36], currently regarded as the most popular software framework for research and development in robotics [37]. The ROS contains a large collection of open-source drivers, algorithms, tools, and libraries for the development of various robotic tasks. MOBOT’s navigation assistance module deployed in this work, was built utilizing ROS’s navigation stack [38], which is extensively used by research teams around the world for autonomous robot navigation. ROS stacks comprise algorithms and tools with a specific objective, as for example, navigation. MOBOT’s navigation assistance task consists of the following sub-tasks: map building, odometry, and localization. To facilitate these sub-tasks, the rollator was equipped with two high-precision quadrature optical encoders on the two rear driving wheels, a laser range finder (Hokuyo UTM-30LX) on the front of the robot, facing towards the direction of motion, and an inertial measurement unit (IMU, XSensMTi-G-700 GPS/INS) mounted on the chassis of the rollator. Mapping was performed using the openSLAM Gmapping library. This creates maps of unknown environments using scan data from laser range finders. Prior to any experiments, a static map of the indoor environment was created. The position and orientation (i.e. pose) estimation of the robot, was performed using odometric information, relative to a starting position. This was achieved by integrating over time the rotation of the rear wheels, as measured by their respective quadrature encoders; a technique called dead reckoning. In this technique, the current pose is estimated by calculating the relative displacements from a previous pose [39]. Odometric pose estimation is a well-established method in robotics, providing good short-term accuracy. However, the integration of incremental motion over time inevitably leads to the accumulation of errors, due to, for instance, wheel slippage and measurement noise. This accumulation increases proportionally to the travelled distance [40] and may result in very large pose estimation errors. To handle this problem, the estimation of the robot’s angular velocity was calculated from the gyroscopic data of the IMU, and consequently fused to the estimation of its linear velocity, provided by the wheel encoders. This data fusion reduced the error estimation error and provided a more robust solution [41]. The estimation of MOBOT’s position and orientation on the known map, created prior to the experiments, is called “Localization”. This was attained using ROS’s Adaptive Monte Carlo Localization module, which provides probabilistic robot localization in 2D maps using a particle filter [42]. The navigation assistance system provides pre-specified auditory localization information and navigation instructions to the user while they have full motion control over the MOBOT rollator (manual control mode). The module considers the known pre-built map, along with a set of predefined guard points which have associated audio cues. Each guard point consists of two states: the “in” stathe and the “out” state. The transition to the “in” state is made when the rollator enters a circle of specified radius centered at the guard point. Conversely, the transition to the “out” state is made when the rollator exits this area. On entering a guard point, a precise directional audio cue is provided guiding the user safely on a direct route to a preselected destination. While inside the guard area, the audio cue is repeated every 3 seconds, until the user exits the specific guard point. The guard are treated as a directed path. This means that when the rollator exits a guard point, this is discarded and removed from the route. Only subsequent guard points are then considered.

The study was approved by the Bioethics and Deontology Committee of the Technological Educational Institute of Athens. A written informed consent was obtained from each participant. The evaluation study took place in a Greek rehabilitation hospital and examined the assistance provided by the robotic rollator focusing on the navigation functionality.

The evaluation scenarios were tailored to meet the needs of the research using both objective and subjective criteria. To this end, a multi-disciplinary team was established to coordinate the project, constituted by a biomedical engineer, a robotic engineer and a physiatrist. The objective evaluation of the robotic rollator was based on the collected various data from the multiple sensors integrated in the robotic rollator-type device. The subjective assessment of the device was another major issue [28], and was achieved by using three already validated questionnaires [43, 44, 45].

2.3 Design and experimental setup

The objective of the cognitive assistance scenario was to assess the integrated navigation functionality of the MOBOT robotic rollator, related to the assistance of the end user when navigating in a trail. MOBOT provided the audial navigational assistance described in the previous section. The experimental premises were mapped prior to the trials, while the navigational trails, along with the guard points and their associated audio cues, were designated on the map and encoded into the system before initiating the experiments.

According to the research hypothesis, the cognitive assistance functionality would assist subjects from Group B (cognitive impairment group, MMSE $>$ 26) to navigate in unfamiliar spaces and achieve comparable results to Group A (cognitive intact group, MMSE $\leqslant$ 26) ones. The cognitive assistance was turned on in both groups. The final navigation trail was complex enough in order to test the capabilities and the assistance provided by the system (Fig. 2). It was divided in three different sections:

Figure 2.

Layout of the navigation trail at the rehabilitation hospital.

•

1 ${}^{\text{st}}$ section: From ICU (point 1) to the occupational therapy room (point 4).

•

2 ${}^{\text{nd}}$ section: From the occupational therapy room (point 4) to the entrance (point 6).

•

3 ${}^{\text{rd}}$ section: From entrance (point 6) to nursing stop in ward C (point 12).

The appropriate audio cues for each point on the navigation trail where imported in the system (Table 1).

Table 1

Audio cues list

Section 1: starting position $\to$ occupational therapy room
Point 1: “Walk straight ahead”
Point 2: “Turn right and walk straight”
Point 3: “Turn right and walk straight”
Point 4: “Turn back and walk straight ahead”
Point 5: “Walk straight ahead towards the front door”
Point 6: “Turn back and walk straight towards the reception desk”
Point 7: “Turn right and walk straight”
Point 8: “Turn left and walk straight”
Point 9: “Turn right and walk straight ahead”
Point 10: “Keep walking straight along the corridor”
Point 11: “Turn left”
Point 12: “You achieved your goal”

Both Group A and Group B performed the navigation trail with the MOBOT rollator-type device used in a leading mode. All participants were instructed about the final targets prior to initiating the experiment. They were instructed to try to reach the designated targets by following the directional audio cues provided by MOBOT. No period of familiarization or practice run using the MOBOT rollator was required before starting the route and no doorways had to be negotiated. The hospital staff was not allowed to give any hint or help during trials. MOBOT’s navigation assistance was provided when MOBOT reached a certain area defined around each one of the points. Each audio cue was repeated every 3 seconds until the participant has left the critical area of the point (Fig. 3). During the experiments, all participants were accompanied and supervised by health professionals in order to help or stop the experiments in case a subject was tired and prevent falling or other unexpected situations. There was no stop for the patients between the targets and the system was automatically recording the time elapsed for each target.

Figure 3.

Critical areas of the trail.

