Typically, four types of human-robot interaction are distinguished: Coexistence, Synchronization, Cooperation, and Collaboration. They difer in the degree of interaction between human and robot and have therefore high impact on safety requirements for a system. Leading to human-robot cooperation and collaboration systems rarely being introduced in industrial practice due to the high risks associated with them. An underlying problem for this, is the lack of understanding for the diferences and necessities of these interaction types on a conceptual level. In this paper, we investigate the diferences between the four human-robot interaction types with respect to the dependencies between human and robot. Although in all interaction types humans and robots depend on each other, we show that these dependencies are very diferent in nature. For future use cases, this analysis helps in binding safety risks to concrete properties of the human-robot dependencies.
CEUR ceur-ws.org
To enhance performance, both industry and academia have been continuously endeavoring to
develop and refine self-assessment models capable of evaluating organizations’ readiness for
Industry 4.0 [
The integration of intelligent robotic systems in Industry 4.0 along with the support of humanrobot interaction (HRI) is seen as a key driver to reduce production costs while simultaneously improving product quality, particularly for producing small batch sizes or highly individualized products [4].
HRI is the field of designing, understanding, and evaluating robotic systems, that involves humans and robots interacting through communication [5]. Safety during HRI is vital, especially the safety of the human, namely, the safe interaction between operators and collaborative robots [6]. In current practice, the introduction of collaborative systems is often hindered by insuficient nEvelop-O (M. Daun) CEUR Workshop Proceedings understanding of the safety implications resulting from the interactions between human and robot in direct collaborative scenarios [7]. Therefore, we see a rising introduction of humanrobot interaction, focusing on a clear separation between the working spaces of human and robot. At the same time, the benefits of collaborative environments are seen as a key for further advancing of flexible manufacturing.
In this paper, we contribute an investigation of the dependencies between human and robot. To do so, we use goal models to highlight the diferences in dependencies between human and robot for diferent interaction types. We conclude by proposing a framework of human-robot dependencies that relate to diferent HRI types. This allows gaining a deeper understanding of the nature of human-robot interaction and its diferent manifestations (i.e. the diferent HRI types). It, furthermore, allows the development of systematic approaches to specify HRI and to assess the risks attached to the diferent dependencies.
The paper is structured as follows, Section 2 gives an overview of the related work. Section 3 investigates the dependencies that exist in diferent types of HRI. Section 4 then discusses and evaluates the diferences in the previously investigated dependencies. Section 5 then concludes the paper along with possible future work.
The field of HRI lacks consensus on the terminology used to classify the diferent types of interactions between humans and robots [8]. We refer to Bauer et al.’s [9] categorization of interaction types, as it covers numerous diferent modalities and is widely used in the field.
The various interaction types are illustrated in Figure 1 and are described in brief below: • Human-robot Coexistence is a type of interaction where both human and robot work in separated workspaces without any protective fences between them [10]. There exist no direct contact between the human and the robot, as both work on separate components in their respective workspaces. • Human-robot Synchronization is the type of interaction where both human and robot work on the same component in the same workspace, but not at the same time. They can have access to the workspace and component one after the other. The contact between them is unnecessary and should be avoided. • Human-robot Cooperation is the type of interaction where both human and robot work on diferent components in the same workspace and at the same time. The contact between human and robot might occur due to working in the same workspace, but it is unnecessary. • Human-robot Collaboration is the type of interaction where both human and robot work on the same component in the same workspace and at the same time. The contact between human and robot is necessary.
Goal modeling is an established approach in requirements engineering [11] and addresses the goals from the very early stages of development [12], already allowing the reasoning of concepts and diferent solution alternatives [ 13]. Common goal modeling approaches are iStar [14, 15, 16], GRL [17] (a light-weight standardized version of iStar) and KAOS [18, 19].
