=Paper=
{{Paper
|id=Vol-3038/paper14
|storemode=property
|title=Hybrid Expert System for Collaborative Decision-Making in Transportation Services of Healthcare Needs
|pdfUrl=https://ceur-ws.org/Vol-3038/short6.pdf
|volume=Vol-3038
|authors=Tetiana Shmelova,Oleksandr Sechko
|dblpUrl=https://dblp.org/rec/conf/iddm/ShmelovaS21
}}
==Hybrid Expert System for Collaborative Decision-Making in Transportation Services of Healthcare Needs==
https://ceur-ws.org/Vol-3038/short6.pdf
Hybrid Expert System for Collaborative Decision-Making in
Transportation Services of Healthcare Needs
Tetiana Shmelovaa, Oleksandr Sechkob
a
National Aviation University, Liubomyra Huzara ave. 1, Kyiv, 03058, Ukraine
b
Uzhhorod National University, Narodna Square, 3, Uzhgorod, 88000, Ukraine
Abstract
The authors presented a Hybrid Expert System for Collaborative Decision-Making (CDM) in
transportation services of healthcare needs. The analysis of the process of delivering medical
supplies to remote areas, in a smart city using UAV, groups of UAVs as a decision-making
process of several participants presented. Analysis multi-DM using network planning models
for all participants of a process and integration of models DM under Risk, and Uncertainty
showed. Optimization with minimal cost and maximum safety example delivering medical
supplies for the smart city using UAVs presented. This is achieved by fullness, precision, and
real-analysis of existing data. Planning of solutions provides using deterministic, stochastic,
and non-stochastic decision-making models; methods of dynamic programming and reflexion
models.
Keywords 1
Artificial Intelligence, Expert System, Decision-Making in Risk, Decision-Making in
Uncertainty, Deterministic Model, Health Care, Collaborative Decision-Making Models,
Transportation/Service/Logistic.
1. Introduction
There are many systems based on knowledge and experience such as expert systems (or decision
support systems) for effective support of medicine, for example, post-disaster patient transportation,
transporting important and urgent cargo. Main modes of transport (air, water, and land transport) are
widely used in civil and military emergency medicine due to the fast speed of transfer, effectiveness in
difficult situations, but air, water, and land transport have various opportunities for effective use [1–3].
Nowadays, unmanned aviation has been rapidly developing especially for the realization of
intelligence solutions, for example in smart-city [3; 4]. New technology and data obtained by modern
methods, such as using UAVs, to monitor cities in near real-time help to simulate minimal risk in
situations proposed for future solutions according to theoretical principles of sustainable urbanism [4].
The development of Unmanned Aerial Systems (UAS) based on Unmanned Aerial Vehicles (UAVs) is
currently being carried out by many industrially developed countries of the world. Until recently, UAVs
had a military purpose, while the use of the UAS is effective both in civilian and military tasks, for
example, in dealing with the effects of emergencies, natural disasters, agriculture, reconnaissance, aerial
photography, and for the healthcare industry [3; 5; 6]. More common UAV applications in healthcare
have started from the provision of disaster estimation where other means of access are strictly limited
[2; 3; 7]; delivering aid medicines packages, medicines, vaccines, blood, and other medical supplies to
remote areas, for example, mountain area [6]; ensuring the safe transport of samples and disease test
kits to highly contaminated areas; potential for quick access to automated defibrillators for patients in
cardiac arrest [1]. Recently work on UAVs for direct transportation and medical evacuation has started
of difficult ill patients [3; 4; 6].
IDDM’2021: 4rd International Conference on Informatics & Data-Driven Medicine, November 19–21, 2021, Valencia, Spain
EMAIL: shmelova@ukr.net (T.Shmelova); aleksander.sechko@gmail.com (O.Sechko)
ORCID: 0000-0002-9737-6906 (T.Shmelova); 0000-0002-4136-5511 (O.Sechko)
©️ 2021 Copyright for this paper by its authors.
Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
CEUR Workshop Proceedings (CEUR-WS.org)
International civil aviation organization (ICAO) supports new applications of aviation and
implement new conceptual models for the search for optimal solutions [8-10]. Interaction can be done
in the form of collaborative decision-making (CDM) by all participants based on the reciprocal
exchange of helpful data [10]. The authors propose to introduce in medicine the methodology of CDM
for improving the effectiveness of transportation that is used in aviation [11; 12] with the application
of integrated models of decision-making (DM) in certainty, risk, and uncertainty, and Artificial
Intelligence (AI) methods.
The purpose of the publication:
• The analysis of the process of delivering medical supplies to remote areas, in a smart city,
and between cities using UAVs and groups of UAVs as the multi-DM process of several
participants.
• Analysis of the multi-DM using network planning models for all participants of a process
and integration of DM models in risk and uncertainty.
• Optimization delivering medical supplies to remote areas, in a smart city, and between cities
using UAVs and groups of UAVs.
2. Collaborative Decision-Making Models for Transportation
2.1. Main scientific results
The CDM is an effective process of exchanging data, individual and collaborative decision-maker
by different contacting members. Determination of potential participants of the process of delivering
medical supplies depends on the purpose of the task, characteristics of medical supplies, the urgency of
delivery, the distance of delivery, etc. As a rule, the participants in the delivery are the sending and
receiving parties (medicine specialists) and specialists in delivering (logistic/transportation company).
It is important to provide an opportunity for CDM with partners at a reasonable level of efficiency
and balance (minimal risk and maximum safety). This is achieved by fullness, precision, and real-
analysis of existing data. Planning of solutions provides using deterministic, stochastic, and non-
stochastic decision-making models; methods of dynamic programming and reflexion models. To
consider the complexity of the factors that affect the human in the expected and unexpected conditions,
a reflexive model of bipolar choice of human has been designed [11; 12; 13]. The result of assessing
unprofessional factors is the definition the social-psychological impact on a person’s DM by revealing
the preferences, diagnostics the individual-psychological qualities of humans during the situation
development, monitoring of the human’s psycho-physiological factors (emotional state) for early
diagnosis of transition to potentially hazardous mental performers and determining stability of patients
in working capacity was obtained [11; 12]. In the “Informational processor of the reflexive intuitive
choice” by human is selected in the directions of positive pole A, negative pole B; mixed selection АВ
according to reflexion theory [13]. The choice of human is described by the function:
X = f (x1 , x2 , x3 ) ,
where Х – is a probability, that a person is ready to choose the positive pole A in reality; x1 – is an
environmental pressure on a person towards a positive alternative at the time of choice, х1 [0, 1]; x2 –
is the pressure of a person's previous experience on a positive alternative at the time of choice, х2 [0,
1]; х3 – the pressure of a person's desire for a positive alternative at the time of choice, х3 [0, 1].
The general technique of DM by the participants in specific transportation/service/logistic (TSL) is
included:
1. Analysis of situation as a complex situation: identification of causal relationships and
determination of potential participants of the complex process of delivering medical supplies.
2. Construction of the algorithms of the potential participants' actions in TSL. Determination of
the average time of each action for all participants in this situation and the rational sequence of all
actions (compilation of a basic structural-temporal table for each participant).
3. Modeling of DM participants' actions in TSL using network planning graphs (Figure 1):
− Network graph of main technology (instruction) for each participant in the situation.
− The main critical time and critical ways of performance of all action.
− Network graph of main technology/instruction.
If it's model difficult with undetermined decisions (many solutions and wishes) may use the
integration of models based on risk and uncertainty DM models.
Figure 1: The deterministic models with controversial actions (S1, S2) of members (A1, A2, A3)
4. Optimization of schedule/plan of performance of main technology and simplification of a
complex model (Figure 2). Identification of difficult points where were several alternative solutions
and in next using the effective method of DM:
− With the existence of Big Data of the process (experiences, statistic data) are used AI methods
for forecasting the responsibility of solutions.
