Health-CAT: Development of a Mobile Robot for Assisting Caregivers

William K. Juel

, Frederik Haarslev

and Leon Bodenhagen

SDU Robotics, University of Southern Denmark, Campusvej 55, Odense C, Denmark

Keywords:

Healthcare Robotics, Human-robot Interaction, Participation.

Abstract:

The demographic change is expected to challenge the healthcare sector which in many countries is already

struggling today leading to, e.g., a shortage of staff. Since robot technology is playing a minor role in health-

care today, robotics is considered to be one mean to mitigate some of the challenges related to the demographic

changes. This paper discusses hurdles for introducing robotics solutions in healthcare and describes the identi-

ﬁcation of a use case as well as the development of a robot prototype. End users have been involved throughout

an iterative development process leading to a prototype that has been tested during normal operations.

1 INTRODUCTION

The demographic change is challenging many health-

care systems. Not only is the share of citizens in the

working age declining but also the demand for health-

care services is increasing since the number of el-

derly citizens is growing and this population group

has a higher prevalence for comorbidities. Already

today a shortage of staff can be observed in the care-

giving domain (Drennan and Ross, 2019) which only

can be expected to worsen given both the general de-

mographic development and, recently, the increase

of the turnover intentions amongst nurses during the

COVID-19 pandemic (Falatah, 2021).

Lately substantial progress has been achieved in

the domain of robot technology in general, includ-

ing related areas such as computer vision or artiﬁ-

cial intelligence. However, unlike the industrial do-

main, robots are currently only playing a minor role in

healthcare, apart from niche applications such as sur-

gical robotics or pilot cases. Therefore robotics can

be considered to present a not yet utilized potential

mean for at least partially mitigating the demographic

challenge (Bodenhagen et al., 2019).

The Health-CAT project

was initiated prior to the

pandemic and focuses on identifying how robot tech-

nology that is largely already available can be ex-

ploited to support care givers and develops and tests

a prototype of such a robot. Throughout the project

https://orcid.org/0000-0001-5046-8558

https://orcid.org/0000-0003-2882-0142

https://orcid.org/0000-0002-8083-0770

Health-Care Assisting Technology, https://healthcat.eu

there was a focus on the involvement of the end-users

for identifying a use case that remains relevant in the

light of the demographic change and generating an

understanding of how a robot should be integrated

into existing workﬂows.

2 BACKGROUND

Given the shortage of staff and an increasing demand

for services, healthcare could be considered to be able

to beneﬁt from the implementation of robotic solu-

tions. However, while the expenditures for health-

care in OECD countries amount to around 10 % of

the GDP (OECD, 2014), it only plays a minor role

in the application of robotics. The majority of ap-

plications outside the industrial domain are related

to logistics where 10 times as many installations are

reported compared to medical robotics (International

Federation of Robotics, 2020).

A large variety of robots for applications within

healthcare can be found in the literature (see, e.g.,

(Holland et al., 2021)), ranging from topics such

as hygiene which got an increased focus during the

covid-19 pandemic where, e.g., mobile robots have

been equipped with UV-light emitting lamps for san-

itizing surfaces, to handling of samples or social

robots. While the development of robotic solutions

for less structured environments can be challenging,

mobile robots are for instance able to operate safely

and largely autonomously in environments shared

with humans which suggests that solutions for mo-

bile robots should be feasible to achieve. However,

Juel, W., Haarslev, F. and Bodenhagen, L.

Health-CAT: Development of a Mobile Robot for Assisting Caregivers.

DOI: 10.5220/0010840800003124

In Proceedings of the 17th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications (VISIGRAPP 2022) - Volume 2: HUCAPP, pages

171-178

ISBN: 978-989-758-555-5; ISSN: 2184-4321

 2022 by SCITEPRESS – Science and Technology Publications, Lda. All r ights reserved

171

Figure 1: Steps taken to come from focus areas to use case

formulation.

besides technical challenges, also various additional

challenges such as a negative reputation of robots,

workﬂows being altered, the fear of jobs to be taken

over by robots or unclear liability are considered im-

portant to be addressed (Cresswell et al., 2018; Radic

et al., 2019). In addition, ethical aspects related to the

consequences, such as the potential reduction of hu-

man contact or the handling of privacy, of introducing

robots are considered important for achieving accept-

ability (Stahl and Coeckelbergh, 2016).

