Safety requirements for symbiotic human–robot collaboration systems in smart factories: a pairwise comparison approach to explore requirements dependencies


Industry 4.0 is expected to deliver significant productivity gain taking advantage of Internet of things (IoT). Smart solutions, enhanced by IoT, are constantly driving revolutionary approaches in multiple domains. Smart factories are one domain where intelligent integrated robotic systems will revolutionize manufacturing, resulting in a complex ecosystem, where humans, robots and machinery are combined. In this setting, human safety requirements are of paramount importance. This paper focuses on symbiotic human–robot collaboration systems (HRC), where human safety requirements are essential. Hence, it aims to explore and prioritize human safety requirement dependencies, as well as their dependencies with other critical requirements of smart factory operation, as effectiveness and performance. Toward this end, the proposed approach is based on SysML to represent the requirements dependencies and pairwise comparisons, a fundamental decision-making method, to quantify the importance of these dependencies. This model-driven approach is used as the primary medium for conveying traceability among human safety requirements as well as traceability from safety requirements to effectiveness and performance requirements in the system model. The analysis is based on the operational requirements identified in the European project HORSE, which aims to develop a methodological/technical framework for easy adaptation of robotic solutions from small-/medium-sized enterprises. Validation of the results is also performed to further elaborate on human safety requirement dependency exploration. The outcomes of this paper may be beneficial for symbiotic HRC systems in the early design stage. As the system is being developed with an emphasis on human safety, all these requirements that have been assessed with highly prioritized dependencies should be taken into account, whereas those with negligible ones have to be ignored since they do not significantly affect the rest of the process. Since operational requirements may be conflicted and incompatible, this approach may be very useful for other systems as well during the system design phase to find the appropriate solution satisfying the majority of the requirements, giving a priority to the ones with highly ranked dependencies and hence facilitating the implementation phase and afterward the production line. The outcomes may be used as a step in developing a model-driven approach which should be able to support the manufacturing process, facilitating the integration of systems and software modeling, which is increasingly important for robotic systems in smart factories incorporating HRC.

This is a preview of subscription content, log in to check access.

Fig. 1
Fig. 2
Fig. 3
Fig. 4
Fig. 5
Fig. 6
Fig. 7
Fig. 8


  1. 1.

    Grefen P, Vanderfeesten I, Boultadakis G (2016) Supporting hybrid manufacturing: bringing process and human/robot control to the cloud (short paper). In: 2016 5th IEEE international conference on cloud networking (cloudnet).

  2. 2.

    Home | HORSE. Accessed 18 Apr 2019

  3. 3.

    Sikora E, Tenbergen B, Pohl K (2012) Industry needs and research directions in requirements engineering for embedded systems. Requir Eng 17(1):57–78

    Article  Google Scholar 

  4. 4.

    Kulić D, Croft EA (2006) Real-time safety for human–robot interaction. Rob Auton Syst 54(1):1–12

    Article  Google Scholar 

  5. 5.

    Vasic M, Billard A (2013) Safety issues in human–robot interactions. IEEE

  6. 6.

    Fryman J, Matthias B (2012) Safety of industrial robots: from conventional to collaborative applications. VDE

  7. 7.

    Westman J, Nyberg M (2019) Providing tool support for specifying safety–critical systems by enforcing syntactic contract conditions. Requir Eng 24(2):231–256

    Article  Google Scholar 

  8. 8.

    Allenby K, Kelly T. Deriving safety requirements using scenarios. In: Proceedings fifth IEEE international symposium on requirements engineering.

  9. 9.

    Hansen KM, Ravn AP, Stavridou V (1998) From safety analysis to software requirements. IEEE Trans Softw Eng 24(7):573–584.

    Article  Google Scholar 

  10. 10.

    Wang J (1997) A subjective methodology for safety analysis of safety requirements specifications. IEEE Trans Fuzzy Syst 5(3):418–430.

