2013 IEEE International Multi-Disciplinary Conference on Cognitive Methods in Situation Awareness and Decision Support (CogSIMA), San Diego

User assistance by situational, task- and model-based process information filtering in interactive, semi-automated human-automation systems
Marcel Langer and Dirk S¨ ffker o Chair of Dynamics and Control
University of Duisburg-Essen
Duisburg, Germany
Email: {marcel.langer; soeffker}@uni-due.de
Abstract

operators representing (human or technical) actions.
Tasks are described by defining final situations that need to be reached and process information are fused to a situation vector representing the actual state of the entire system. Suitable information is identified by comparison of differences in future situations and used for the detection of valuable information with respect to the defined (sub)task. The hierarchical task analysis [4] of a semi-automated molding process using No-Bake-Technique is additional discussed as an illustrative application example that is completely mapped into the SOM-framework. Furthermore, the results of a user study of selected functionality of the developed process guiding and supervision assistance system are presented.

Nowadays, the integration of users into technical systems becomes necessary due to reasons of automation concepts that are not capable of completely integrating process knowledge and manufacturing skills for full automation of technical processes. In such interactive, semi-automated systems human operators are facing a high amount of process information that need to be evaluated according to individual skills and knowledge to situational and necessary actions including the evaluation of alternatives etc. In order to benefit from human abilities of information fusion and sophisticated comprehension of process coherence in complexity gaining environments, integrated approaches as framework for guiding and supervision assistance systems of semi-automated, technical systems can be developed. Thereby, the advantage of having an uniform description of the entire, interactive system
(consisting of human, machine, and technical process) as framework allows the implementation of methods for a situational, task-oriented assessment of existing information and corresponding filtering, respectively.
The resulting reduced set of process-relevant information allows the human operator to focus during the decision making process and supports maintaining process reliability and quality.
In this contribution, recent modeling approaches of Human-Automation-Interaction, e.g. Message-based
Part State Graphs [1] or Enriched Labeled Transition
Systems [2] are discussed with respect to their ability of being suitable to serve as framework and are compared to the Situation-Operator-Model (SOM) introduced by [3]. Using the SOM approach, considered interactive processes can be represented by a set of situations defined by characteristics connected through

978-1-4673-2437-3/13/$31.00 ©2013 IEEE

1. Introduction
Developing advanced technical systems aims specific goals, such as satisfaction of human factor requirements or increasing productivity. The challenge during the development or technical system (re)design process concentrates on solving the occurring conflicts to meet either requirements. In such cases automation concepts may become necessary, that are capable to fuse advantages of the combination of human and automation capabilities leading to a certain level of automation [5]. The achievable level of safe and reliable automation thereby depends on the considered processes (tasks) and is initially proposed by early work of Sheridan and Verplank and recently discussed in [5]. When analyzing technical processes that are under investigation these days, several applications can be categorized as semi-automated processes [5–10].
Recent case studies [8] confirm the value and the advantage of Human-Centered-Automation (HCA) ap-

106

proaches, detail challenges of their implementation [9], and identify recommendations to improve Situation
Awareness (SA) of human operators in complex environments. Therefore, determining the relevant information [10] is an essential strategy to reduce user’s input and thus allowing human operators to be focused when interacting with dynamical systems. Catching necessary attention and enhancing the user’s attention allocation by reducing the input set of information is essential to improve user assistance in semi-automated technical systems. The Case-based Situation Awareness (CBSA) approach [11] is successfully applied to design context-aware user interfaces [6].
The CBSA approach is a suitable descriptive technique and capable of forecasting cases, while not modeling interaction itself or formulating tasks natively in terms of formulating cases to reach. The combination of information collection, situation creation, case identification, situation projection, and corresponding information filtering to support users action selection process with a single descriptive modeling technique is still an open question and are discussed within this contribution. Recent interaction models for model-based, stateoriented, and event-discrete description of semiautomated (cooperative) processes are introduced by
Message-based Part State Graphs (MPSG) referring to
[1] or (Enriched) Labeled Transition Systems (ELTS) referring to [2]. Considering the ELTS approach, the modeling concept is based on a pre-defined minimal, machine-based mental model that is completely known
(by learning and training) to the operator (‘full control property’ refer to [2]). In difference the MPSG approach is using the state of the work piece (or considered part), identical to the state of the considered system, as representation. The state transitions (interactions) are triggered by messages that are exchanged between human operator and system.
In order to implement a state information-oriented assistance system to support human operators in their decision making process when dealing with semiautomated systems, ELTS and MPSG are not suitable as only subsets of available state information is used for the description. In ELTS the state is marked by a pointer defined within the minimal mental model of the process. In MPSG the position of the considered work piece is located within the process model and used as state description. Thus, both approaches are not capable to identify the actual process state by measurements or initiating routines based on the known
(sensed) process information to support information filtering or offering user assistance - refer to [11].
As framework for the implementation of user assis-

tance systems addressing issues in detecting relevant information to support human operators, the SituationOperator-Model approach [3] is used in specifically semi-automated scenarios. As application example the molding process using the No-bake technique, in which human workers need to be separated locally from the process to take into account human safety requirements and to achieve a repeatable process quality is chosen.