2.4 Recruitment

The recruitment process was very time consuming since the right subjects had to fit the evaluation strategy. The participation was on a volunteering basis with no other incentives provided. Due to the health conditions of the target group there was also a great risk of drop out that could be caused due to fatigue by the requested tasks. An extended period of screening and evaluation of the potential subjects and their medical files was conducted. Both inpatients and outpatients of the rehabilitation hospital as well as of other collaborating geriatric settings were examined. The inclusion criteria of the subjects were: (i) being users of an assistive device (rollator type) for at least a month, (ii) having a moderate motor impairment (habitual use of rollator, gait speed $<$ 0.6 m/sec unassisted), (iii) having a moderate impairment (unable to stand up and sit down unassisted on a normal chair (standardized 100% leg length) without problem, 5-chair stand $>$ 16.7 sec or able to stand up from a chair from elevated chair (120% lower leg length), (iv) persons aged over 65 years old, (v) with no cognitive impairment (MMSE $\geqslant$ 26–30) or mild to moderate cognitive impairment (MMSE 17–25).

2.5 Performance metrics

The MOBOT rollator-type device was assessed during the experiments using both objective and subjective performance measures.

Objective measures:

•
success rate (first/intermediate/final target position)
•
task completion time (first/intermediate/final target position)
•
stopping time

Subjective measures:

•
ATD PA Greek version [44]
•
QUEST 2.0 Greek version [42]
•
PYTHEIA questionnaire [43, 45]

Additional objective measures (walking trajectories, gait parameters captured by laser sensor) which can be applied in future work were also collected during the experiments.

Table 2
Sample descriptive characteristics

Variable Total sample ( $n=$ 30)

Age, years, mean (SD) 78.6 (6.9)

Female, $n$ , (%) 17 (56.7)

MMSE score, mean (SD) 24.8 (4.2)

Barthel-index score, mean (SD) ${}^{{\dagger}}$ 76.8 (14.5)

GDS score, median (range) 4.0 (0–14)

BBS score, mean (SD) 29.2 (10.7)

Recent history of falls, $n$ (%) ${}^{{\dagger}}$ 22 (73.3)

Short FES-I score, median (range) 13.0 (7–28)

Living situation, $n$ (%)

Community-dwelling 15 (50.0)

Institutionalized 15 (50.0)

POMA score, mean (SD) 15.0 (4.2)

Habitual gait speed, m/s (SD) ${}^{{\dagger}{\dagger}}$ 0.24 (0.13)

Note. $n=$ number of cases; MMSE $=$ Mini-Mental State Examination; GDS $=$ Geriatric Depression Scale, 15 items; SF-12 $=$ 12-item Short Form Health Survey; BBS $=$ Berg Balance Scale; FES-I $=$ Falls Efficacy Scale – International, 7-item version; POMA $=$ Performance Oriented Mobility Assessment. ${}^{{\dagger}}$ Based on interview-led self-reports; ${}^{{\dagger}{\dagger}}$ Assessed by 4-Meter-Walk-Test.

Table 3
Total time of the walk

Group statistics

Gender N Mean Std. deviation Std. error mean

Total time Female 15 306.64756 99.083074 25.583140

Male 12 337.99944 171.980798 49.646580

Test statistics ${}^{\text{a}}$

Total time

Mann-Whitney U 89.000

Wilcoxon W 167.000

Z $-$ 0.049

Asymp. sig. (2-tailed) 0.961

Exact sig. [2*(1-tailed sig.)] 0.981 ${}^{\text{b}}$

a. Grouping Variable: Gender. b. Not corrected for ties.

Figure 4.
Total time of the walk.

2.6 Statistical analysis

Variable	Total sample ( $n=$ 30)
Age, years, mean (SD)	78.6	(6.9)
Female, $n$ , (%)	17	(56.7)
MMSE score, mean (SD)	24.8	(4.2)
Barthel-index score, mean (SD) ${}^{{\dagger}}$	76.8	(14.5)
GDS score, median (range)	4.0	(0–14)
BBS score, mean (SD)	29.2	(10.7)
Recent history of falls, $n$ (%) ${}^{{\dagger}}$	22	(73.3)
Short FES-I score, median (range)	13.0	(7–28)
Living situation, $n$ (%)
Community-dwelling	15	(50.0)
Institutionalized	15	(50.0)
POMA score, mean (SD)	15.0	(4.2)
Habitual gait speed, m/s (SD) ${}^{{\dagger}{\dagger}}$	0.24	(0.13)

Group statistics
Total time	Female	15	306.64756	99.083074	25.583140
	Male	12	337.99944	171.980798	49.646580

For the statistical analysis and inference regarding the collected data, we first examined their distribution. The normally distributed data variables were analyzed using $t$ tests (for independent or paired samples) and the rest of them using the Mann-Whitney U or Kruskal-Wallis non-parametric tests. In most cases, regardless of the normality, non-parametric tests were used along $t$ tests since our sample was limited to 30 responses. We used analysis of variance (ANOVA, parametric or not) to examine the possible effects of the different variables measured during the experiment. The Pearson correlation was also applied to investigate certain associations between variables. Finally, a principal component analysis (PCA) was performed to extract the variable classes that explain the overall behavior of our sample. The IBM SPSS Statistics for Windows, version 25.0 (IBM Corp., Armonk, NY, USA) was used for the purpose of the analysis.

3. Results

After a long period of recruitment process, a total number of 30 subjects (out of the 425 screened) participated in the study. The subjects were appropriately distributed in order to have two equal groups of participants: 15 with cognitive impairment (MMSE $<$ 26) and 15 being cognitive intact (MMSE $\geqslant$ 26).

3.1 Sample description

The study did not focus on pathologies rather on the functional status of the participants as described in the inclusion criteria. The sample basic descriptive characteristics are presented in Table 2.

The average patient is 78.6 years of age and this value reflects the whole population since the age of patients is normally distributed ( $p=$ 0.338 $>$ 0.05 for the Kolmogorov-Smirnov normality test). We accept that our findings correspond to a population consisted equally of male and female patients, as well as of equally institutionalized and non-institutionalized patients ( $p=$ 0.885 $>$ 0.05 and $p=$ 0.856 $>$ 0.05 for the two corresponding binomial tests, Table 3). Moreover, male and female patients exhibit the same distribution of age and, therefore, the same mean and variance ( $p=$ 0.408 $>$ 0.05 for the non-parametric Mann-Whitney test, Table 3).

Figure 5.

Total time of the walk and age.