The Health-CAT project focuses on the devel-

opment of a healthcare robot involving end users

throughout the process to develop a use case that is

both feasible but also relevant and where the integra-

tion of the robot into existing workﬂows has been a

integrated part of the design process. Ethical chal-

lenges have been taken into account during the selec-

tion of the use case to minimize the risk for a later

implementation of the robot to be inhibited by ethical

aspects and potential negative publicity.

In the following the overall approach is outlined

(section 3) followed by a needs analysis (section 4)

that culminates in the formulation of a speciﬁc use

case and a test of solution for this use case. Details of

the technical development are provided in section 5.

3 METHOD

When developing and implementing new technology,

it is essential to have an understanding of the end

users and the context of the developed solution. It en-

sures that the solution will fulﬁl an actual need rather

than, potentially driven by technology, formulating

new applications for existing technologies. In addi-

tion, it ensures a preparation of possible challenges of

the solution - for instance physical limitations, differ-

ent levels of competencies among users or unwilling-

ness to adopt new technology amongst others. Fur-

thermore, involving end users of the technology at an

early stage often pays off in the implementation pro-

cess, because the users get a sense of ownership to the

ﬁnal solution.

Health-CAT focuses on the general needs of the

healthcare sector and its users. We involved the users

(staff at German and Danish hospitals and care cen-

ters) during every step in the development process to

determine actual requirements for a robot prototype.

The feedback of the users was utilized in the subse-

quent steps in the development process. Thereby we

ensure to always meet the needs of the users. In the

following (section 4 the needs analysis, starting with

desk research and progressing via multiple mock-ups

to the ﬁnal prototype, which was performed in the

Health-CAT project will be described.

4 NEEDS ANALYSIS

The overarching goal in the Health-CAT project is to

investigate use cases in the healthcare sector where

robot technology can create value and help mitigat-

ing the challenges related to the demographic change.

We want to ensure that the quality of care does not

decrease in our future healthcare system that is chal-

lenged by to the expected issues arising from the soci-

etal challenges described in section 1. A needs analy-

sis was conducted based on broad involvement of pa-

tients and healthcare employees. The analysis con-

sist of desk research, ethnographic observations, in-

terviews and focus-group interviews. Initially desk

research was conducted to ﬁnd societal challenges

that later was further investigated through observa-

tions and interviews at different healthcare institu-

tions in Denmark and Germany. In total 20 inter-

views with patients and 62 interviews with employees

at healthcare institutions ranging from nurses, care-

givers, doctors, service- and kitchen staff was con-

ducted. The needs analyses attempts to:

• Identify focus areas where healthcare profession-

als have a need to be relieved in their daily work.

• Evaluate, how these workﬂows can be supported

by robot technology

• Map the barriers and challenges with implemen-

tation of robot technology.

Finally the needs analysis points towards several

scenarios where robot technology could play a role.

In order to conduct a qualiﬁed identiﬁcation of

needs, it was crucial to gather input from many differ-

ent types of employees in the healthcare sector so that

the needs analysis will represent needs across differ-

ent parts of the healthcare sector. The ﬁrst step in the

HUCAPP 2022 - 6th International Conference on Human Computer Interaction Theory and Applications

172

needs analysis was to identify existing potential prob-

lems within the healthcare sector in our case in Danish

hospitals and German nursing homes, see ﬁgure 1 for

an overview of our process. The context is analyzed

at three levels macro, meso and micro. The macro

and meso levels was studied by conducting literature

search and expert interviews with e.g. union repre-

sentatives. The output of researching these two levels

was a list of 10 potential focus areas within Health-

care that could potentially be supported by robot tech-

nology:

1. Hygiene: poor cleaning

2. Poor oral hygiene: pneumonia

3. Malnutrition and lack of follow-up

4. Overweight and obesity

5. Transporting equipment and patients

6. Inappropriate medication

7. Rehabilitation

8. Documentation

9. Cultural diversity

10. Measuring health parameters

These areas are then qualiﬁed by the project part-

ners as well as relevant hospital employees. After

conducting interviews, it was possible to map com-

mon focus areas of potential interest, which was then

agreed on by project partners, relevant stakeholders

and users. This ﬁnal qualiﬁcation was conducted in

a workshop, and resulted in a list where 6 of the ini-

tial focus areas got discarded and 4 was kept, these 4

focus areas was further discussed and detailed into 7

sub areas, see ﬁgure 2.

With these relevant focus areas identiﬁed, we fur-

ther wanted to validate them by interacting with rele-

vant employees and observing the work ﬂow within

the potential focus areas at both hospitals and care

centers.

4.1 Field Work

A ﬁeld work protocol including an interview guide

and an observation guide was created. The protocol

deﬁnes who to interact with, how many, where, and

how to document the interactions. The ﬁeld work

was conducted in a Danish hospital and in German

care centers, we identiﬁed speciﬁc workﬂows, user

needs, potentials and challenges, within the focus ar-

eas of interest. An observer’s ability to observe can

be inﬂuenced on whether the observer is in advance

familiar with the observed workﬂows or not. The

research team therefore included both a Healthcare

professional and an engineer with a technical back-

ground conducting the ﬁeld work. The ﬁeld work

process lasted 1.5 month in Germany and 2.5 month

in Denmark with a total of 63 people from relevant

roles being observed and interviewed. The intention

was to interview both employees and patients in hos-

pitals, as well as employees and residents in nursing

homes. However, several of the patients in the geri-

atric department at the hospital suffer from cognitive

impairment, which made it challenging to interview

the patients. As a consequence only two patients was

interviewed, see table 1 for more information about

the ﬁeld work.

The data received from the ﬁeld work was further

analyzed at a workshop and 5 focus areas was further

detailed in the form of concrete robotic use cases:

• Cleaning of wheelchairs, toilet chairs, beds, night-

stands

• Transportation of food

• Transportation of small equipment

• Transportation of large equipment

• Hand hygiene of patients and employees

4.2 Use Case Analysis

In the following the focus areas will be further intro-

duced and potential use cases are outlined. In the con-

ceptualization of the use cases ethical challenges in

regard to introduction of robot technology was highly

important and likewise was technical challenges, the

solutions had to be technically feasible with today’s

technology while at the same time not cause any kind

of ”harm” to the employees. This could be altering

their workﬂow excessively or creating technology that

could end up being a liability.

Area: Cleaning of Wheelchairs, Toilet Chairs,

Beds or Nightstands

During observations and interviews, it was clear that

the health carers do not consider cleaning as their pri-

mary task. Furthermore, there is not allocated any

space for cleaning and the consequence is cleaning

of equipment takes place in the hallways or in bath-

rooms. Another consequence is bad work postures,

as the employees do not have any tools available for

easing the cleaning process. Therefore, the employ-

ees have to reach all parts of the equipment manually,

including difﬁcult accessible parts like wheels or the

bottom of a nightstand and we observed that it was not

cleaned properly due to several of the above issues.

Potential Use Case: Automated Washing Machine

Transport to Automated Washing Machine: The em-

ployees use different types of equipment as part of

nursing tasks. Before or during use, the employee

Health-CAT: Development of a Mobile Robot for Assisting Caregivers

173

Figure 2: The 10 initial focus areas have been ﬁltered to 4 (top row), excluding those that have been considered to be infeasible,

and detailed further. This leads to 7 reworked focus areas (bottom row).