2. Modeling interactive processes
For modeling of interactive processes it is necessary to represent the physical (or technical) process including sensors, actuators, etc., the human operator as process knowledge keeper, the interactions between human and process, and the interface(s) as illustrated in Fig. 1(a), whereby different methods can be used.
The ELTS [2] or MPSG [1] methods are capable of representing each part. However, the representation itself is not taken into consideration valuable process information from sensors or actuators on a fundamental level. The proposed modeling method, the SituationOperator-Model, on the contrary allows a differentiated representation in this sense [3].

2.1. Situation-Operator Model
For implementing an user-centered assistance system to support human operators while guiding the process, the technical process is modeled by the
Situation-Operator-Model (SOM) approach [3]. The
SOM methodology allows a formal description of complex processes by taking real world scenes and assigning descriptive characteristics that may underlie further relations to form situations (refer to Fig. 1(b)).
Each situation consists of an unique combination of characteristics values that can be of an arbitrary type
(e.g BOOLEAN, descriptive, etc.). Regarding a technical process, a situation is understand as the mapping of a (frozen, problem-oriented) process scene described by process variables (as characteristics).
Thereby human interventions are interpreted as actions and correspondingly described by so-called operators.
An example of a process describing situation is given in Fig. 1(c).
The segments of the process are modeled by a set of characteristics that can be changed by fulfilling a process task. A process task corresponds to a sequence of operators formulated in the SOM formalism. Thus, in the SOM-based model, operators are actions released automatically or by human intervention transferring an initial situation to a target situation. The situation transition can also include releasing a series of operators to

107

S

Alarms
Notifications
Measurements
Optional actions
Necessary actions
Active tools
Process values
Progress values
Status values

Process value 1

c

Process value m
Progress value 1

o r Progress value q
Status value 1

c SRS 2012
(a) Information flow in handcrafting processes (back) and (semi)automated processes (front)

Status value n

c SRS 2012
(b) Definition of a situation S by characteristics c, relations r, and an operator o

c SRS 2012

(c) Example of a situation vector Figure 1. Representation of human-automation systems using SOM [3]

reach a final situation by passing several intermediate situations. In this case the released operator is denoted as meta-operator. Regarding a technical process representing task, several actions are available yielding to a net of situations interconnected by operators building up the action space of the human operator realizing desired process tasks (Fig. 2).
The action space is representing the technical process, the human operator, and the interaction. The interface is implicitly described by the information
(characteristics) and operators included in the model.
Consequently and in contrary to recent developments
[1, 2], the entire interactive system is described by a single modeling method (SOM), which can also be used to implement assistance systems.

S6

S4

Si
S2

S3
S5

Sf

S7
S8

c SRS 2012

Figure 2. Action space example with Si denoting the initial situation and Sf the desired final situation
Detecting filling areas (a,m)

Molding process Filling mold sand

Determining filling volume (a)
Defining filling path (a,m)
Distributing sand (a)
Detecting compression areas (a,m)

2.2. Hierarchical task analysis

Compressing mold sand

Common to analyze human-automation systems is the Hierarchical Task Analysis (HTA). Applying HTA allows structuring and identifying working routines and thereby improve collaboration of human operators and automation systems [12]. In the considered application of the automation of a handcrafting molding process, the HTA is performed by studying the manual manufacturing process. A small group of human workers is performing several repeated steps in order to build up a mold. The analysis leads to the superficial structure illustrated in Fig. 3, whereby the two tasks ‘Filling mold sand’ and ‘Compressing mold sand’ are repeatedly performed until the mold cabinet is completely filled. In fact, these repetitive procedures offer a certain potential of automation as these procedures are process and product quality-related and also skill-/experiencebased, respectively.
In Fig. 3 the whole process is given in detail.
Considering human factors within the given subtasks leads to necessary automation when dealing with the

Determining compression energy (a)
Performing compression (a)
Assessing compression quality (a,m)
Detecting filling status (m)

Leveling surface Assessing filling status (a,m)
Leveling surface (a)
Assessing planeness (a,m)

Figure 3. Superficial HTA of manual nobakemolding process (a - automated, m - manual)

mold sand itself due to hazardous material exposure.
When focusing on economical goals, all mold quality related working steps need to be maintained or improved (by automation). Consequently, HTA leads to sub-processes, tasks [12], scenarios [13], cases [11], or scenes [3] that are defined differently, however all definitions are leading to a discrete framework with certain advantages and disadvantages as already discussed. 108

2.3. From HTA to SOM

The support of the user in approaching and solving the addressed tasks is to improve the decision making process and situation awareness (SA) that need further treatment in order to be capable of user assistance.