− In a large amount of statistical data and probabilities are used DM methods in risk (Figure 3).
− In the absence of a large amount of statistical data and probabilities, are used DM methods in
conditions of uncertainty (Table 1).
Figure 2: The deterministic models with decisions A1, A2, A3 - simple Network graph
5. DM models in risk conditions: evaluation of risk R for different decisions (tool – decision tree).
The DM periods are described by decisions (A = {A1; A2; …, An}), a time t of the evolution of the
situation on each stage, and added value β, that depends on the period of the evolution of the situation
and timely DM for countering the situation (Figure 3).
When solving the problem of minimizing the risks during each period, added risks growth (+βk), the
threats are increasing with time t:
𝑅𝑘 = 𝑡𝑘 ∑𝑛𝑖=1 𝑝𝑖 𝑢𝑖 ± 𝛽𝑘 ,
where ti – is a time of the period k; βk – is an added risk during the period k; pi – are the
probabilities of the evolution of the situation, ∑𝑛𝑖=1 𝑝𝑖 = 1; ui – are the anticipated outputs.
The DM model in risk is shown in Figure 6. Step-by-step correction of the decision matrix is carried
out in risk assessment [17].
Figure 3: The periods of the evolution of situation and DM on the decision tree
5. Application of the method of the objective-subjective decision in conditions of uncertainty for
the formation of CDM.
6. Decision-making matrix under uncertainty for each participant in the process. In the matrix
(Table 1), factors are external circumstances influencing decision-making, alternative decisions are
possible actions. The optimum decision is found using the criteria of Wald, Laplace, Hurwitz, Sevij -
minimum losses and maximum safety during transportation. Each of the criteria has a set of differences
in application. The main difference is the different levels of problem uncertainty, types of situations
(often, rare, first time), transport opportunities, and complexity of care situation. For instance, the
Laplace criterion is based on more upbeat opinions (same situations what were); the Wald criterion is
based on more pessimistic opinions and is applied to find the optimum decision for the first moment.
The optimism-pessimism coefficient is applied in the Hurwicz criterion that can be adapted in various
accesses from the most optimistic to the most pessimistic grade. The Savage criterion is applied for
decisions recounting to minimize the losses after completion of the situation.
Table 1
Decision-making matrix in uncertainty for each member
{А} F1 - objective factors Results according to criteria
f1 f2 … fj … fn Wald Laplace Hurwitz Sevij
А1 U11 U12 … U1j … U1n F11 F11 F1 F11
Alternative А2 U21 U22 … U2j … U2n F21 F21 F2 F21
solutions … … … … … … ….
Аi Ui1 Ui2 … U ij … U in Fi1 Fi1 Fi1 Fi1
… … … … … … …
Аm Um1 Um2 … Umj … Umn Fm1 Fm1 Fm1 Fm1
7. Decision-making matrix under conditions of uncertainty of all participants in the process. In
the matrix (Table 2), factors are the opinions of participants in the transportation process, alternative
solutions are joint possible actions. The optimal decision – minimum losses and maximum safety during
transportation, taking into account all partners-parties.
Table 2
Decision-making matrix in uncertainty for CDM
{А} F2 - subjective factors Results according to criteria
f1 f2 … fj … fn Wald Laplace Hurwitz Sevij
А1 F11 F12 … F1j … F1n AS1 AS1 AS1 AS1
Alternative А2 F21 F22 … F2j … F2n AS2 AS2 AS2 AS2
solutions … … …
Аi Fi1 Fi2 … Fij … Fin ASi ASi ASi ASi
… … …
Аm Fm1 Fm2 … Fmj … Fmn ASm ASm ASm ASm
8. Expert system (ES) for assessment of the operability of all types of transport (air, water, land)
for solving various intentional goals in urban areas [4; 5].
