Substantiation of the Required World Model Configuration for

Autonomous Mobile Medical Service Robots

Dmitry Rogatkin, Liudmila Lapaeva and Denis Lapitan

Moscow Regional Research and Clinical Institute “MONIKI” named after M.F.Vladimirskiy, 61/2 Shepkina str., 129110,

Moscow, Russian Federation

Keywords: Medicine, Service, Mobile Robot, World Model, Concept, Requirements, Artificial Intelligence.

Abstract: Development of smart mobile service medical robots is the actual and interesting problem in the modern

biomedical instrumentation. This study is aimed at development of the theoretical basis of the modern

medical robotics, in particular, of the theory of autonomous mobile service medical robots. Based on the

expert analysis of typical scenarios of behaviour of service medical robots in clinics, the necessity of

development of reasoning-based robots with built-in world model was substantiated. It was shown, that the

development of the world model should be the key start point of all the process of development and design

of such robots. General technical requirements for the robot’s world model, for its composition and

functionality, were formulated. It was shown also, that the world model of such robots must maintain a

database of precedents of various events and must work taking into account both embedded instincts of the

robots and general rules of the World.

1 INTRODUCTION

One of the sections of the contemporary robotics is

medical robots, and in particular – autonomous

mobile service medical robots (AMSMR) (Butter et

al., 2008; Wang et al., 2006; Rogatkin et al., 2013).

The first AMSMR named ”AMS-car” appeared in

the middle 1970s in the Fairfax hospital (USA) to

move containers with food trays for patients. From

this moment, the idea of introduction of AMSMRs

into the health care system has been flying up in the

air, because there was a great need in medicine for

diligent and exact implementation of plenty of

different routine transport, information, and other

auxiliary service procedures. Today, there are

hundreds of articles in journals as well as in Internet

devoted to different AMSMR including one of the

first projects – «HelpMate» – the project of the

transport service robot for delivery of different

drugs, analyses, materials, and documents inside a

hospital (Evans, 1994). Also, it is well-known the

Russian robot «R-Bot» for telepresence in hospitals

(http://www.plasticpals.com/?p=26770); the project

of remote-presence medical robots «RP-7» and «RP-

8» – information robot-manipulators allowing oral

and visual communication of physician with a

patient from any point of the Globe (Wang et al.,

2006), and a number of other similar projects. The

last generation of the “RP” robot family – robots of

the “RP-VITA” series – was approved in 2013 by

FDA, and its invasion into clinics was really started

(see in http://www.intouchhealth.com/products-and-

services/products). Robots “RP-VITA” can move

themselves round the clinic territory, find up the

room of interest and a patient’s bed without human

participation. A doctor must only point out the aim

for the robot by pressing the corresponding button.

However, the majority of publications about all

these projects in different conference proceedings as

well as in Internet have neither scientific nor

engineer or technical features. They are of only

advertising (demonstrational) character in the vast

majority of the publications. They describe different

robot resources and demonstrative examples of its

usage without detailed and systemic technical

description of concrete scenarios and algorithms of

the robots’ behaviour in clinics, without their target

functions, without general fundamental regularities

and (or) scientific background for development and

design of both hardware and software for such

robots. Meanwhile, in reality, AMSMR is an

automatic, electronic, and mechanical medical

equipment. Therefore, both the general system of

medico-technical requirements according to existing

Rogatkin, D., Lapaeva, L. and Lapitan, D.

Substantiation of the Required World Model Conﬁguration for Autonomous Mobile Medical Service Robots.

DOI: 10.5220/0005650801250130

In Proceedings of the 9th International Joint Conference on Biomedical Engineering Systems and Technologies (BIOSTEC 2016) - Volume 1: BIODEVICES, pages 125-130

ISBN: 978-989-758-170-0

125

standards and specialized demands, for example, for

design of software exclusively specific for medical

applications, in some degree or other, must be

applied to AMSMR. Thus, there is an obvious gap in

the engineer theoretical background for AMSMR.