The mean MMSE score was 24.8 $\pm$ 4.2. The two groups (Groups A and B) were exactly divided (15 members in each group) according to the participants’ cognitive status. As far as the activities of daily living (ADL) and mobility are concerned, the majority of the subjects (20/30) were moderate dependent (need for physical assistance), 5 had severe dependency and 5 more had only slight dependency [46]. Only 10% of the participants indicated depressive symptoms [47, 48]. According to the collected data, 93,4% of the subjects had moderate (9 $\leqslant$ Short FES-I $\leqslant$ 13) to high (14 $\leqslant$ Short FES-I $\leqslant$ 28) concerns about falling and only 6.6% of them had low concerns [49]. Twenty of them (77.7%) reported at least one fall during the last year. The average POMA score (15 $\pm$ 4.2) and the walking speed (0.24 $\pm$ 0.13) indicated frailty and low motor performance for the participants [50]. Furthermore, half of the subjects were institutionalized.

3.2 Navigation assessment

The accomplishment time (total time) for each participant was measured by adding the times spent in each point (Point 1 to 12) along with the ones spent while walking between the points until reaching the final target (Point 12). The total time is illustrated in the left boxplot and histogram of Fig. 4, with mean value 320.5 s and median 287.3 s. In a 95% confidence interval, their mean times range from 267 s (4.5 min) to 373 s (6.2 min). Note that the mean value is higher than the median since two patients were a lot slower than the rest (over 600 s), which is also confirmed by the aforementioned boxplot, and hence the median is more informative about the central tendency of the total time.

Male and female patients show no difference in their average time of the walk ( $p=$ 0.981 $>$ 0.05 for the independent samples Mann-Whitney U test, Table 3); see the right boxplot and histograms in Fig. 4. Note that the dashed line all the figures represents the corresponding (total) mean value.

In relation to their age, there is no difference in their average walk time ( $p=$ 0.413 $>$ 0.05 for the independent samples Kruskal-Wallis test, as well as $p=$ 0.078 $>$ 0.05 for the corresponding Pearson’s correlation, Table 4). For details, see boxplot and the mean bar chart in Fig. 5.

Table 4
Total time of the walk with respect to age

Correlations
		Total time	Age
Total time	Pearson correlation	1	0.359
	Sig. (2-tailed)		0.078
	N	25	25
Age	Pearson correlation	0.359	1
	Sig. (2-tailed)	0.078
	N	25	30

Test statistics ${}^{\text{a,b}}$
	Total time
Kruskal-Wallis H	1.769
df	2
Asymp. sig.	0.413

a. Kruskal Wallis Test. b. Grouping Variable: Decades of age.

Table 5

Section completion times of the trail

Tests of normality
	Section no.	Kolmogorov-Smirnov ${}^{\text{a}}$			Shapiro-Wilk
		Statistic	df	Sig.	Statistic	df	Sig.
Completion time	1	0.085	28	0.200 ${}^{*}$	0.973	28	0.670
	2	0.171	25	0.057	0.919	25	0.048
	3	0.166	26	0.063	0.943	26	0.158

Descriptives
Completion time
					95% confidence interval for mean
	N	Mean	Std. deviation	Std. error	Lower bound	Upper bound	Minimum	Maximum

1	29	86.27943	59.750385	11.095368	63.55159	109.00726	30.083	358.000
2	28	77.83607	52.408938	9.904358	57.51401	98.15814	23.800	225.000
3	27	160.53148	56.366226	10.847685	138.23375	182.82922	78.250	329.950
Total	84	107.33171	66.799301	7.288401	92.83537	121.82804	23.800	358.000

Paired samples test
		Paired differences
		Mean	Std. deviation	Std. error mean	95% confidence interval of the difference		$t$	df	Sig. (2-tailed)
					Lower	Upper
Pair 1	Section 1 time – Section 2 time	9.009881	47.682972	9.011235	$-$ 9.479645	27.499407	1.000	27	0.326
Pair 2	Section 2 time – Section 3 time	$-$ 86.331111	52.004370	10.008246	$-$ 106.903355	$-$ 65.758867	$-$ 8.626	26	0.000

Test statistics ${}^{\text{a}}$
N	27
Chi-square	32.667
df	2
Asymp. sig.	0.000

${}^{*}$ This is a lower bound of the true significance. a. Lilliefors Significance Correction.

a. Friedman test.

Figure 6.

Section completion times of the trail.

Considering the three sections of the trail, they exhibit different patterns; see boxplot and means bar plot in Fig. 6. This pattern of mean times can be (statistically) confirmed for the whole population ( $p<$ 0.01 for the independent samples Kruskal-Wallis test in Hypothesis Test Summary sub-table of Table 5). In a more detailed analysis, the completion time of the first and second sections shown the same average behavior, while the time for the third section does not ( $p=$ 0.326 $>$ 0.05 for the paired $t$ test between the 1 ${}^{\text{st}}$ and the 2 ${}^{\text{nd}}$ section completion times, and $p<$ 0.01 for the paired $t$ -test between the 2 ${}^{\text{nd}}$ and the 3 ${}^{\text{rd}}$ section completion times, Table 5; all were normally distributed in K-S test).

Only one participant failed to complete the first section (Point 1 to Point 4) of the trail, and one more failed at the second section (Point 4 to Point 6). All other participants completed the task.

Table 6

Walk time in relation to institutionalized patients

Group statistics
	Institutionalized	N	Mean	Std. deviation	Std. error mean
Total time	No	13	330.40308	171.520339	47.571183
	Yes	14	311.46190	93.602917	25.016432

Tests of normality
	Institutionalized	Kolmogorov-Smirnov ${}^{\text{a}}$			Shapiro-Wilk
		Statistic	df	Sig.	Statistic	df	Sig.
Total time (no outliers)	No	0.241	11	0.073	0.859	11	0.056
	Yes	0.214	14	0.081	0.919	14	0.215

Independent samples test
		Levene’s test for equality of variances		$t$ -test for equality of means
		F	Sig.	$t$	df	Sig. (2-tailed)	Mean difference	Std. error difference	95% confidence interval of the difference
									Lower	Upper
Total time	Equal variances assumed	4.228	0.050	0.360	25	0.722	18.941	52.63	$-$ 89.469	127.352
	Equal variances not assumed			0.352	18.265	0.729	18.941	53.748	$-$ 93.861	131.744

Test statistics ${}^{\text{a}}$
	Total time
Mann-Whitney U	79.	000
Wilcoxon W	170.	000
Z	$-$ 0.	582
Asymp. sig. (2-tailed)	0.	560
Exact sig. [2*(1-tailed sig.)]	0.	583 ${}^{\text{b}}$

Test statistics ${}^{\text{a}}$
	Total time
Mann-Whitney U	24.	000
Wilcoxon W	69.	000
Z	$-$ 2.	932
Asymp. sig. (2-tailed)	0.	003
Exact sig. [2*(1-tailed sig.)]	0.	002 ${}^{\text{b}}$

a. Lilliefors Significance Correction.

a. Grouping Variable: Institutionalized. b. Not corrected for ties.

a. Grouping Variable: Fear of falling. b. Not corrected for ties.