9. Expert system (ES) for assessment of the operability of UAVs flights (single and group) for
solving various intentional goals in urban areas:
1) Assessment of the efficiency of the intentional goals of the next systems applying: a group of
individual UAVs controlled by individual operators; UAV group controlled by call detail records
(CDR)-UAV; single UAV controlled by one operator. If there is a group of UAVs with control from
CDR, then it’s necessary:
• Decomposition of the complicated system into subsystems “network topology –
intentional goals”, a description of the subsystems’ specifications, and an assessment of
the efficiency of network topologies for execution of the concrete intentional goals.
• The efficiency of the topologies of the network for execution of the intentional goals and
determination of evaluation criteria (determination of the appropriate weights for the
topology effectiveness).
• Assessment of the efficiency of network topologies of the UAV group for the concrete
intentional goals applying Expert Judgment Method (EJM) (determination of the experts’
preferences and consistency).
2) Assessment of the urban areas for applying UAVs and methods (GRID analysis of a sectoral
UAV flight, fuzzy logic, and EJM for “risk” assessment).
3) Aggregation of subsystems into a new system (additive or multiplicative aggregation,
whichever the “intentional goals” type).
4) Graphical performance of the ES results, for example, assessment of the efficiency of the
topologies of network for execution the intentional goal “transportation” by UAV group, single UAV
or group of single UAVs (Figure 4 To assess the safety of UAV flights in a city, it is necessary to get
quantity values of flight risks in various parts of the city.
Figure 4: Graphical performance of the ES results, of the efficiency of the topologies of network for
execution of the intentional goal “transportation” by UAVs
For example, appraisal and finding the way with the minimal cost W1 for UAV1 in Figure 5, W1=39
for the territory fragment which is depicted in Figure 6.
Figure 5: The territory fragment for appraisal of the minimal cost and safety of UAVs motion
Figure 6: Grid cell risk assessment and calculation of the minimal cost of way W1 for UAV1
The model of synchronization of actions of all participants in the delivery are the sending and
receiving parties (medicine specialists) and specialists in delivering (logistic/transportation company)
in the difficult conditions can be designed applying artificial intelligence (AI) techniques. The
formalization of all participants’ actions in TSL with the assistance of AI would allow DM methods to
define the optimum subsequence and the time of the performance of the procedure for resolving non-
standard situations.
To prepare participants to correct and effective DM action in non-standard conditions, training
procedures must be able to simulate situations closely approximating real-life events. The steps in the
process of creating such simulated environments are:
− a thorough and deep analysis of the emergency case;
− intellectual processing of data;
− situation identification;
− formalization of the situation with the help of integrated models;
− decomposition of the complicated situation into subclasses;
− synthesis of adapted deterministic models for actions determined by AI.
For example, urgent delivery of medical supplies from the starting point (Uzhgorod) to the terminal
point Khust is required. The distance between the cities of Uzhgorod and Khust is about 100 km,
mountainous terrain, the UAV delivers urgent cargo before the arrival of the doctor with the minimal
cost (Figure 7).
Algorithm of definition minimal cost and maximum safety of UAVs movement ways between towns
in next:
1) Grid-analysis - cells are superimposing on a fragment of map (Figure 7).
2) Risk assessment of Grid cells depending on the type of area (“Track area - TA”, “Restricted
area - RA” or “Dangerous area - DA”).
3) Finding the minimum cost path for a UAV1 using the Dynamic Programming method for
planning a flight in a first level:
Wi (yi ) = yi −1 ( RA; DA; TA) + min ( yi ( RA; DA; TA) )
The minimum cost path is 21 conventional units (Figure 7).
Figure 7: Prototype of hybrid monitoring and situation management system ML-DM
In cases of big and complicated data, techniques can be integrated into traditional and hybrid DM
systems of the next generation by processing uncontrolled data of situations in the deep landscape
models (Figure 8), potentially with high data transfer and almost in real time, creating a structured
presentation of input data by clusters corresponding to the types of general situations [14; 15; 16].