Table 1: Field work information.

Country Germany Denmark

Location of Observation Tabea, nursing home and UKSH, hospital SHS Aabenraa, hospital

Ward Orthopedics and Trauma Surgery Geriatric

# of employee observations 14 28

# of employee interviews 14 27

# of interviews with patients or residents 19 2

calls a robot to place itself right outside the pa-

tient’s room. When the employee is done using the

equipment, the employee places the equipment on

the robot, and the robot transports the equipment to

the automated washing machine. After cleaning, a

robot transports the equipment back to the designated

equipment space.

The equipment cleaning: The robot automatically

places the equipment on a conveyor belt. The belt will

transport the equipment into the automated washing

machine, where it is automatically cleaned and dried.

Area: Transportation of Food

Transportation of food is a time consuming task due

to registration of intake, ordering and serving of food

and it requires a lot of walking. In the morning the

kitchen staff have to prepare and transport food to

all wards at the same time which creates a crowded

kitchen area. At noon and in the evening the food is

transported to the ward by the kitchen staff and the

health carers helps the patients eat. The health carers

and the kitchen staff do not coordinate their tasks so

the patients’ meals is sometimes stranded in the hall-

way for a while before the health carers have time to

serve food to patients. Furthermore, the registration

of patients’ food intake is often a challenging tasks

since it is time consuming and meals are brought to

the patient by one employee and picked up by another.

Use Case: Food Transport and Registration

Transportation of Food: Health carers order meals

and kitchen staff loads a robot (with a combined

heating- and cooling cabinet) with trays of meals for

a speciﬁc ward. The robot registers food for the each

patient or resident and drives to the speciﬁc ward. The

health carers pick up and serve trays for patients or

residents one by one.

Registration of food: When the patients or residents

are ﬁnished eating, the health carers pick up the trays

and load the robot. The robot registers food intake

(calories) for each patient or resident and uploads data

automatically (e.g. in Cosmic or Dankost) and returns

to the kitchen.

Area: Transportation of Small Equipment

As part of care-giving, the health carers need small

equipment, such as venﬂons, bandages etc. The

health carers have to leave the patients, walk to the

storage, pick up small equipment, walk back to the

patient and continue their original task. If equipment

is damaged the procedure is repeated. The health car-

ers cannot bring spare equipment, due to a policy to

minimize waste. Collection of small equipment is an

issue since it is a time consuming task that requires

walking up to 10 km a day. Patients and residents of-

ten experience the situation as both time consuming

and confusing since the staff interrupts their routine.

HUCAPP 2022 - 6th International Conference on Human Computer Interaction Theory and Applications

174

If small equipment has been brought along to a pa-

tient room without being used, it has to be discarded

due to risk of infection. One speciﬁc observation that

illuminates this issue was patch being fetched which

took up to 5 minutes.

Potential Use Case: Mobile Dispenser

Each ward has a number of robots with a dispenser

mounted that is ﬁlled with small equipment, speciﬁc

to the ward’s requests. The reﬁlling of all robots is

done by e.g. a service assistant in the basement. Af-

ter reﬁlling, the robots drive to the ward and wait in

allocated spaces in the hallway. When equipment is

needed, the health carers visit the robot or call it to the

room via. a call system installed on the ward. When

the robot is low on inventory, it automatically calls a

another robot to replace it. The empty robot drives to

the storage, where it will be reﬁlled and cleaned by a

service assistant.

Area: Transportation of Large Equipment

When equipment is needed for a nursing task, the

health carers start searching in hallways before look-

ing in the allocated space since equipment is often not

positioned at the allocated space. In German nurs-

ing homes, we observed that health carers interrupt

their colleagues just to locate equipment. Once the

equipment is located, and if not in use, the health car-

ers transport the equipment manually. Due to lack of

time, equipment is often left without proper cleaning,

which mean the next person have to clean it before

using it.