Goal models provide a systematic approach to displaying stakeholder or system goals in their context. The goals capture, at diferent levels of abstraction, the various objectives the system under consideration should achieve[20]. The iStar framework is, among others, beneficial in capturing and analyzing properties of complex systems in terms of actors, their intentions, and their relationships [21]. Goal models emphasize the relation between the various elements, which allows one to specify the hierarchical decomposition of goals, contributions, and dependencies. GRL and iStar have already been shown to be useful for modeling robotic systems in early phases (e.g., [22]). For instance, Morales et al. [23] use iStar to successfully specify teleo-reactive robots. Their extension is later on systematically integrated into the work by Gonçaleves et al. [24].
In previous work, we proposed the use of a lightweight GRL-compliant iStar extension to model collaborative cyber-physical systems [25]. Among the chosen system types to which the extension was applied, a fleet of collaborative transport robots used in a modern factory for evaluation. In more recent work, we extended this goal modeling extension to support safety analyses of human-robot collaborations [26, 22] and the specification of digital twins for human-robot collaboration systems [27]. 3. Investigating Dependencies in Human-Robot Interaction In this paper, we investigate what dependencies exist in diferent HRI scenarios. We, therefore, compare how dependencies difer for the four types of HRI: Coexistence, synchronization, cooperation, and collaboration. To do so, we introduce and then analyze a case example from the industry domain that includes all four types in one assembly process.
To visualize the diferences and maintain relation to the real world, we chose a small scale industry case example from the manufacturing domain, the assembly of a toy truck. The idea is that the truck is assembled on a collaborative workspace, where a human and a cobot work together in close proximity. A projector is used to display instructions for the human to follow, and it also displays boundaries of the workspaces for the human, the cobot and the area for both human and cobot. During the assembly of the toy truck, diferent human-robot interaction scenarios are applied, which correspond to four types of human-robot interaction: coexistence, synchronization, cooperation, and collaboration as shown in Figure 1. A detailed explanation of the tasks involved in the assembly process for each interaction type is provided in the subsections 3.2, 3.3, 3.4 and 3.5
While the human prepares four axle holders by inserting two screws in each holder, the robot assembles the base of the truck by inserting the cabin, the load carrier, and the chassis into the mounting bracket. Figure 2 shows the goal model for this interaction type. There exist no new dependencies except the normal ones between the actors during the execution tasks, as there is currently no communication between them.
Pick and place 8 screws Pick and place 4 axle holders Prepare axle holder AND
Assemble 4 axle holders
AND Certain tasks of the cobot are dependent on the tasks of the human and vice versa. These tasks are not necessarily bidirectional. The dependencies here illustrate how one actor must wait for the other actor to complete a task on a certain component. If there are diferent components involved, there exist no dependency between both actors. For example, the task of the actor human ‘Place one axle holder on the left of rear axle’ is dependent on the task of the actor cobot ‘Place rear axle in the back of the chassis’ meaning the cobot has to finish working on the part ‘rear axle’ before the human could start working on it.
AND
human-robot synchronization is similar to that of human-robot cooperation. The diference is Pick and place 8 screws
AND that, unlike cooperation, synchronization requires maintaining spatial restrictions. The human is not allowed to enter the shared workspace while the robot is working in the said workspace and vice versa.
After the cobot picks and places the front axle, it moves out of the shared workspace. The human then enters the
workspace and screws two axle holders to the left and right of the front axle. The human again leaves the area so that the cobot can place the rear axle. It is then mounted in the same way.
In Figure 4, the dependency shown in color blue is the one representing human-robot synchronization. The dependencies are considered from the fact that when the human is working in the shared area, the cobot is not allowed to enter the area and when the cobot is working, the human is not allowed to enter the shared area. For example, task ‘Place one axle holder on the right of front axle’ is dependent on ‘Place front axle in the front of the chassis’. Another example is that the task of the actor cobot ‘Place rear axle in the back of the chassis’ is dependent on the task of the human ‘Screw 2 screws to the right front axle holder’. The dependency explained here states that one task has to be completed for the second task to begin. Another dependency type states that the task must be completed, and the human must exit the shared workspace for the robot to enter and perform its task, and vice versa. Synchronization is restricted in comparison to cooperation due to these dependencies.