Institutionalized patients may seem to vary their total walk time in our sample (Fig. 7), however we infer that the population of the institutionalized or not patients (using the MOBOT device) have on average the same total walk time ( $p=$ 0.722 $>$ 0.05 for independent samples $t$ test while the total walk time of the institutionalized and non-institutionalized patients are normally distributed with $p>$ 0.05 for both categories in Kolmogorov-Smirnov and Sapiro-Wilk tests). Note, however, that our sample has less than 30 responses for the institutionalized and the non-institutionalized patients. Alternatively, the above result can also be confirmed since the distributions of their walk time (institutionalized or not) are statistically the same ( $p=$ 0.583 $>$ 0.05 for the independent samples Mann-Whitney U test, Table 6). In relation to their total walk time, the subjects facing fear of falling are on average slower than those who have no fear, as it was expected ( $p=$ 0.02 $<$ 0.05 for the independent samples Mann-Whitney U test, Table 6).

Table 7

Walk time and MMSE

Mental status (MMSE categories)
		Frequency	Percent	Valid percent	Cumulative percent
Valid	No mental issue	13	43.3	43.3	43.3
	Mental issue	17	56.7	56.7	100.0
	Total	30	100.0	100.0

Binomial test
		Category	N	Observed prop.	Test prop.	Exact sig. (2-tailed)
Institutionalized	Group 1	Mental issue	17	0.57	0.50	0.585
	Group 2	No mental issue	13	0.43
	Total		30	1.00

Test statistics ${}^{\text{a}}$
	Total time
Mann-Whitney U	89.	000
Wilcoxon W	209.	000
Z	$-$ 0.	049
Asymp. sig. (2-tailed)	0.	961
Exact sig. [2*(1-tailed sig.)]	0.	981 ${}^{\text{b}}$

Correlations
		Total time	MMSE (mental ability)
Total time	Pearson correlation	1	$-$ 0.079
	Sig. (2-tailed)		0.696
	N	27	27
MMSE (mental ability)	Pearson correlation	$-$ 0.079	1
	Sig. (2-tailed)	0.696
	N	27	30

a. Grouping Variable: Mental status (MMSE categories). b. Not corrected for ties.

Figure 7.

Walk time in relation to institutionalization.

Figure 8.

Walk time and MMSE.

Figure 9.

Total time of the walk and functional independence (BARTHEL).

The histogram (top left sub-figure) and the boxplot next to it in Fig. 8 illustrate the MMSE values of the patients’ sample with mean 24.67. In our sample, 56.7% of the patients have mental issues (MMSE $\leqslant$ 26), but we can assume that half of the test population have mental issues ( $p=$ 0.585 $>$ 0.05 for the binomial test, Table 7) and, thus, our findings correspond the populations where half of them has some mental issue. In relation with the total time of the patients’ walk, we strongly infer that there is no difference on the average total time between patients with mental issues or not ( $p=$ 0.981 $>$ 0.05 for the independent samples Mann-Whitney U test, Table 7); see also the boxplots in Fig. 8. Therefore, the MMSE score (mental status) of the patients does not affect their average time. The above result can also be confirmed by the fact that there is statistically no correlation between the MMSE values themselves and the total time of the walk ( $p=$ 0.696 $>$ 0.05 for the Pearson’s correlation between MMSE and total walk time, Table 7).

Similar to the results for the MMSE, the BARTHEL values do not have a statistically significant association with the total time of the walk ( $p=$ 0.066 $>$ 0.05 for the Pearson’s correlation). Moreover, regarding their BARTHEL functional independent categories (severe for BARTHEL values $\leqslant$ 60, moderate for values $>$ 61 and $\leqslant$ 90 and slight for values $\geqslant$ 91), we infer also that their average total walking time does not differ ( $p=$ 0.207 $>$ 0.05 for the corresponding Kruskal-Wallis independent samples test, Table 8), although the corresponding boxplot shows some time differences between the BARTHEL categories in our sample (Fig. 9).

Table 8

Total time of the walk and functional independence (BARTHEL)

Correlations
		BARTHEL	Total time
BARTHEL	Pearson correlation	1	$-$ 0.374
	Sig. (2-tailed)		0.066
	N	30	25
Total time	Pearson correlation	$-$ 0.374	1
	Sig. (2-tailed)	0.066
	N	25	25

Test statistics ${}^{\text{a,b}}$
	Total time
Kruskal-Wallis H	3.147
df	2
Asymp. sig.	0.207

a. Kruskal Wallis Test. b. Grouping Variable: Functional independence (BARTHEL categories).

Table 9

Satisfaction ratings for MOBOT

Correlations
		Rating (PYTHEIA)	Total time
Rating (PYTHEIA)	Pearson correlation	1	$-$ 0.051
	Sig. (2-tailed)		0.805
	N	29	26
Total time	Pearson correlation	$-$ 0.051	1
	Sig. (2-tailed)	0.805
	N	26	27

Correlations
		Rating (PYTHEIA)	BARTHEL
Rating (PYTHEIA)	Pearson correlation	1	$-$ 0.051
	Sig. (2-tailed)		0.793
	N	29	29
BARTHEL (functional independence)	Pearson correlation	$-$ 0.051	1
	Sig. (2-tailed)	0.793
	N	29	30

Test statistics ${}^{\text{a}}$
	Rating (PYTHEIA)
Mann-Whitney U	53.500
Wilcoxon W	144.500
Z	$-$ 2.218
Asymp. sig. (2-tailed)	0.027
Exact sig. [2*(1-tailed sig.)]	0.025 ${}^{\text{b}}$

a. Grouping Variable: Gender. b. Not corrected for ties.

Figure 10.

Satisfaction ratings for MOBOT.

Figure 11.

Scree plot for the PCA.