Potential Use Case: Equipment Caddy

When large equipment is needed, the healthcarers call

a robot via. a tablet, to bring required item to the de-

sired area. The healthcarers can also locate equip-

ment on the tablet, and pick it up and manually. If

equipment is already in use, it will display a waiting

list. After use, equipment is loaded on the robot. The

robot transports the equipment to the next user or to

allocated free space. Thereby healthcarers can con-

tinue there routines and be more efﬁcient.

Area: Hand Hygiene of Patients and Employees

As part of the morning routine in hospitals in both

Germany and Denmark, the patients get a cloth or

a wet tissue. It is up to the patients to assess what

to clean. From our observations, most patients for-

get to clean their hands thoroughly. During the day

there was little or no focus on cleaning of patient’s

hands, neither before meals and after toilet use. In

nursing homes, the residents are responsible for their

own hand hygiene, but at the geriatric ward the health-

carers had to ensure proper hand hygiene of weak

patients. The healthcarers disinfect their own hands

between each patient visit to avoid the spread of in-

fections, however, it is also important to clean hands

thoroughly with water and soap. We observed that

the healthcarers tend to disinfect their hands instead

of washing, as this is faster.

Potential Use Case: Mobile Hand Disinfection

A mobile robot drives into the patients’ rooms and

offers disinfection of hands. The patients place their

hands in robot and disinfection is sprayed directly on

patients’ hands. An important feature of the robot is

that it motivates/nudge the patients to disinfect hands.

These 5 early use cases were introduced to a fo-

cus group in a Danish hospital and a German nursing

home. The group of Healthcare professionals quali-

ﬁed and ranked the use cases with regard to impor-

tance. Based on the inputs from the focus group, two

use cases were chosen. Use case 1 dispensing of small

equipment and use case 2 tracking of large equipment.

Use case 2 is a non robotic use case consisting of a

Bluetooth tracking system placed on all large equip-

ment in order for healthcarers to track it. In the fol-

lowing only use case 1 will be focused on.

Use Case 1 Flow: Mobile Robot Dispensing Small

Equipment

The goal of the use case is to support the care staff

by giving them a storage room on wheels, that is al-

ways where it’s needed with needed equipment. All

the necessary equipment are sterile and/or clean, until

they are pulled from the dispenser. The ﬂow of the

use case is outlined below:

1. The robot is ﬁlled with the required items in the

central storage.

2. The robot drives to its designated department and

positions itself centrally in the corridor.

3. The robot can be called to the entrance of a patient

room using call buttons installed in each room.

4. When the robot arrives the nurse is informed and

the necessary items from the robot can be taken.

In the following section, a description of the tests

process of the Health-CAT robot is given.

4.3 The Test Process

Three rounds of tests in the geriatric ward at SHS

Aabenraa have been conducted. The testing consisted

of two, one-day long mock-up tests and ﬁnal test with

the developed prototype that spanned a full week.

The goal is to verify in the clinical setting that the

robot complies with the use case and to document

the effects of its use. With each test a new iteration

of the robot has been developed, incorporating ﬁnd-

ings of the earlier tests and progressing from a simple

mock-up to an actuated robot. We started with some-

thing very simple and then in each test iteration, we

added more technical details. Thereby We avoided the

Health-CAT: Development of a Mobile Robot for Assisting Caregivers

175

Figure 3: Timeline of the development and test cycles in Health-CAT with three different prototypes: a passiv mockup (left),

a mockup with relevant sensor capabilities and user interfaces (center), and the ﬁnal, actuated prototype.

robot getting a negative reputation early and the work-

ﬂow being altered drastically before we knew that we

could technically solve the actual use case. The hos-

pital staff reported, that the existing robot technology

deployed there often also was a nuisance. The way it

behaved was often unclear and we wanted to ensure

the opposite in Health-CAT. Therefore we recorded

data about the environment to be fed into the vision

and navigation system. The hospital employees also

reported fear of losing their jobs to robots. We in-

cluded them after each test iteration. This made the

users feel ownership of the robot since they were part

of the development process and we could continu-

ously explain and show them that this robot was a tool

for them and not a replacement. On ﬁgure 3 a time-

line of the overall test process in Health-CAT can be

seen. In the following, a general description of the

test process is listed.