Undoubtedly, during past several years, the

theory of service robots for different fields of human

activity including medicine became already an

object of investigations (Fiorini et al., 2013;

Rogatkin et al., 2013; Towle et al., 2014), though

quite recently the majority of classical monographs

on robotics (Young, 1979, for example) didn’t refer

to such studies. Voluminous articles and books on

this subject described particular problems of

construction of mechanical manipulators and (or)

clamps, sensors, elements of mobile parts of the

robot, and so on, but had none of the general

theoretic problems of software elaboration for

autonomous mobile service robots, especially for

medicine. Only since recently, a special attention is

paid to fundamental scientific mechanisms of the

software development for service robots (Iborra et

al., 2009; Somby et al., 2009; Towle et al., 2014).

But, in the first place, it still concerns the design of

the specialized software (ROS, Microsoft Robotics

Studio, Skilligent) for developers and development

of additional languages for high level programming

in robotics (LABVIEW Robotics Module, for

example). Until now, really smart AMSMR don’t

exist in our hospitals due to the complex problem of

creation of intelligent machines. Today, there aren’t

general methodical manuals and rules for engineers

how to create them. What should be done as the first

step on this way? Should we create a hardware or a

software first? In fact, this situation is preventing the

worldwide adoption of robots as efficient service

machines (Mastrogiovanni and Chong, 2013). For

overcoming this situation, the real integration

between techniques of artificial intelligence and a

practical robotics should be reached.

The main objective of our theoretical study is the

investigation and elaboration of general scientific

approaches to the formal engineer description of the

problem of AMSMR projecting and development, in

particular – a formulation of scientifically grounded

demands to the software for AMSMR. Firstly, in the

section 2 of the present article, the basic concept

(scenario) of AMSMR functioning in clinics is

considered. In robotics research, in the general case,

a formation of a scenario of a robot behaviour is not

a novel clime. But for AMSMR we don’t know such

concept and results, which have been published and

accepted by all. Secondly, we analyze the results of

the proposed concept and formulate a number of its

consequences. The main one - the generalized task

of clinical AMSMR functioning is a finite task,

limited (restricted) both in physical building space

and in a space of possible decisions and actions, and

is of the closed type, i.е. all possible situations and a

spectrum of various methods of their resolution can

be exhaustively described in the closed form.

Another important consequence – only smart,

intelligent AMSMR can follow the scenario in a full

scale. So, AMSMR should have in its “brain” a full-

scale World Model (WM), containing all necessary

information about each object as well as about each

possible situation in the robot’s World. Formation of

the internal WM of the “reasoning-based” AMSMR

as the key element of its software is discussed in the

section 3. Needed WM configuration and its

functionality are substantiated there, as well. All

these gives the answer on the question about the

order of the AMSMR development and design.

2 CONCEPT OF AMSMR

FUNCTIONING IN HOSPITALS

At the initial stage of our study, we interviewed

physicians and patients of different departments of

the multi-clinical and multi-functional medical

research centre “MONIKI” and received the verbal

description of the most prospective AMSMR

applications in the health care system (from the

point of view of both physicians and patients). Then,

all results of these interviews were classified,

generalized, and analyzed by the method of the

expert assessment. On the basis of the outcome, in

addition to the telepresence (remote-presence)

function, which is often mentioned in publications as

the main useful function for AMSMR, the following

important and prospective tasks were also

determined which are very interesting to solve and

which can be solved with the help of AMSMR in

real clinical practice:

- interactive lecture-excursion (10-15 minutes)

for patients at the moment of admission to the

hospital about daily routine in it, about building plan

(wards, shower-bath, dining-room, etc.), therapy

regimen, feeding, walking in the air, etc.;

- information of patients about the time of

diagnostic and (or) treatment procedures, repeated

reminding about it 15-20 minutes before the fixed

time in the concrete day;

- searching within the clinics for a given patient,

physician, or a nurse to inform them about concrete

instructions, problems, including the urgent ones;

BIODEVICES 2016 - 9th International Conference on Biomedical Electronics and Devices

126

- delivery of drugs and medical materials to

patients in their wards;

- delivery of food to the bed patients, removal of

waste and scrap products of them;

- listening (recording) to requirements and

complaints of patients and passing these information

to medical personnel;

- a series of other similar informational,

searching, and (or) transport tasks.