Regarding the overall subjective satisfaction form the use of the MOBOT Rollator, Fig. 10 demonstrates the corresponding histogram and boxplot of the PYTHEIA values with mean 3.98 measuring the overall satisfaction of the users with MOBOT. When responding to Part B of the PYTHEIA questionnaire, subjects assessed the cognitive assistance functionality with 4.46 and the audio-gestural interaction functionality with 4.30, indicating their high level of satisfaction from the experience of using these features [43, 51, 52]. These ratings are not statistically correlated with their total time of the walk, nor their MMSE and BARTHEL values ( $p=$ 0.805 $>$ 0.05, $p=$ 0.703 $>$ 0.05 for the Pearson correlation between PYTHEIA values and MMSE, and BARTHEL values, Table 9). On the other hand, PYTHEIA ratings are affected by the gender ( $p=$ 0.025 $<$ 0.05 for the Mann-Whitney U test), with the women rating, on average, the MOBOT Rollator higher than men; see Fig. 10.

Table 10

Principal component analysis

KMO and bartlett’s test
Kaiser-meyer-olkin measure of sampling adequacy.		0.585
Bartlett’s test of sphericity	Approx. chi-square	148.179
	df	78
	Sig.	0.000

Total variance explained
Component	Initial eigenvalues			Extraction sums of squared loadings			Rotation sums of squared loadings ${}^{\text{a}}$
	Total	% of variance	Cumulative %	Total	% of variance	Cumulative %	Total
1	4.014	30.880	30.880	4.014	30.880	30.880	3.828
2	2.308	17.757	48.637	2.308	17.757	48.637	2.503
3	1.482	11.402	60.039	1.482	11.402	60.039	2.066
4	1.348	10.365	70.404	1.348	10.365	70.404	1.472
5	0.980	7.540	77.944
6	0.800	6.155	84.099
7	0.759	5.837	89.937
8	0.386	2.968	92.904
9	0.308	2.368	95.273
10	0.228	1.752	97.025
11	0.158	1.213	98.238
12	0.146	1.126	99.363
13	0.083	0.637	100.000

Pattern matrix ${}^{\text{a}}$
	Component
	1	2	3	4
FIM (functional ability)	0.936
BERG BS (functional balance)	0.805
BARTHEL (functional independence)	0.798
Total time	$-$ 0.656
P.O.M.A. – TINETTI (mobility test)	0.646
Gender		0.886
Rating (PYTHEIA)		$-$ 0.747
FES-7 (concerns of falling)
Fear of falling			0.850
Age			0.725
Institutionalized			0.604
MMSE (mental ability)				0.939
GDS (depression)

Extraction Method: Principal Component Analysis. a. When components are correlated, sums of squared loadings cannot be added to obtain a total variance.

Extraction Method: Principal Component Analysis. Rotation Method: Promax with Kaiser Normalization. a. Rotation converged in 13 iterations.

Applying the Principal Component Analysis (PCA with Promax Rotation Method), we obtained four components that explain 70.4% of the total variance in our sample ( $p<$ 0.01 for the KMO and Bartlett’s test, Table 10); see also the corresponding scree plot in Fig. 11. We observe that the MMSE variable is a group by itself, meaning that MMSE participate to the (total variance of the) behavior of the patients in a unique and distinctive way (in relation to the other factors) as they use the MOBOT device; see Pattern Matrix in Table 9. In contrast, BARTHEL, FIM, BERG and POMA values (1 group) influence the total variance more or less in the same way. Note that the satisfaction variable (PYTHEIA) is grouped only with gender, which means that the overall satisfaction of the patients it is not related to some other characteristics of the patients and thus we can safely derive that they all more or less have the same level of satisfaction which is only distinguished by the gender.

4. Discussion

The present study focused on the assessment of the cognitive assistance provided by a robotic rollator utilizing navigation functionalities. The target group was frail seniors with and without cognitive impairment. The study design was based on the findings of previous research activities [28] in order to follow an appropriate methodological approach. Clinical measurements and observations were the base for selecting an adequate number and sample of participants. The subjects, following the inclusion criteria, are the targeted end-users of the evaluated robotic assistive device. Their physical health condition was also taken into account in the navigation trail design in order to avoid dropouts of the participants due to the effort, tiredness and other related factors. According to the results, almost all participants reached the final target following the navigation trail. Only two subjects didn’t make it to achieve the final target (point 12 of the trail). The first one failed from the beginning and the second one finished only the first section of the trail.

As indicated by the current research, a major outcome is that their cognitive status did not influence the results of the two different groups (cognitive impaired and not). As a common ground, all participants had no previous knowledge of the places where the navigation took place. Their only help was by the MOBOT system (using the audio cues presented in Table 1, which were triggered automatically when they stopped in a position for some time or when they were lost). The results reveal that even cognitive impaired people can achieve results similar to normal groups when using innovative technologies like the one provided by the MOBOT robotic rollator. The functionality offered (navigation assistance) is really helpful when navigating in unknown spaces where no signs may appear or when orientation problems appear.

The short completion time of the cognitive impaired group was apparently achieved based on the navigation assistance given by the system. Participant’s disorientation in several points was immediately improved by the given assistance, resulting in the achieved completion time which was comparable with the one of the not cognitive impaired group (MMSE $<$ 26). The present study suggests that the integration of navigation assistance in rollators or other mobility assistance devices can help elderly to effectively orient themselves in challenging and unknown environments. The directional audio cues provided by MOBOT at critical waypoints led also to smoother walking paths. As risk of falling is related with the walking pattern [53], this can be considered as another valuable contribution towards safety issues. Furthermore, audio assistance seems to be more effective for navigation than providing visual cues. This was confirmed by the high rate of satisfaction given by the participants in this feature. The combination of a challenging and attention-demanding orientation task along with the requirement to walk while navigating contribute to lower motor and cognitive performances [54]. Our results suggest that implementing navigation assistance in rollators can overcome the mentioned obstacles.

In this study we have chosen to provide audio cues to convey navigation assistance messages, which was made possible because all of the subjects included in the study had no major hearing impairment. To cope with mild to moderate impairments, the volume/loudness of the audio cues was adjusted to the needs of each participant, which proved to be a helpful personalization feature of the MOBOT navigation assistance system. The use of headphones is of course another available and easy to install option, which could provide an additional solution also coping with environment regulations regarding loudness and noise level (e.g. in hospital settings where it would not be proper to use very loud audio cues). In future research work, it is envisioned to further explore other features related to online personalization of the system, such as automatically adjusting the volume of audial cues in different areas depending on the anticipated (or detected) background noise level (e.g. large crowded hallways etc.), in order to improve the user-friendliness of such a navigation assistance modality. We also plan to further explore and compare the use of other feedback modalities including combinations of audio-visual navigation cues.