Figure 8: Prototype of hybrid monitoring and situation management system ML-DM
In Figure 8 above, a deterministic model of actions is focused on a concrete type of situation. One
more advantage of this model is its potential ability to study to define the interconnections between
various types of situations, almost entirely in self-monitoring learning modes with very limited demands
to reliable data. Possible uses of these opportunities of models of machine intelligence can spread, for
example, on developing the abilities to discover early signs or symptoms of emerging situations through
relationships between types of situations, as well as the ability to create notifications and early warnings,
which a person can take in advance before the situation develops.
2.2. Experiments & discussions
So, AI is a framework of methods, models, and practices that is capable of performing some human
intellectual or physical activities related to the perception and processing of information, reasoning, and
DM, communications with natural Intelligence (human), rational support of human in the efforts. These
processes of building AI include training (obtaining information and rules for applying the information);
reasoning, evaluating, and modeling (applying rules to get approximate or final results); self-correction
(assessment of the resulting models); automated systems and human-computer systems; image
recognition systems, speech recognition, and machine vision, etc. Particular applications of AI include
next systems: Expert systems (ES); Decision support systems (DSS); human-computer systems (HCS);
Automated systems (AS); AI systems. To design and develop an AI system, it is necessary to create an
Expert system. An Expert system is an informal model of the system being created, with the help of
expert assessment, on small data it's possible to create a demo version of the AI system. The
accumulation of data creates a real AI system. Ready-made AI systems have varying degrees of
performance and DM. The degree of productivity changes from simple (simple actions) to complex
(creative actions):
• simple AI actions - repeating actions;
• complicating AI actions - repeating actions and creating new actions according to existing
rules;
• complex AI actions - repeating actions, creating new actions, changing the rules for
performing actions;
So, steps for building an AI system:
1. Expert system - a data description information – using experts (according to statistics,
experience, skills data too).
2. DM and CDM models – to improve and prepare data.
3. AI systems without training data and effective DM in difficult processes.
4. Big Data to create AI systems with training data and more effective DM /CDM.
5. Big Data to create an AI system with Machine Learning and IDM (Intelligence DM).
6. Big Data to create an AI system with Deep Learning and IDM (Intelligent systems of DM),
DM models, models of forecasting of development situations and optimal solutions.
7. Intelligent systems of DM – combine natural and AI – Hybrid DSS.
3. Conclusion
Optimum solutions planning should provide using DM different models such as deterministic,
stochastic, and non-stochastic models. After analyzing the situation, it is necessary to synthesize
stochastic models to correct an indefinite deterministic model with a set of solutions. A continuous
reporting and CDM process is required to synchronize the decisions made by participants and exchange
information between them involving natural intelligence and AI as a combined hybrid intelligence for
effective DM. It is essential to provide the ability to develop a joint, comprehensive solution with
partners at a sufficient efficiency level. The obtained integrated DM and CDM models can be applied
in the DSS and ES of physicians to serve patients in future medical AI-based systems.
The example of the transportation medical service situation of patients in healthcare, the search for
an optimal solution for effectively delivering medical supplies (with minimal cost and maximum safety)
and delivering timely medical care of patients presented.
The obtained DM models can be applied in the DSS of physicians to serve patients in future medical
AI-based systems.
4. References
[1] Xiang-Hui Li, Jing-Chen Zheng. Efficient post-disaster patient transportation and transfer:
experiences and lessons learned in emergency medical rescue in Aceh after the 2004 Asian
tsunami. Military medicine, 179, 8:913, 2014, journal-article. doi: 10.7205/MILMED-D-13-00525
[2] C.G. Lowe. Pediatric and neonatal interfacility transport medicine after mass casualty incidents.