These functional duties of AMSMR (service and

assistant functions) appeared to be common for the

majority of departments of the multi-functional

medical centre, and, consequently, they are typical

to every medical institution of the health care

system. Moreover, it becomes clear from the list

above mentioned, that the vast majority of

prospective scenarios of AMSMR behaviour is

associated with a transport as well as with

information-searching tasks within the limits of the

room-type territory. Shortly speaking, AMSMRs are

needed in hospitals to implement quite compact and

rather unified set of simple instructions, such as: to

implement different errands associated with seeking

an object of interest, interaction with different

objects, delivery of objects, receiving and storage

information from people, information processing

and transmission to other people.

It makes the generalized task of AMSMR

functioning a finite one, limited (restricted) both in

physical building space and in the space of possible

decisions and actions, and of the closed type, i.е. all

possible situations and a spectrum of methods of

their resolution can be described exhaustively in the

closed form. It allows us to create an abstracted and

strictly formalized description of the generalized

concept of AMSMR functioning in hospitals (Fig.1):

All tasks should be resolved by AMSMR

algorithmically at the expense of its moving through

the room-type territory to search for and then to

contact with external objects (physician, patient,

medical equipment, furniture, etc.), which all are

programmed inside its WM as a set of variables for

identification and communication. The robot moves

around a clinic in order to solve the concrete task Т

which is put before it by an external operator

(physician or a patient). Various situations S

rise up

during the robot motion including urgent ones (for

example, a primary passage is blocked by an

obstacle) which should be solved by AMSMR itself,

without any operator participations. Also, AMSMR

must be able to put and to solve intermediate

situational tasks and subtasks t

, to create a

situational behavioural strategy Str

, as well as to

formulate its own situational target functions Q

with special particular purposes (going round the

obstacle, accumulator recharging, etc.). And only in

the absolute insoluble situation the robot should stop

completely and give an alarm signal of calamity

(feebleness).

Figure 1: Generalized concept of AMSMR functioning in

hospitals.

This general concept shows, that AMSMR must

operate with various objects and terms (territory,

room, passage, obstacle) in a complex environment,

must “understand” both the regularities of mutual

positions of these objects and possible changes in

their locations in the space of clinics, as well as must

take into account the logic and the history of such

changes, providing a semantically meaningful

interpretation of all these object interactions and a

motion within a scene, especially if one of the

participants within the scene is a man. Therefore,

only smart, intelligent AMSMR can follow the

scenario in a full scale; and the key element of

AMSMR should be the full-scale WM in its “brain”,

containing all necessary information about each

object as well as about each possible situation in the

robot’s World.

Today a number of approaches is known to

formation of such WM for autonomous mobile

robots (Roth et al., 2003; Burgard et al., 2008;

Elfring et al., 2013). However, the overwhelming

majority of such WM are de-facto a simple

geometrical description of the territory map and

objects location in the indoor environment

(Kułakowski et al., 2010). They describe statistical

Substantiation of the Required World Model Conﬁguration for Autonomous Mobile Medical Service Robots

127

panorama (situation) of the territory, but tell nothing

of the history, logic, and the possible ways of this

situation development with time. Such WM are,

definitely, needed, but it is only one of the elements

of the really required, full-scale, effective and

workable WM for AMSMR.

Some publications (Burgard et al., 2008; Elfring

et al., 2013)) indicate that a set of special semantic

attributes of the objects describing sense (semantic)

associations between all objects are needed to create

WM of full workable. To our opinion, WM

presented by B. Coltin et al (Coltin et al., 2010), is

the most similar to the real required WM for “smart”

AMSMR. It offers a modelling approach that

differentiates among the motion models of different

objects in terms of their dynamics, namely the static

landmarks (e.g., goal posts, lines, corners), the

passive moving ball, and the controlled moving

robots, both teammates and adversaries. In their

article WM is a tuple {O, X, S, M, H, U}, where: O

is the set of labels of objects in the World, X is the

set of possible object states, S is the set of possible

sensor readings, M is the set of models of the

objects, H: MxO→X is a hypothesis function that

returns the current state of an object, U: MxOxS→M

is the model update function. Perception of the state

of the World by the robot with such WM is the most

complete today, but, unfortunately, this WM is

restricted by the objects of the football playing

situation only. It is based on the concrete football-

game data of robot’s sensors (tuple S), and,

therefore, it is not widely applicable for the more

general types of robots, for AMSMR, for example.