Such a technology can be proved helpful in several environments. For example, a MOBOT-like device can help carers, physiotherapists, occupational therapists and nurses in rehabilitation hospitals, since they will no longer be required to accompany their patients at every step. The decongestion of paramedical staff is considered an extremely important issue in hospitals that also affects their economics. The hospital’s staff can focus on more important tasks in their daily routine. However, safety issues should always be taken into account and a more in-depth research study is needed before deciding to leave a patient totally unsupervised by a carer using any such device. Moreover, the integration of navigation assistance in a rollator could reduce the fear of getting lost or falling in the targeted group. This could also lead to improved physical and social activities. The same device could also be used in a non-controlled environment after some modifications. Even though the initial designed MOBOT robotic rollator was used in a constrained environment (hospital setting), it is envisioned that after appropriate mapping of other places, using for example predefined guard points, would allow its use in less controlled places for more typical use. Current and future research work in follow-up projects include extending the use-case scenarios to address a larger scope of application environments, as well as integrating the technologies on a more commercial-oriented prototype with improved functionalities, much smaller size and more elegant than the original MOBOT prototype used in this study.

In the early years of many technical fields, the research community often utilizes a wide range of metrics that are not comparable due to a bias towards application specific measures. The primary difficulty in defining common metrics is the incredibly diverse range of human-robot applications. Thus, although metrics from other fields (HCI, human factors, etc.) can be applied to satisfy specific needs, identifying metrics that can accommodate the entire application space may not be feasible [55]. Attempts to categorize both objective and subjective metrics have been made. In our study, we combined both objective and subjective measurements in order to conclude on the usefulness and acceptance of the provided service/product. However, as reported in the Introduction section, the subjective measurement of user satisfaction is not an easy task. As reported in [28] most of the studies are utilizing either custom-made questionnaires or interviews that are neither valid nor reliable instruments to represent the subjective opinion and perception of the end users. The absence of standard scales/questionnaires for the subjective assessment of robot-based devices makes it difficult to design products that meet exactly the needs of the intended end users, to further improve prototypes, or to compare the results from different researchers. A great advantage of our study was the fact that all three questionnaires used (PYTHEIA, ATDPA, QUEST 2.0) were already tested for their reliability and validity in the Greek population [42, 43, 44, 45], thus claiming valid results. Moreover, PYTHEIA’s capability to capture the satisfaction of the end users with assistive and robotic technologies, while being able to evaluate any individual features/functionalities implemented was proved to be valuable. This is not present in any other scale.

The early abandonment of assistive devices is also a major issue. Certain studies show that a big number of patients don’t use (or abandon early) their gait assistive device, when other studies speak of patients using more than one mobility aid. Two of the most common limitations or dissatisfaction reasons while using the mobility aid are: handling the rollator, social stigmatizing. According to the results of the presented study, the users not only enjoyed using the robotic rollator, but also were feeling more safe and secure, while thinking of it as a hi-tech gadget. Thus, they wanted and really enjoyed to participate in the study even without any other incentives given. Social stigmatizing seemed not to be concern due to the innovative characteristics of the robotic rollator.

The collected results on objective performance metrics of the cognitive assistance scenario are quite promising, indicating that cognitively impaired participants achieved comparable performance (i.e. task completion time) with non-cognitive impaired subjects, using the MOBOT’s cognitive assistance system when completing a navigation task in a complex real-world application scenario. Results on the subjective user satisfaction are also very positive and suggest that the use of this cognitive assistance functionality provides a benefit for the users, do not cause feelings of insecurity, is an interesting challenge for the users, and may have a positive effect on the user’s quality of life. These results indicate the potential added value of such a robotic functionality for cognitively impaired, frail older people. Of course, regarding the positive results obtained in relation to the feeling of safety when using the device, we should point out the following. Though there has been a lot of research conducted in the frames of the MOBOT project focusing on topics such as human gait tracking, analysis, and characterization of pathological and stability features (using data provided by multiple on-board sensors), there was in fact no specific “fall-prevention” functionality integrated on the MOBOT rollator platform, which constitutes actually an important topic for future research. Therefore, we could attribute the feeling of safety perceived by MOBOT users as partly related to the apparent mechanical properties (mass, inertia) of the device, which have been actively controlled by means of an adaptive admittance controller (exploiting data provided by force/torque sensors mounted on the rollator handles); thus, apparently providing a reassuring feeling of safety when handling and maneuvering the device, assessed in the context of the subjective evaluation procedures of this study.

We should also point out here that, as it has been assessed by early experimental studies conducted in the frames of the MOBOT project, the use of the MOBOT rollator (with actively controlled admittance) does not have by itself any significant effect in the walking speed of the user as compared to a typical rollator. We can thus argue that the MOBOT device does not seem to constitute a limiting factor to the walking performance of rollator users, meaning that the results achieved by the non-cognitively impaired group can indeed be considered to constitute a measure of comparison, for assessing the effect of the navigation assistance functionality to the cognitively impaired group. Previous studies [56] have also confirmed that the performance achieved by cognitively impaired users, without any navigation assistance provided, is indeed significantly inferior compared to non-cognitively impaired users, a finding which also adds value to the results obtained by the study presented in this paper.

As far as the limitations of the study are concerned, we can refer the limited number of participants and the complexity of the navigation trail. Moreover, the hi-tech nature and novelty of the device with the natural tendency of the users to want to give positive responses to please the research team conducting the tests may also have bias the results to give more favourable results in the context of a subjective evaluation without the possibility of comparison with an alternative system. In a future research, we plan to test MOBOT taking into account the above parameters so as to have more accurate and generalized results. A new research effort has already begun in order to build a more elegant and smaller in size system able to be adopted in more typical environments outside a hospital.

5. Conclusions

The current research presents the integration of navigation assistance in rollator type devices. According to the results presented, this can effectively support orientation and navigation of cognitive impaired users.

Footnotes

Acknowledgments

We would like to thank the Diaplasis Rehabilitation center, the physiatrist A. Karavasili and her team, for helping us in the recruitment process and in collecting all the relevant data from the patients participating in the study. The work leading to the presented results has received funding from the European Union under grant agreement no. 600796.

Conflict of interest

None to report.

References

WHO. Chapter 1: Global Health: today’s challenges [Internet]. Who.int. 2020 [cited 30 January 2020]. Available from: https://www.who.int/whr/2003/chapter1/en/index1.html

United Nations, Department of Economic and Social Affairs, Population Division; 2017. World Population Prospects: The 2017 Revision, Key Findings and Advance Tables. (ESA/P/ WP/248).