Human-robot collaboration difers from those for human-robot synchronization and cooperation, as an additional task of holding the axles is performed by the robot while the human is screwing the axle holders. After the cobot picks and places the front axle, it holds the axle in place for the human to screw the axle holders. Once both the axle holders are fixed, the cobot then moves on to performing the next task. The rear axle is then assembled the same way. Thus, for collaboration, human and cobot are working in the shared area, on the same component and at the same time.
In figure 4, the dependencies for human-robot collaboration are shown in the color orange. When the robot is holding and the human is screwing, both the human and robot are working together, which means they are working on the axles of the truck at the same time in the shared workspace. This is specified in the goal model, with a bidirectional dependency between tasks like ‘Screw 2 screws to the right rear axle holder’ of the human and ‘Hold rear axle’ of the robot. Another example of the bidirectional dependency can be seen between the task ‘Screw 2 screws to the left front axle holder’ and ‘Hold front axle’. The collaboration also involves an active interaction where the task of robot ‘Pick rear axle’ is dependent on the task of the human ‘Screw 2 screws to the left front axle holder’ which means the robot cannot pick the rear axle until the human is done screwing the axle holder to the right of the front axle. 4. Dependencies in Human-Robot Interaction
In human-robot coexistence, there exist no dependency between the actors human and robot. As for the rest, the dependencies depend on the sequence of execution of tasks and the restrictions of the workspace. In human-robot synchronization, tasks are performed one after the other by both actors, stating one actor has to wait for the other actor to finish a task. The actor also has to wait for the other actor to leave the part and the shared workspace for them to perform their own tasks.
In summary, for human-robot synchronization, two diferent dependency types need to be distinguished: process sequence dependencies and spatial dependencies. Process sequence dependencies refer to dependencies resulting from the assembly process (i.e. one task has to be ifnished before another task can be performed). Spatial dependencies result from the sharing of the workspace (e.g., the human must have left the workspace, before the cobot is allowed to enter).
For human-robot cooperation, again process sequence dependencies are as seen for humanrobot synchronization, as the tasks of both actors depend on other tasks to be executed by that time. As human and cobot do not work on the same component at the same time, it is not necessary to leave the shared workspace. However, it is necessary to observe spatial time dependencies, which difer from those used in human-robot synchronization. The actors are not dependent on a cleared space; rather, they must adapt their behavior when the other actor is present. For example, a robot must work more slowly if the human is present in a certain area.
In human-robot collaboration, both actors are dependent on each other when performing a task on a part in the shared area. Thus, we see bidirectional synchronization dependencies, where both actors need to synchronize to work on a task at the same time. Thus, the work of the cobot depends on the human and vice versa. In addition, we again have process sequence dependencies and spatial time dependencies for certain tasks.
In summary, five types of dependencies to be diferentiated to properly describe and diferentiate HRI are identified: • Normal dependency: In addition, to the dependencies related to HRI, normal dependencies still exist, which are known from other system types as well. For instance, actors or tasks being dependent on the availability of a resource. In this paper, we do not discuss these normal dependencies to avoid confusion with the dependencies related to the HRI types. • Process sequence dependency: Dependency rooted in the production process, where one task can only be executed after the successful completion of another task. • Spatial dependency: Dependency where a task by one actor can only be executed if the workspace has been prepared, meaning the actor who performed the previous task has vacated the workspace, allowing the current actor complete access to the entire area. • Spatial time dependency: Behavioral dependency, where the tasks of two actors are interdependent and must adhere to specific time related conditions during their execution. • Bidirectional synchronization dependency: Dependency where a task by one actor is dependent on a task by another actor, and reciprocally, that task by the other actor is dependent on the task by the first actor.
Figure 5 visualizes the diferent dependency types and their relations to the diferent HRI types using a simplified GRL meta-model based on [ 15]. We introduce the actors human and collaborative robot, which are central for HRI. In addition, the four newly identified dependency types are highlighted.