4.3.1 Mock-up 1 and 2

The ﬁrst two mock-up tests were designed as non

robotic and was situated at the geriatric ward at SHS

Aabenraa. A nurse was pushing a drawer on wheels

and together they functioned as the robot, see ﬁg-

ure 3. In mock-up test 1 we did some very basic

and early testing, we installed Bluetooth buttons in

half of the patient rooms at the ward. These buttons

was connected to a similar Bluetooth button outside

the room that was visually on the hallway, when the

button inside the patient room was pushed, the but-

ton outside lighted up with green light, and the nurse

acting as the robots actuators pushed the mock-up to

the speciﬁc room. The testing was done during the

morning round, afternoon round and during the night

shift. The ﬂow of the mock-up test basically follow

the earlier described use case; when a nurse enters

a patient room, the installed button was clicked and

the button outside the patient room lighted green, re-

sembling that the robot was notiﬁed that this speciﬁc

room needed equipment. The nurse controlling the

”robot” would then drag the drawer to the speciﬁc

room that lighted green. This early mock-up test was

primarily conducted to get a general understanding of

how this robot solution could be build and how it po-

tentially would affect the workﬂow. Interviews with

the staff involved in the mock-test was conducted and

data from this was feed into the following testing.

Mock-up test 2 is an extension of the ﬁrst test, see

ﬁgure 3. We performed the test on the full ward in-

cluding every patient room. We automated more fea-

tures in the process such as the calling system. We

added a feedback mechanism on the button so the

nurses could see that the robot is on its way and that

the robot has arrived outside the patient room. We

added visual feedback (LED strip around the robot)

that indicates the internal state of the robot, blue when

it drives and green when it can be accessed by a nurse.

We added later a laser scanner and a camera hub on

top of the robot so we could gain data about the real

environment to use for improving and developing our

navigation system. Again, interviews with the staff

involved in the mock-test was conducted and data

from this was feed into development of the prototype.

4.3.2 Prototype Test

The ﬁnal test was also performed at the geriatric ward

at SHS Aabenraa. The Robot prototype shown on ﬁg-

ure 4 was tested for 1 week during every shift of the

day. In this test we focused on measuring and ob-

serving the general work processes and compare it to

data collected during our ﬁeld work. Some of the core

ﬁndings are listed below:

• Change in Nurse Walking Distance: One of the

core ﬁndings was that the nurses took way less

HUCAPP 2022 - 6th International Conference on Human Computer Interaction Theory and Applications

176

steps during a work day working with the robot

vs not having the robot.

• Change in Used Time Picking up Equipment:

There was a large time save in picking up equip-

ment and they spend less time locating equipment.

• The General Flow of Their Work with a Robot:

Several nurses reported an increase of empower-

ment and that the environment was less hectic in

general since the robot was always at the right

place at the right time. The risk of spreading in-

fections was considered reduced using the robot.

• Change in Disturbances from Other Employ-

ees: The nurses were disturbed less by each other

and therefore more time was spend with the pa-

tients.

• Use of Freed up Time: We measured that the

nurses spend more time in the patient rooms.

5 ROBOT PROTOTYPE

The robotic system consists of two major parts – a

robot prototype and a call system – which both con-

sists of various subcomponents. The prototype con-

sists of a mobile base, a cabinet, a GUI head, and a

handle for manual movement of the robot in emer-

gency situations (see ﬁg. 4). The base has two ac-

tuated differential drive wheels in the middle of the

base, and two safety lasers in opposite corners. The

cabinet is made by Standard Systemer

and is ap-

proved for clinical use. The two narrow sides of the

cabinet (front and back of the robot) each has a Re-

alSense D415

RGB-D sensor, which are used for de-

tecting people in the hospital hallways. The GUI head

has a screen which is used to show the control inter-

face for the robot. It also contains processing units

for sensor processing and robot control. The handle

is attached to the brake system and disengages the

brakes when pulled, which allows for manual move-

ment of the robot. The robotic prototype is based on

ROS (Quigley et al., 2009) and uses the ROS naviga-

tion stack (Marder-Eppstein et al., 2010) for navigat-

ing the hallways of the hospital.