World Health Organization [Internet]. Ageing and Health. [cited 30 January 2020]. Available from: https://www.who.int/news-room/fact-sheets/detail/ageing-and-health.

United Nations, Department of Economic and Social Affairs, Population Division; 2015. World Population Ageing 2015 (ST/ESA/SER.A/390).

United States Census Bureau [Internet]. Mobility is Most Common Disability Among Older Americans, Census Bureau Reports. [cited 30 January 2020]. Available from: https://www.census.gov/newsroom/press-releases/2014/cb14-218.html.

Walsh

Caraceni

Fainsinger

Foley

Glare

Goh

, et al. Palliative Medicine. 1st ed.. Elsevier; 2009, pp. 570-573.

Charron

Kirby

MacLeod

. Epidemiology of walker-related injuries and deaths in the united states. American Journal of Physical Medicine & Rehabilitation. 1995; 74(3): 237-239. doi: 10.1097/00002060-199505000-00011.

Roman de Mettelinge

Cambier

. Understanding the relationship between walking aids and falls in older adults. Journal of Geriatric Physical Therapy. 2015; 38(3): 127-132. doi: 10.1519/jpt.0000000000000031.

Fortin

Yeaw

Campbell

Jameson

. An analysis of risk assessment tools for falls in the elderly. Home Healthc Nurse. 1998; 16: 624-629.

10.

Bateni

Maki

. Assistive devices for balance and mobility: benefits, demands, and adverse consequences. Arch Phys Med Rehabil. 2005; 86(1): 134-145.

11.

Stevens

Thomas

The

, et al. Unintentional fall injuries associated with walkers and canes in older adults treated in U.S. emergency departments. J Am Geriatr Soc. 2009; 57(8): 1464-1469.

12.

Martins

Frizera-Neto

Urendes

dos Santos

Ceres

Bastos-Filho

. A novel human-machine interface for guiding: The NeoASAS smart walker. Proceedings of the 2012 ISSNIP Biosignals and Biorobotics Conference: Biosignals and Robotics for Better and Safer Living (BRC); Manaus; 2012, pp. 1-7. 2.

13.

Cetin

Muzembo

Pardessus

, et al. Impact of different types of walking aids on the physiological energy cost during gait for elderly individuals with several pathologies and dependent on a technical aid for walking. Ann Phys Rehabil Med. 2010 Aug-Sep; 53(6-7): 399-405.

14.

Thomas

Halbert

Mackintosh

, et al. Walking aid use after discharge following hip fracture is rarely reviewed and often inappropriate: an observational study. J Physiother. 2010; 56(4): 267-272.

15.

Scherer

. The impact of assistive technology on the lives of people with disabilities. In: Gray

Quatrano

Lieberman

, eds. Designing and using assistive technology: the human perspective. Baltimore: PH Brooks; 1998, pp. 99-116.

16.

Akinbo

Sokunbi

Ogunbameru

. Factors contributing to possession of walking aids among persons with osteoarthritis of the knee. Nig Q J Hosp Med. 2008 Apr-Jun; 18(2): 64-68.

17.

Lezzoni

Rao

Kinkel

. Patterns of mobility aid use among working-age persons with multiple sclerosis living in the community in the United States. Disabil Health J. 2009 Apr; 2(2): 67-76.

18.

Van der Loos

Reinkensmeyer

. Rehabilitation and Health Care Robotics. In: Siciliano

Khatib

, eds. Springer Handbook of Robotics. Springer, Berlin, Heidelberg.

19.

Frizera

Ceres

Pons

, et al. The Smart Walkers as Geriatric Assistive Device. the Simbiosis Purpose. In Proceedings of the 6th International Conference of the International Society for Gerontechnology; 2008. pp. 1-6.

20.

Martins

Santos

Frizera-Neto

, et al. Assistive mobility devices focusing on smart walkers: classification and review. Robotics and Autonomous Systems. 2012 Apr; 60(4): 548-562.

21.

Agrawal

. Walk-Assist Robot: A Novel Approach to Gain Selection of a Braking Controller Using Differential Flatness. Proceedings of the 2010 American Control Conference; Baltimore, MD; 2010, pp. 2799-2804.

22.

Kulyukin

Kutiyanawala

LoPresti

, et al. iWalker: Toward a Rollator-Mounted Wayfinding System for the Elderly. In 2008 IEEE International Conference On RFID; 2008; pp. 303-311.

23.

Spenko

Dubowsky

. Robotic personal aids for mobility and monitoring for the elderly. IEEE Trans Neural Syst Rehabil Eng. 2006 Sep; 14(3): 344-351.

24.

Shin

Itten

Rusakov

, et al. SmartWalker: Towards an Intelligent Robotic Walker for the Elderly. In 2015 International Conference on Intelligent Environments, Prague; 2015, 9-16.

25.

Geunho

Jung

Ohnuma

, et al. JAIST Robotic Walker Control Based on a Two-Layered Kalman Filter. In 2011 IEEE International Conference on Robotics and Automation, Shanghai; 2011, pp. 3682-3687.

26.

Graf

Matthias

Rolf

. Care-o-bot II – development of a next generation robotic home assistant. Autonomous Robots. 2004; 16(2): 193-205.

27.

Thordardottir

Fänge

Lethin

Gatta

Chiatti

. Acceptance and Use of Innovative Assistive Technologies among People with Cognitive Impairment and Their Caregivers: A Systematic Review. BioMed Research International. 2019; Article ID 9196729, 18 pages. doi: 10.1155/2019/9196729.

28.

Koumpouros

. A Systematic Review on Existing Measures for the Subjective Assessment of Rehabilitation and Assistive Robot Devices. Journal of Healthcare Engineering. 2016; Article ID 1048964. doi: 10.1155/2016/1048964.

29.

MOBOT project website [Internet]. MOBOT project. [cited 23 January 2020]. Available from: http://www.mobot-project.eu/?Description.

30.

Geravand

Korondi

Werner

, et al. Human sit-to-stand transfer modeling towards intuitive and biologically-inspired robot assistance. Auton Robots. 2017; 41: 575-592.

31.

Peer

Geravand

Tzafestas

Moustris

. Deliverable D3.2: Report on autonomous navigation algorithms, online configuration optimization, motion control, and physical human-robot interaction and collaboration. 2015.

32.

Werner

Moustris

Tzafestas

, et al. User-oriented evaluation of a robotic rollator that provides navigation assistance in frail older adults with and without cognitive impairment. Gerontology. 2018; 64(3): 278-290.