3 GENERAL CONCEPT OF THE

FULL-SCALE WM FOR

AMSMR

In our study, the general concept of the object-

oriented WM is being elaborated for AMSMR. It is

based on the abstraction from the concrete hardware

of AMSMR and underlain by an assumption that all

sensor organs of the ideal AMSMR as well as its

object-recognition system allow separating from the

information entering AMSMR sensors all necessary

object identification signs installed within WM

about the objects being observed. This, in its turn,

enables exact identification of all objects by the

robot, i.e. AMSMR is assumed to have ideal sensors

and perfect object-recognition systems. How do the

sense organs and the object-recognition systems

work – it is a question beyond the concept under

consideration, because only general fundamental

principles of the “high level” WM functioning are an

important subject of our investigation without

detailing of how the sensor system and the object-

recognition one works.

To make it simpler, the first stage of our study

was restricted by a static description of the WM of

2D space. The general regularities of situational

location of the robot and external objects in

environment as well as general rules of their

possible associations and an interrelationship were

studied. It was shown that such a task for AMSMR

requires formation of reasoning in the robot’s

“mind”. The robot is to act on the basis of situational

cases and extrapolation of a situation development.

According to classification, such a robot is attributed

to intellectual dynamic systems acting situationally

in multi-agent surroundings based on adaptive logic

reasoning and precedents (reasoning-based adaptive

AMSMR). Therefore, formulizing causal rules of the

World within the frames of the logic of the 1

order

predicates calculation is necessary for the “smart”

robot WM, with regard to the time and modality

factors (temporal and modal logics). In addition, a

list of environmental objects and its properties, as

well as the knowledge of the intra-environment

behaviour of them, definition of the dictionary of

predicates, functions, constants, and a database of

typical situational tasks in the environment are

required. The block of prognosis and planning of

situation development should be one of the blocks of

WM functionalizing, as well. Shortly, WM of

AMSMR, in the general case, should involve not

only a structuralized set of categories (objects) of the

environment and their characteristics but also should

involve the means of description of the possible

semantic associations between all objects, their

possible interrelations, rules of such interrelations.

All these determine the needed structure of WM

(Table 1). It must include, in addition to models of

external objects (rooms, furniture, medical

equipment, etc.), some general non-object and

procedural (functional, semantic, etc.) terms: way

(distance, trajectory), direction, motion, danger, to

be present, to search for, to find, to implement,

cause, quantity, etc. It distinguishes our full-scale

WM from WM of other authors. In our WM, the

general relations and kindred associations between

objects are also determined: faster/slower,

part/whole, friendly/alien, etc. Based on these terms,

the basic rules and relations in the World are

formulated.

As it was shown, all rules (regulations) existing

in WM can be conditionally divided in 2 categories.

BIODEVICES 2016 - 9th International Conference on Biomedical Electronics and Devices

128

Table 1: World model composition.

General World regulations and rules independent of

the concrete robot environment (general philosophic

regulations of the world) can be attributed to the first

category. To the second category attributed are laws

and rules currently in force for the given concrete

environment (clinic) and for the given concrete robot

design. All these rules may be formulated as the

sense logical statements (predicates) as follows:

1. There is a closed-type set of non-crossing

concepts (terms) in the world: object, property, rule,

quantity, cause, consequence, time, space, etc.

2. Any task or situation in the World can be

described using these concepts.

3. In the World there are many external objects,

characterized by different sets of its properties,

which can be detected by sensors (sense organs).

4. Objects of the World can be simple and

complex (consisting of other objects).

5. There is a basic set of strong (non-changing)

rules and restrictions in the World.

6. The life goes on in the environment

determined by the territory map.

7. Each existing object has its own location in the

environment.

8. There is a single location for each object in the

given moment of time.

9. Two different objects cannot be located in one

place in the given moment of time.

10. It is impossible to return back in time.

Such system of knowledge and rules enables

description of WM in the most natural way for the

human beings. In total, the set of the object models

and the world rules must allow description of all

typical situations (scenes) possible in the chosen

environment. The sequence of scene sets organized

in time must allow modelling possible situations,

predicting and planning the robot behaviour.