The call system consists of WiFi enabled buttons

which are placed in each patient room. Once a button

is pressed it transmits a unique ID to the robot through

the MQTT messaging protocol, and the room corre-

sponding to that ID is placed in a queue. An RGB

LED on the button provides feedback to the user: blue

if that room is ﬁrst in line and yellow otherwise.

When the queue is not empty, a navigation goal

in front of the door to the patient room at the top of

https://standardsystem.dk/

https://www.intelrealsense.com/depth-camera-d415/

Figure 4: Final prototype that has been evaluated.

the queue is sent to the navigation stack. While nav-

igating to the room the robot detects and tracks peo-

ple in the hallway (Juel et al., 2020), and the tracked

people are used as input to a social navigation behav-

ior (Haarslev et al., 2021). The tracker is based on

the CenterNet (Zhou et al., 2019) human pose estima-

tion network and the Deep SORT (Wojke et al., 2017)

object tracker. The system works by feeding images

from the two cameras to the human pose estimator,

which estimates bounding boxes and joint locations

for each person in the view. The joints are projected to

3D, and the torso center and facing direction of each

person is estimated using the 3D shoulder joints. The

pose is transformed to the map frame using localiza-

tion data from the navigation stack. The transforma-

tion to the map frame serves two purposes: it uni-

ﬁes the detection frames from the two cameras, and it

eliminates the problem of egomotion when tracking.

The transformed detections are used as input to a

modiﬁed Deep SORT algorithm. Deep SORT tracks

objects based on deep appearance features found by

inputting images of each detected person to a Siamese

CNN trained with triplet loss (Schroff et al., 2015). It

incorporates a Kalman ﬁlter , which predicts the mo-

tion of the tracked people. A cost matrix is created

using the distance between the deep appearance fea-

tures of the tracks and the new detections. The cost

matrix is gated by setting costs to inﬁnity for detec-

tion/track pairs where the detection is too far away

from the predicted position of the track. Association

is done by solving the cost matrix using the Hungar-

ian algorithm. The Deep SORT algorithm is modiﬁed

by changing the state space of the Kalman ﬁlter from

image coordinates to 3D map frame coordinates.

The tracked people are input to a social navigation

method, which is integrated with the ROS navigation

stack. The method works by creating costs represent-

ing social spaces (Lindner and Eschenbach, 2011) at

estimated collision points between the robot and the

Health-CAT: Development of a Mobile Robot for Assisting Caregivers

177

tracked people. The collision points are estimated us-

ing the tracked motion of the people, and the motion

of the robot. By projecting costs to future positions of

the people the robot preemptively avoids entering the

personal space of the people, resulting in a socially

acceptable navigation behavior.

6 CONCLUSIONS

We have described the process of developing a robot

to solve a use case within healthcare. The goal of

the Health-CAT project is to ensure that the quality of

care does not decrease despite the healthcare system

being challenged due to the expected issues arising

from the societal challenges. This process is based

on a needs analysis using literature search and ethno-

graphic studies to identify relevant focus areas. From

these needs, seven use cases were formulated with

ethical and technical challenges, concerning the im-

plementation and introduction of robot technology, as

a focus point. We ended up choosing one use case, the

transportation of small equipment, where the robotic

system consists of two major parts - a mobile robot

prototype and a call system enabled staff to call the

robot to any patient room.

This robot concept was tested in three iterations.