The Journal of Trauma: Injury, Infection, and Critical Care: August 2009 - Volume 67 - Issue 2;
journal-article. doi: 10.1097/TA.0b013e3181af6086
[3] Anna Konert, Jacek Smereka, Lukasz Szarpak. The Use of Drones in Emergency Medicine:
Practical and Legal Aspects. Emergency Medicine International, 2019, journal-article.
doi.org/10.1155/2019/358979
[4] J. A. Tenedório, R. Estanqueiro, C. Delgado Henriques. Methods and Applications of Geospatial
Technology in Sustainable Urbanism. International Publisher of Progressive Information Science
and Technology Research, USA, Pennsylvania, 2021. doi: 10.4018/978-1-7998-2249-3
[5] T. Shmelova, V.Lazorenko, O. Burlaka, Methods and Applications of Geospatial Technology in
Sustainable Urbanism. Chapter 15. Unmanned Aerial Vehicles for Smart Cities: Estimations of
Urban Locality for Optimization Flights. International Publisher of Progressive Information
Science and Technology Research, USA, Pennsylvania, 2016. doi: DOI: 10.4018/978-1-7998-
2249-3.ch015
[6] W. Bo, V. Kharchenko, A. Nakhaba. Mathematical model of automatic flight of polikopter UAV
NAU PKF "AURORA" Proceedings of National Aviation University 69(4), 2016, journal-article,
DOI:10.18372/2306-1472.69.11050
[7] A. Claesson, L. Svensson, P. Nordberg et al. Drones may be used to save lives in out of hospital
cardiac arrest due to drowning, Resuscitation, vol. 114, 2017, journal-article.
doi.org/10.1016/j.resuscitation.2017.01.003
[8] Safety Management Manual (SMM) (3rd ed.). Doc. ICAO 9859-AN 474. Canada, Montreal. 2013.
[9] Manual on Collaborative Decision-Making, 2nd ed., Doc. 9971. Canada, Montreal. ICAO. 2014
[10] Manual on Flight and Flow Information for a Collaborative Environment (FF-ICE), 1st ed., Doc.
9965. Canada, Montreal. ICAO. 2012
[11] T.Shmelova. Information Technology Applications for Crisis Response and Management. Chapter
10: Collaborative Decision Making in Emergencies by the Integration of Deterministic, Stochastic,
and Non-Stochastic Models. International Publisher of Progressive Information Science and
Technology Research, USA, Pennsylvania, 2021
[12] T. Shmelova, A. Sterenharz, Yu. Sikirda (Eds.) Handbook of Artificial Intelligence Applications
in the Aviation and Aerospace Industries. Chapter 1 T. Shmelova, A. Sterenharz, S. Dolgikh,
Artificial Intelligence in Aviation Industries: Methodologies, Education, Applications, and
Opportunities. International Publisher of Progressive Information Science and Technology
Research, USA, Pennsylvania. 2020 doi: 10.4018/978-1-7998-1415-3.ch001
[13] Vladimir A. Lefebvre Theoretical modeling of the subject: Western and Eastern types of human
reflexion Computer Science, Medicine Progress in biophysics and molecular biology June, 2017,
journal-article. DOI:10.1016/j.pbiomolbio.2017.06.006 Corpus ID: 26607071
[14] T.Shmelova, A.-B.M. Salem, V.Smolanka, O. Sechko, Collaborative deterministic and stochastic
decision-making models in health care In CEUR Vol 2753 Workshop Proceedings.
[15] S.Dolgikh Spontaneous Concept Learning with Deep Autoencoder In International Journal of
Computational Intelligence Systems, Volume 12, Issue 1, November 2018, pp. 1 – 12, Canada.
[16] P. Prystavka, O.Cholyshkina, S.Dolgikh, D.Karpenko. Automated Object Recognition System
based on Convolutional Autoencoder. 10th International Conference on Advanced Computer
Information Technologies, ACIT 2020 Proceedings.