    Article  Google Scholar 

  11. 11.

    Kotronis C, Nikolaidou M, Dimitrakopoulos G, Anagnostopoulos D, Amira A, Bensaali F (2018) A model-based approach for managing criticality requirements in e-health IoT systems. In: 2018 13th annual conference on system of systems engineering (SoSE).

  12. 12.

    Firesmith D (2004) Engineering safety requirements, safety constraints, and safety–critical requirements. J Obj Technol 3(3):27.

    Article  Google Scholar 

  13. 13.

    Firesmith D (2004) Prioritizing requirements. J Obj Technol 3(8):35.

    Article  Google Scholar 

  14. 14.

    Zhang W, Mei H, Zhao H (2005) A feature-oriented approach to modeling requirements dependencies. In: 13th IEEE international conference on requirements engineering (RE’05).

  15. 15.

    Anderson S, Felici M (2001) Requirements evolution from process to product oriented management. In: Product focused software process improvement, pp 27–41.

  16. 16.

    Winkler S, von Pilgrim J (2010) A survey of traceability in requirements engineering and model-driven development. Softw Syst Model 9(4):529–565.

    Article  Google Scholar 

  17. 17.

    Zhang H et al (2014) Investigating dependencies in software requirements for change propagation analysis. Inf Softw Technol 56(1):40–53.

    Article  Google Scholar 

  18. 18.

    Kulić D, Croft E (2007) Pre-collision safety strategies for human–robot interaction. Auton Robots 22(2):149–164

    Article  Google Scholar 

  19. 19.

    Thoben K-D, Wiesner S, Wuest T (2017) ‘Industrie 4.0’ and smart manufacturing—a review of research issues and application examples. Int J Autom Technol 11(1):4–16

    Article  Google Scholar 

  20. 20.

    Wang L, Törngren M, Onori M (2015) Current status and advancement of cyber-physical systems in manufacturing. J Manuf Syst 37:517–527

    Article  Google Scholar 

  21. 21.

    Ikuta K, Ishii H, Nokata M (2003) Safety evaluation method of design and control for human-care robots. Int J Rob Res 22(5):281–297

    Article  Google Scholar 

  22. 22.

    Geisberger E, Broy M (2012) agendaCPS: integrierte forschungsagenda cyber-physical systems, vol 1. Springer, Berlin

    Google Scholar 

  23. 23.

    Wiesner S, Gorldt C, Soeken M, Thoben K-D, Drechsler R (2014) Requirements engineering for cyber-physical systems. Springer, Berlin

    Google Scholar 

  24. 24.

    dos Santos J, Martins LEG, de Santiago VA Junior, Povoa LV, dos Santos LBR (2019) Software requirements testing approaches: a systematic literature review. Requir Eng 14:247

    Google Scholar 

  25. 25.

    Ellis-Braithwaite R, Lock R, Dawson R, King T (2017) Repetition between stakeholder (user) and system requirements. Requir Eng 22(2):167–190

    Article  Google Scholar 

  26. 26.

    Aisbl EUR (2014) Robotics 2020 multi-annual roadmap n for robotics in Europe, call 1 ICT23–horizon 2020. Initial Release B 15.01 (2014)

  27. 27.

    Grefen P, Vanderfeesten I, Boultadakis G. Developing a cyber-physical system for hybrid manufacturing in an Internet-of-Things context. In: Advances in business information systems and analytics, pp 35–63.

  28. 28.

    Vanderfeesten ITP, Erasmus J, Grefen PWPJ (2018) The HORSE project: IoT and cloud solutions for dynamic manufacturing processes. In: 5th European conference on service-oriented and cloud computing presented at the ESOCC 2016, pp 303–304

  29. 29.

    Modugno F, Leveson NG, Reese JD, Partridge K, Sandys SD (1997) Integrated safety analysis of requirements specifications. Requir Eng 2(2):65–78

    Article  Google Scholar 

  30. 30.