(Classical) Dependency Bidirectional Synchronization Dependency Spatial Time
Dependency Process Sequence
Dependency
Spatial Dependency
XOR‐ Refinement 0..1 to to
Contribution Type 2..* 1..* 1..* 00..1..11 <I<n0AEt.lbe.e*nsmttrieaoncntta>l> 00....**contributesTo {xor}
IOR‐ Refinement
0..1 Goal Softgoal
Task Resource
Belief Human‐Robot Coexistence
Human‐Robot Synchronization
Human‐Robot Cooperation
Human‐Robot Collaboration
A comparative analysis of the interdependencies between humans and robots reveals that the specific interdependencies involved vary depending on the interaction type. While only the normal dependencies need to be taken into account for coexistence, synchronization also involves process sequence and spatial dependencies, and cooperation involves process sequence and spatial time dependencies. In the context of collaboration, the same dependencies that apply to cooperation are also relevant, with the additional consideration of bidirectional dependencies (Figure 5).
It is essential to consider the potential safety risks associated with each dependency between humans and robots when performing tasks together. Therefore, it is crucial to identify these dependencies at the outset of the assembly sequence planning process, including how they may be influenced by the choice of interaction modality. Goal models ofer a straightforward method for identifying task-specific dependencies and the overall number of dependencies to be considered in the assembly process. This approach allows planning experts to evaluate the relative merits of diferent interaction modalities, taking into account the additional efort required to manage dependencies at the task level within the assembly process.
In future manufacturing scenarios, HRI is crucial, as evidenced by increasing research in this area. Despite strong research interest in intensive human-robot collaborations involving physical contact, there is reluctance in industry applications due to safety concerns and a lack of understanding of how these interactions afect each other. This paper investigates the diferences in dependencies between humans and robots across various interaction types.
In general, it is accepted, that there are four major types of HRI: Coexistence, Synchronization, Cooperation, and Collaboration, each varying in the degree of interaction and dependencies between humans and robots. In this paper, we used goal models to investigate the dependencies between human and robot for the diferent types of HRI. We conclude that there exist diferent types of dependencies as well. These dependencies can be categorized into four types excluding the normal dependency: Process sequence, Spatial, Spatial Time, and Bidirectional, which are based on the sequence of task execution and workspace restrictions. Process sequence refers to the order in which tasks are performed, spatial to the physical workspace shared, spatial time to the timely behavior in shared workspaces, and bidirectional to mutual communication. We then showed how these relate to the interaction types. It can be said that we do not have a 1:1 relation between interaction types and dependency types. However, not every dependency type is used for every interaction type and vice versa. Which we believe to have an impact on the safety assessment of the HRI types and needs to have an influence on structured approaches to develop human-robot interactions.
Understanding and managing these dependencies is crucial for ensuring both the safety and successful completion of tasks in HRI. These factors play a significant role in designing eficient workflows and enhancing the overall synergy between human workers and robotic systems. For the future, safety and potential threats must be taken into account concerning these types of dependencies. Therefore, an extensive analysis on the correlation between the dependency and the safety must be conducted. In addition, generalizability of our findings must be assured by application to diferent use cases in HRI. Furthermore, HRI are also used outside the manufacturing domain, for which transferability of the results to other domains (e.g., service robotics) should be proved. [3] G. J. Hamlin, A. C. Sanderson, Tetrobot modular robotics: Prototype and experiments, in: Proceedings of IEEE/RSJ International Conference on Intelligent Robots and Systems.