The AMSMR model itself is, also, an important

element of WM. It allows a robot to determine itself

as an element (object) of the World able to move,

communicate with other objects, and simultaneously

(in a certain way) as a special “subject” of the

World, which means impossibility to ordinary

communicate with itself (with the object “I am”) like

with other objects. For example, it is impossible to

search for itself. The latter requires the presence in

WM several models of the robot’s “physical” and

“analytical” opportunities. The important and

necessary element characterizing WM of AMSMR

is an interface of a target designation, i.е. the “robot-

physician” interface, which allows to realize all

information interchanges with physicians and to

perceive from them all orders for implementation.

However, the given model tells nothing yet about

motives of the robot activity. It isn’t locked in the

sense of the absence of the tasks for the robot. For

efficient AMSMR functioning, a target designation

or other motive for its activity is required. But it is

well-known, that the task of the target designation

cannot be solved today in a frame of the artificial

intelligence theory (Russel et al., 1995, 2003).

Therefore, it has to be solved additionally by

formulation or determination of the external tasks

and external motivations for the robot activity

(motivation from “outside”). In this case, all

motivations should be ranged by priorities to

exclude indefiniteness in the case of multitask

situations. In our project, we offer the following

postulate as a rule for any action motivation: the

robot acts according to the “instincts” programmed

within it. The example of instincts (with priority

diminishing) may be as follows:

- instinct of saving the human being (avoiding

collision, warning about danger)

- instinct of self-preservation (avoiding collision,

falling down, overheating, etc.);

- instinct of hunger (taking urgent steps to satisfy

the hunger when batteries are discharged);

- fear of the unknown situations (avoiding

situation non-modulated within the given WM);

- instinct of execution of the target designation:

to act in the case of the target designation from a

WorldModel



Model of objects

Modelofenvironment:

‐territory(rooms,passages)

‐items(furniture,devices)

‐otherdwellers(people,robots)

‐feedingzone(theplaceforrecharging

accumulators)

‐……..

Modelofitselfasanobject:

‐overalldimensions

‐sensorsofexternalobjects

‐sensor(s)ofadanger

‐sensorsofamotionvelocity



Modeloftargetdesignator

(robot‐physicianinterface)

General non-object (subjective) and procedural terms:

way,trajectory,direction,motion,turninground,tostop,danger,tobepresent,tofind,

objective,toreach,time,cause,quantity……

Typical situations (scenes) in the given environment:

‐tostandwithoutanymotion

‐amotionthroughthepassage

‐collision

‐……………

Model (possible) scenarios of behavior - precedents:

‐turningroundtheobstacle;

‐searchingfortheitem;

‐………………………

Relations (kindred and functional associations, comparisons):

Part/thewhole;more/less;friendly/alien;initial/terminal(final)…….

Rules of the World

Therulesforthegivenenvironmentand

givenrobot:

‐motionrules(brakeway,etc)

‐actionrulesatdanger

‐routineoftheday(schedule)

‐………………………..

GeneralrulesoftheWorld:

‐impossibletomovebackintime

‐impossibletopassthroughthewall

‐thereisapassagebetweenrooms

‐…………………..

Substantiation of the Required World Model Conﬁguration for Autonomous Mobile Medical Service Robots

129

doctor or a patient;

- instinct of “laziness” (to act if a motive is

present; otherwise, to return home and “sleep”) – a

condition for “locking” the general task of the

AMSMR functioning. Here, the “home” is the place

for charging batteries.

The last instinct is the main procedure for the

AMSMR to act. It determines the main loop of the

robot’s activity in its software.