The ﬁrst two iterations were highly focused on iden-

tifying the requirements, issues, and beneﬁts of the

robot. The last test involved the actual robot proto-

type in a one-week-long integration of it at the hos-

pital ward. The Health-CAT robot showed that the

daily work life for nurses improved. They walked

less, which decreased the physical stress, and experi-

enced an increase of the time spent with the patients.

Furthermore, nurses reported an increase in perceived

empowerment and that their work environment was

less hectic in general.

ACKNOWLEDGEMENTS

This research was supported by the project Health-

CAT, funded by the European Fund for regional de-

velopment.

Furthermore we would like to thank the involved

staff at the Hospital Sønderjylland, UKSH L

ubeck

and AWO Haus am M

uhlenteich Lensahn for their

valuable input and support.

REFERENCES

Bodenhagen, L., Suvei, S.-D., Juel, W. K., Brander, E.,

and Kr

uger, N. (2019). Robot technology for future

welfare: Meeting upcoming societal challenges – an

outlook with offset in the development in scandinavia.

Health and Technology.

Cresswell, K., Cunningham-Burley, S., Sheikh, A., et al.

(2018). Health care robotics: qualitative exploration

of key challenges and future directions. Journal of

medical Internet research, 20(7).

Drennan, V. M. and Ross, F. (2019). Global nurse shortages:

The facts, the impact and action for change. British

medical bulletin, 130(1):25–37.

Falatah, R. (2021). The impact of the coronavirus disease

(covid-19) pandemic on nurses’ turnover intention:

An integrative review. Nursing Reports, 11(4):787–

810.

Haarslev, F., Juel, W. K., Kollakidou, A., Kr

uger, N., and

Bodenhagen, L. (2021). Context-aware social robot

navigation. In ICINCO.

Holland, J., Kingston, L., McCarthy, C., Armstrong, E.,

O’Dwyer, P., Merz, F., and McConnell, M. (2021).

Service robots in the healthcare sector. Robotics,

10(1).

International Federation of Robotics (2020). World

Robotics (Summary). VDMA, Frankfurt am Main,

Germany.

Juel, W. K., Haarslev, F., Kr

uger, N., and Bodenhagen, L.

(2020). An integrated object detection and tracking

framework for mobile robots. In ICINCO.

Lindner, F. and Eschenbach, C. (2011). Towards a formal-

ization of social spaces for socially aware robots. In

Spatial Information Theory, pages 283–303.

Marder-Eppstein, E., Berger, E., Foote, T., Gerkey, B. P.,

and Konolige, K. (2010). The ofﬁce marathon: Ro-

bust navigation in an indoor ofﬁce environment. 2010

IEEE Int. Conf. on Robotics and Automation.

OECD (2014). Health expenditure indicators.

Quigley, M., Gerkey, B., Conley, K., Faust, J., Foote, T.,

Leibs, J., Berger, E., Wheeler, R., and Ng, A. (2009).

Ros: an open-source robot operating system. In Proc.

of the IEEE Intl. Conf. on Robotics and Automation;

Workshop on Open Source Robotics.

Radic, M., Vosen, A., and Graf, B. (2019). Use of robotics

in the german healthcare sector. In Social Robotics,

Cham. Springer International Publishing.

Schroff, F., Kalenichenko, D., and Philbin, J. (2015).

Facenet: A uniﬁed embedding for face recognition

and clustering. 2015 IEEE Conference on Computer

Vision and Pattern Recognition.

Stahl, B. C. and Coeckelbergh, M. (2016). Ethics of health-

care robotics: Towards responsible research and inno-

vation. Robotics and Autonomous Systems, 86.

Wojke, N., Bewley, A., and Paulus, D. (2017). Simple on-

line and realtime tracking with a deep association met-

ric. In 2017 IEEE Int. Conf. on Image Processing.

Zhou, X., Wang, D., and Kr

ahenb

uhl, P. (2019). Objects as

points. ArXiv, abs/1904.07850.

HUCAPP 2022 - 6th International Conference on Human Computer Interaction Theory and Applications

178