    RIoT Control (2017)

  31. 31.

    Operational requirements, August 2013

  32. 32.

    Bartneck C, Kulić D, Croft E, Zoghbi S (2009) Measurement instruments for the anthropomorphism, animacy, likeability, perceived intelligence, and perceived safety of robots. Int J Soc Robot 1(1):71–81

    Article  Google Scholar 

  33. 33.

    Kossiakoff A, Sweet WN (2002) Systems engineering principles and practice.

  34. 34.

    Berenbach B, Paulish D, Kazmeier J, Rudorfer A (2009) Software & systems requirements engineering: in practice. McGraw Hill Professional, New York

    Google Scholar 

  35. 35.

    Filiopoulou E, Mitropoulou P, Tsadimas A, Michalakelis C, Nikolaidou M, Anagnostopoulos D (2015) Integrating cost analysis in the cloud: a SoS approach. In: 2015 11th international conference on innovations in information technology (IIT).

  36. 36.

    Friedenthal S, Moore A, Steiner R (2014) A practical guide to SysML: the systems modeling language. Morgan Kaufmann, Burlington

    Google Scholar 

  37. 37.

    OMG Systems Modeling Language (OMG SysMLTM) (2015) Version 1.4

  38. 38.

    Hause M (2009) OMG systems modeling language (OMG SysML™) tutorial. In: INCOSE international symposium, vol 19, no 1, pp 1840–1972.

  39. 39.

    Roques P (2015) Modeling requirements with SysML. Requir Eng Mag, no. 2015–02

  40. 40.

    Dede G, Kamalakis T, Sphicopoulos T (2016) Theoretical estimation of the probability of weight rank reversal in pairwise comparisons. Eur J Oper Res 252(2):587–600.

    Article  MATH  Google Scholar 

  41. 41.

    Saaty TL (2003) Decision-making with the AHP: why is the principal eigenvector necessary. Eur J Oper Res 145(1):85–91.

    MathSciNet  Article  MATH  Google Scholar 

  42. 42.

    Dede G, Kamalakis T, Varoutas D (2011) Evaluation of optical wireless technologies in home networking: an analytical hierarchy process approach. J Opt Commun Netw 3(11):850.

    Article  Google Scholar 

  43. 43.

    Supriyono H, Sari CP (2018) Developing decision support systems using the weighted product method for house selection.

  44. 44.

    Brans JP, Vincke P, Mareschal B (1986) How to select and how to rank projects: the Promethee method. Eur J Oper Res 24(2):228–238.

    MathSciNet  Article  MATH  Google Scholar 

  45. 45.

    Triantaphyllou E (2000) Multi-criteria decision making methods: a comparative study. Appl Optim.

    Article  MATH  Google Scholar 

  46. 46.

    Yager RR (2004) Modeling prioritized multicriteria decision making. IEEE Trans Syst Man Cybern Part B (Cybern) 34(6):2396–2404.

    Article  Google Scholar 

  47. 47.

    Saaty TL, Vargas LG (2001) Models, methods, concepts & applications of the analytic hierarchy process. Int Ser Oper Res Manag Sci.

    Article  MATH  Google Scholar 

  48. 48.

    Dede G, Kamalakis T, Sphicopoulos T (2015) Convergence properties and practical estimation of the probability of rank reversal in pairwise comparisons for multi-criteria decision making problems. Eur J Oper Res 241(2):458–468.

    MathSciNet  Article  MATH  Google Scholar 

Download references


The research leading to these results has received funding from the European H2020-FoF-2015 Project “Smart Integrated Robotics System for SMEs Controlled by Internet of Things Based on Dynamic Manufacturing Processes (HORSE)” (680734).

Author information



Corresponding author

Correspondence to Georgia Dede.

Additional information

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.




Requirements definition

In this section, Table 1 of Sect. 4.1 is presented to the experts in order to understand the requirements and their description, before proceeding to the questionnaire. Moreover, the SysML diagram without the weights is also given to understand the requirements relationships.