4 CONCLUSIONS

We have tried in our study to develop the new,

formal and scientifically based engineer approach

for soundness of technical requirements to AMSMR

“from the first principles”. The expert analysis of the

basic scenarios of AMSMR behaviour in clinics

enabled classification of such robots as intellectual

dynamic systems acting according to a situation in a

multi-agent environment and basing on adaptive

logical reasoning. AMSMR are expected in clinics

to implement different errands associated, mainly,

with seeking an object of interest, interaction with it,

receiving and storage information from people,

information processing and transmission to other

people. As it was shown in the study, to follow the

scenario in a full-scale, such a robot must have well

developed model of our World – the World Model

(WM). Studying the principles of WM functioning

showed that WM should involve, in addition to the

models of all external objects, some general non-

object and procedural terms (to find, to be present, to

execute). Semantic relations between objects should

be also determined, as well as the basic world rules

(impossible to pass through the wall, etc.) should be

formulated. The model of AMSMR itself should

also be given in WM, and its behavioural instincts as

general motivations for a robot to act should be

formulated and determined, as well. It distinguishes

our WM from other ones. Correctness and efficiency

of the offered approach should be proved by further

experiments, but it is evidently now, that the first

and key step of AMSMR design and development is

the WM configuration design.

ACKNOWLEDGEMENTS

The reported study was funded by RFBR according

to the research project No. 14-08-01127.

REFERENCES

Burgard, W., Hebert, M., 2008. World modelling. In

Springer Handbook of Robotics, B. Siciliano and O.

Khatib eds. 853-869.Springer. Berlin.

Butter, M., Rensma, A., Boxel, J., et al., 2008. Robotics

for Healthcare. Final report of the study within

framework of the e-Health activities of the EU

Comission. EU Comission, DG Information Society.

Brussel.

Coltin, B., Liemhetcharat, S., Meriçli, Ç., Tay, J., Veloso,

M., 2010. Multi-humanoid world modelling in

standard platform robot soccer. In Proceedings of

2010 IEEE-RAS International Conference on

Humanoid Robots. Nashville. TN. USA.

Elfring, J.,Van den Dries, S., Van de Molengraft, M.,

Steinbuch, M., 2013. Semantic world modeling using

probabilistic multiple hypothesis anchoring. Robotics

and Autonomous Systems. 61(2). 95-105.

Evans, M.,1994. Helpmate: An autonomous mobile robot

courier for hospitals. In Proceedings of International

Conference on Intelligent Robots and Systems (IROS

'94). 1695-1700. Munich. Germany.

Fiorini, P., Botturi, D., 2008. Introducing service robotics

to the pharmaceutical industry. Intelligent Service

Robotics. 1(4). 267-280.

Iborra, A., Caceres, D.A., Ortiz, F.J., Franco, J.P.,Palma,

P. S., Alvarez, D., 2009. Design of service robots.

IEEE Robotics & Automation Magazine. 3. 24-33.

Kułakowski,K., Wąs, J., 2010. World Model for

Autonomous Mobile Robot - Formal Approach. In

Proceedings of 18th International Conference on

Intelligent Information Systems (IIS’2010). 37-45.

Poland.

Mastrogiovanni, F., Chong, N. Y., 2013. The need for a

research agenda in intelligent robotics. Inelligent

Service Robotics. 6. 1-3.

Rogatkin, D. A., Lapitan, D. G., Lapaeva, L. G., 2013.

Conception of the mobile autonomous service medical

robots. Biomeditcinskaya radio-elektronika

[Biomedical radioelectronics]. 5. 46-56. (in Russian).

Roth, M., Vail, D., Veloso, M., 2003. A real-time world

model for multi-robot teams with high-latency

communication. In Proceedings of IEEE/RSJ

International Conference on Intelligent Robots and

Systems. 2494–2499. Las Vegas. Nevada.

Russel, S., Norvig, P., 2003. Artificial Intelligence. A

Modern Approach. Prentice Hall, New Jersey, 2

edition.

Somby, M., 2009. Software platforms for service robots.

Servo. 1. 50-60.

Towle Jr, B., Nicolescu, M., 2014. An auction behavior-

based robotic architecture for service robotics.

Intellegent Service Robotics. 7(3). 157-174.

Wang, Y., Butner, S., Darzi, A., 2006. The Developing

market for medical robotics.In Special issue “Medical

Robotics”. Ed. by T. Kanade, B. Davis and C.N.

Riviere, Proc. IEEE. 94(9). 1763-1771.

Young, J., 1979. Robotics. Butterworths. London.

BIODEVICES 2016 - 9th International Conference on Biomedical Electronics and Devices

130