IROS’96, volume 2, IEEE, 1996, pp. 390–395. [4] J. Scholtz, Theory and evaluation of human robot interactions, in: 36th Annual Hawaii International Conference on System Sciences, 2003. Proceedings of the, IEEE, 2003, pp. 10–pp. [5] M. A. Goodrich, A. C. Schultz, et al., Human–robot interaction: a survey, Foundations and
Trends® in Human–Computer Interaction 1 (2008) 203–275. [6] A. Zacharaki, I. Kostavelis, A. Gasteratos, I. Dokas, Safety bounds in human robot interaction: A survey, Safety science 127 (2020) 104667. [7] R. Galin, R. Meshcheryakov, Collaborative robots: development of robotic perception system, safety issues, and integration of ai to imitate human behavior, in: Proceedings of 15th International Conference on Electromechanics and Robotics” Zavalishin’s Readings” ER (ZR) 2020, Ufa, Russia, 15–18 April 2020, Springer, 2021, pp. 175–185. [8] F. Vicentini, Terminology in safety of collaborative robotics, Robotics and Computer
Integrated Manufacturing 63 (2020) 101921. [9] W. Bauer, M. Bender, M. Braun, P. Rally, O. Scholtz, Lightweight robots in manual assembly– best to start simply, Frauenhofer-Institut für Arbeitswirtschaft und Organisation IAO, Stuttgart 1 (2016). [10] C.-P. Lam, C.-T. Chou, K.-H. Chiang, L.-C. Fu, Human-centered robot navigation—towards a harmoniously human–robot coexisting environment, IEEE Transactions on Robotics 27 (2011) 99–112. doi:10.1109/TRO.2010.2076851. [11] A. Van Lamsweerde, Goal-oriented requirements engineering: A guided tour, in: Proceedings fith ieee international symposium on requirements engineering, IEEE, 2001, pp. 249–262. [12] A. M. Grubb, M. Chechik, Formal reasoning for analyzing goal models that evolve over time, Requirements Engineering 26 (2021) 423–457. [13] J. Horkof, E. Yu, Interactive goal model analysis for early requirements engineering,
Requirements Engineering 21 (2016) 29–61. [14] E. S. Yu, Towards modelling and reasoning support for early-phase requirements engineering, in: Proceedings of ISRE’97: 3rd IEEE International Symposium on Requirements Engineering, IEEE, 1997, pp. 226–235. [15] D. Amyot, J. Horkof, D. Gross, G. Mussbacher, A lightweight grl profile for i* modeling, in: Advances in Conceptual Modeling-Challenging Perspectives: ER 2009 Workshops CoMoL, ETheCoM, FP-UML, MOST-ONISW, QoIS, RIGiM, SeCoGIS, Gramado, Brazil, November 9-12, 2009. Proceedings 28, Springer, 2009, pp. 254–264. [16] F. Dalpiaz, X. Franch, J. Horkof, istar 2.0 language guide, arXiv preprint arXiv:1605.07767 (2016). [17] ITU International Telecommunication Union, Recommendation ITU-T Z.151: User Requirements Notation (URN), Technical Report, 2018. [18] A. Dardenne, A. Van Lamsweerde, S. Fickas, Goal-directed requirements acquisition,
Science of computer programming 20 (1993) 3–50. [19] A. Van Lamsweerde, Requirements engineering: From system goals to UML models to software, volume 10, Chichester, UK: John Wiley & Sons, 2009. [20] A. van Lamsweerde, Goal-oriented requirements engineering: a guided tour, in: Proceedings Fifth IEEE International Symposium on Requirements Engineering, 2001, pp. 249–262. doi:10.1109/ISRE.2001.948567. [21] E. Yu, Modeling strategic relationships for process reengineering (2010). [22] M. Manjunath, J. Jesus Raja, M. Daun, Early model-based safety analysis for collaborative robotic systems, IEEE Transactions on Automation Science and Engineering (2024). [23] J. M. Morales, E. Navarro, P. Sánchez, D. Alonso, Tristar: an i* extension for teleo-reactive systems requirements specifications, in: Proceedings of the 30th Annual ACM Symposium on Applied Computing, 2015, pp. 283–288. [24] E. Goncalves, J. Araujo, J. Castro, Prise: A process to support istar extensions, Journal of
Systems and Software 168 (2020) 110649. [25] M. Daun, J. Brings, L. Krajinski, V. Stenkova, T. Bandyszak, A grl-compliant istar extension for collaborative cyber-physical systems, Requirements Engineering 26 (2021) 325–370. [26] M. Daun, M. Manjunath, J. Jesus Raja, Safety analysis of human robot collaborations with grl goal models, in: International Conference on Conceptual Modeling, Springer, 2023, pp. 317–333. [27] J. Jesus Raja, M. Manjunath, M. Daun, Towards a goal-oriented approach for engineering digital twins of robotic systems., in: ENASE, 2024, pp. 466–473.