Questionnaire Completion

Table 3 of Sect. 4.4 is also given to the experts in order to understand the nine-level scale and fill in the pairwise comparison matrices.

How to complete the questionnaire

The following questionnaire aims at prioritizing the dependencies between the requirements in order to evaluate their importance. For example, if S-FRQ01 depends on S-FRQ02 and S-FRQ03, we have to evaluate the importance of the dependencies in order to examine whether S-FRQ02 affects more S-FRQ01 than S-FRQ03 or the opposite.

Toward this end, you have to compare the requirements in pairs of two (pairwise comparisons) by allocating a value from the nine-level scale presented in Table 2. Please read carefully Table 3 (nine-level scale) and Table 1 (brief description of requirements) in order to complete the questionnaire.

Making the following pairwise comparisons, please allocate a number from the nine-level scale at each box. You compare the requirement presented in each row with all the other requirements presented in the columns, keeping in mind which requirement has more or less strong dependency and how much to the requirement that they affect.

For example, we know that both S-FRQ02 and S-FRQ03 requirements affect (derive/relate) the requirement S-FRQ01, then if we compare the S-FRQ02 with S-FRQ03 and put in the box the value 3, we mean that S-FRQ02 slightly affects more than the S-FRQ03 the requirement S-FRQ01.

Pairwise comparison for S-FRQ01 Requirement




Example of dependency computation

The estimation of dependencies is based on the PWC procedure described in Sect. 4.4. We want to explore and prioritize the dependencies of requirements S-FRQ03, S-FRQ01, S-FRQ02, S-FRQ17 and S-FRQ04. We denote these safety requirements as Sk (1 ≤ k ≤ 5) and the requirements with which are related in terms of relate and derive relationship as Ri. Toward this end, each expert m from a group of M experts fills in the PWC matrices mentioned in the above section of “Appendix” in order to explore the dependencies of each Sk requirement mentioned in the title of the PWC matrices. Each PWC matrix depicts the dependencies of the Sk requirement with the requirements presented in the matrix. Each of the aforementioned PWC matrices, filled in by the mth expert, corresponds to the P(m) matrix of the PWC process. The estimated weights w(m)i of the matrix (according to Eq. 1) are the dependency of the requirement Ri of the mth expert with the related Sk requirement of each PWC matrix. Then, the average weights wi for the M experts are estimated, based on Eq. (2). The weights wi define the weights of dependencies of the requirements Ri with the related Sk.

For example, we consider the first PWC of the questionnaire, namely the matrix depicting the dependencies of the requirement S1, namely S-FRQ03. We want to find the dependencies with the requirements S-FRQ01, E-FRQ07, E-FRQ13 and E-FRQ16. These requirements are the Ri requirements of the PWC process, where 1 ≤ i ≤ 4. Each expert 1 ≤ m ≤ M fills in this matrix, and hence, we have M PWC matrices for the dependencies of S-FRQ03. We consider each of these matrices as P(m). For each P(m), we apply the eigenvalue method and we estimate the weights w(m)i (according to Eq. 1) which is the dependency of the requirement Ri of the mth expert to the S1. We then estimate the average of these weights, say wi, based on Eq. (2). The weights wi are the weights of dependencies of the Ri to the S1. These are the weights depicted in the SysML diagram (Fig. 4) as well as in Figs. 5, 6 and 7.

The procedure is the same for the exploration and prioritization of the dependencies of the other requirements.

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Dede, G., Mitropoulou, P., Nikolaidou, M. et al. Safety requirements for symbiotic human–robot collaboration systems in smart factories: a pairwise comparison approach to explore requirements dependencies. Requirements Eng (2020).

Download citation


  • Symbiotic human–robot collaboration systems
  • Safety
  • Requirement analysis
  • Dependencies
  • SysML
  • Decision making
  • Pairwise comparison