33.

Moustris

Tzafestas

. Intention-based front-following control for an intelligent robotic rollator in indoor environments. In 2016 IEEE Symposium Series on Computational Intelligence (SSCI); 6–9 Dec. 2016, pp. 1-7.

34.

Geravand

Werner

Hauer

, et al. An integrated decision making approach for adaptive shared control of mobility assistance robots. Int J Soc Robot. 2016; 8(5): 631-648.

35.

Efthimiou

Fotinea

Goulas

, et al. The MOBOT Platform – Showcasing Multimodality in Human-Assistive Robot Interaction. In Universal Access in Human-Computer Interaction Techniques and Environments: 10th International Conference, UAHCI 2016, Held as Part of HCI International 2016, Toronto, ON, Canada, July 17–22, 2016, Proceedings, Part II. Edited by Antona M, Stephanidis C. Cham: Springer International Publishing; 2016, pp. 382-391.

36.

Quigley

Conley

Gerkey

, et al. ROS: An Open-Source Robot Operating System. In ICRA Workshop on Open Source Software. Kobe, Japan; 2009.

37.

Emmi

Gonzalez-de-Soto

Pajares

, et al. New Trends in Robotics for Agriculture: Integration and Assessment of a Real Fleet of Robots. The Scientific World Journal. 2014; Article ID: 04059. doi: 10.1155/2014/404059.

38.

ROS.org [Internet]. Navigation – ROS Wiki. [cited 28 January 2020]. Available from: http://wiki.ros.org/navigation.

39.

Klancar

Zdesar

Blazic

, et al. Wheeled Mobile Robotics: From Fundametals Towards Autonomous Systems. Butterworth-Heinemann; 2017.

40.

Ballagi

Tihamér

. Fuzzy odometry error approximation of a mobile robot. Annals of Craiova University, series. Automation, Computers, Electronics and Mechatronics. 2007; 4(31): 13-17.

41.

Thrun

Fox

Burgard

, et al. Robust monte carlo localization for mobile robots. Artificial Intelligence. 2001; 128: 99-141.

42.

Koumpouros

Karavasili

Papageorgiou

, et al. Validation of the Greek version of the device subscale of the Quebec user evaluation of satisfaction with assistive technology 2.0 (Quest 2.0). Assistive Technology. 2016; 28: 152-158.

43.

Koumpouros

Papageorgiou

Karavasili

. Development of a new psychometric scale (PYTHEIA) to assess the satisfaction of users with any assistive technology. In Duffy

Lightner

, eds. Advances in Human Factors and Ergonomics in Healthcare Proceedings of the AHFE 2016 International Conference on Human Factors and Ergonomics in Healthcare, July 27–31. Walt Disney World, Florida, USA, Springer, 2016.

44.

Koumpouros

Papageorgiou

Karavasili

, et al. Translation and validation of the assistive technology device predisposition assessment in Greek in order to assess satisfaction with use of the selected assistive device. Disabil Rehabil Assist Technol. 2017 Jul; 12(5): 535-542.

45.

Koumpouros

Papageorgiou

Karavasili

, et al. PYTHEIA: A scale for assessing rehabilitation and assistive robotics. In ICPMRSM 2016: 18th International Conference on Physical Medicine, Rehabilitation and Sports Medicine, Nov 3–4. Cape Town, South Africa, 2016.

46.

Shah

Vanclay

Cooper

. Improving the sensitivity of the barthel index for stroke rehabilitation. Journal of Clinical Epidemiology. 1989; 42: 703-709.

47.

Sheikh

Yesavage

. Geriatric Depression Scale (GDS): Recent evidence and development of a shorter version. In Brink

, ed. Clinical Gerontology: A Guide to Assessment and Intervention. New York, Haworth Press; 1986, pp. 165-173.

48.

Fountoulakis

Tsolaki

Iacovides

, et al. The validation of the short form of the geriatric depression scale (GDS) in greece. Aging Clin Exp Res. 1999; 11(6): 367-372.

49.

Delbaere

JCT

Mikolaizak

, et al. The falls efficacy scale international (FES-I). A comprehensive longitudinal validation study. Age and Ageing. 2010; 39(2): 210-216.

50.

Tinetti

. Performance-oriented assessment of mobility problems in elderly patients. J Am Geriatr Soc. 1986; 34: 119-126.

51.

Koumpouros

Papageorgiou

Karavasili

, et al. PYTHEIA: a scale for assessing rehabilitation and assistive robotics. International Journal of Medical, Health, Biomedical, Bioengineering and Pharmaceutical Engineering. 2016; 11: 522-526.

52.

Koumpouros

Karavasili

Efthimiou

, et al. User evaluation of the MOBOT rollator type robotic mobility assistive device. Technologies. 2017; 5(4): 73.

53.

Schwenk

Hauer

Zieschang

, et al. Sensor-derived physical activity parameters can predict future falls in people with dementia. Gerontology. 2014; 60: 483-492.

54.

Montero-Odasso

Muir

Speechley

. Dual-task complexity affects gait in people with mild cognitive impairment: the interplay between gait variability, dual tasking, and risk of falls. Arch Phys Med Rehabil. 2012; 93: 293-299.

55.

Steinfeld

Fong

Kaber

, et al. Goodrich, Common Metrics for Human-Robot Interaction. In HRI ’06 Proceedings of the 1st ACM SIGCHI/SIGART Conference on Human-robot Interaction; 2006, pp. 33-40.

56.

Hauer

Werner

Tzafestas

Moustris

Maragos

Pitsikalis

Kardaris

Rodomagoulakis

Koumpouros

. Deliverable D5.4: Report on performance metrics and final evaluation study. 2016.

Test statistics ${}^{\text{a}}$
	Total time
Mann-Whitney U	89.000
Wilcoxon W	167.000
Z	$-$ 0.049
Asymp. sig. (2-tailed)	0.961
Exact sig. [2*(1-tailed sig.)]	0.981 ${}^{\text{b}}$

Assessment of an intelligent robotic rollator implementing navigation assistance in frail seniors

Abstract

BACKGROUND:

OBJECTIVE:

METHODS:

RESULTS:

CONCLUSIONS:

Keywords

1. Introduction

2. Materials and methods

2.1 Intelligent mobility assistance robot

2.3 Design and experimental setup

2.5 Performance metrics

3. Results

3.1 Sample description

Table 4 Total time of the walk with respect to age

5. Conclusions

Footnotes

Acknowledgments

Conflict of interest

References

Table 4
Total time of the walk with respect to age