Similar to the situation-based task design approach illustrated in [6] which is slightly modified from the scenario-based approach [13], the SOM methodology
[3] allows not only the description of the process itself, but also of the interaction of human and machines.
Also the human’s ability of cognition and learning is considered, which is not further treated in this contribution. Using the HTA as description in form of finite processes, the SOM methodology can be applied by using specific correspondences. The related application shows some advantages in dealing with the complexity of modeled process information in comparison to other works [1, 2].
According to [14], tasks can be mapped to the SOM framework by considering corresponding characteristics or relations within the situation vector, e.g. state of the environment. As measure for the task complexity the number or the complexity of the used characteristics, e.g. Boolean characteristics vs. descriptive strings and their underlying relations can be used. A task can be represented by a situation to re ach by an operator. More complex tasks (with several steps) can be mapped by using meta-operators, whereby the number of changed characteristics or the structural change of the situation can be used as measures of the task’s complexity. From the SOM point of view a task is defined by an initial situation that is transferred by an operator to a final situation solving the task.
The initial situation represents the state of the entire environment and the process, the operator denotes the applied action. The final situation is representing the goal of the applied task. Thus, the task defines the necessary characteristics to build initial and final situation and also the characteristic values that are changed by the task action (operator).
Considering all finite tasks with all the necessary descriptions leads to a resulting situation vector collecting relevant information of the considered (global) process and suitable operators as representatives of available actions (refer to Fig. 2 and Fig. 1(c)).

3.1. Action space-oriented SA
Referring to [15] the SA in terms of user’s understanding consists of acquiring data, comprehension of their meaning, and forecasting their change as time passes by (projection).
In [11] a method merging SA approaches with case-based reasoning (CBR) methods to case-based situation awareness (CBSA) is introduced. The resulting CBSA model has certain parallels to the proposed method. Since the considered application aims semi-automated processes modeled by SOM, tasks are clearly defined by situations to reach and the goal of the process as well, the CBSA-case library that is used for forecasting can be replaced by the action space. The action space represents the entire process and provides necessary and optional actions connecting the discrete framework of process situations. Thus, by available information the situation vector can be built and located within the action space meaning that by measurement and evaluation the current state of process can be determined.
By identifying the current situation suitable and available operations are known with respect to the defined process goal. Furthermore, SA can be assumed as long as human operators are aware of the current location within the underlying action space. Arguing the other way around, the approach allows to assess the user’s SA, if user are conscious of their and the process situation.

3.2. Information filtering methods
In order to support the user’s SA, the unique set of information forming the current situation needs to be available to the human operator. Referring to [15], this process belongs to the comprehension phase of given information. The proposed approach is supporting the comprehension in identifying and filtering of information correspondingly that need to be illustrated in the human-machine-interface (HMI). In contrary to [6] focusing on adaption of the user interface, the proposed method focuses on identifying those information to be displayed in terms of formulated tasks or goals.
The idea of assisting the human operator in comprehension and projection the process state is realized by a situational information filter illustrated in Fig. 4 that is supporting the choice of operators to apply. Here,

3. Situational user assistance by task/model-based information filtering
Complex processes are based on sophisticated user skills in order to perform well during each step of handcrafting processes. So far, the description of the process and the related interactions is available as
SOM-framework. The resulting net is understood as the working process describing action space, in which tasks can be formulated as situations to be reached.

109

o1

o1

o2

o2

(a)

(b)

o1

get situations (applicable if certain characteristics are changed by the available set of operators Fig. 4(b)),
• characteristic differences between target situation
(applicable if a complete set of characteristics is changed in a different way by the available set of operators - Fig. 4(c)), and
• characteristic differences of predefined process quality-related characteristics (applicable if a process quality containing characteristic is changed by the available set of operators - Fig. 4(d)) to be used for supporting the comprehension (by illustrating reduced sets of information) and projection of the current process situation (by illustrating future situations). Applying this method allows reducing information, prioritization of information, or highlighting of information, in order to improve SA of human operators.
The introduced information filter strategies can also be combined to achieve an efficient and problem-related reduction of the necessary information with respect to the actual situation of the complex technical process or the defined task to accomplish.
Similar to [11], the approach allows the human operator to focus on reduced set of information that concentrates on upcoming situations of interest, danger etc. The process-, task- or operator-related information are available in a database structure to be used for assistance and supervision. Here, the user assistance is not realized by restriction of actions, but on an available information base, which action to be performed with respect to the (known) process goal is appropriate.
Regarding the conceptual framework of SOM as tool to address issues in automated processes and with increasing amount of information available or easy to have, human operators can be supported in fulfilling their role in human-automation systems. However, human operators are always understood as keeper of related process knowledge and of manufacturing skills that need to be considered within automation processes. o1

o2
(c)

c SRS 2012

o2

(d)

Figure 4. SOM-based information extraction and filtering with o1 representing the preferred process action based on the current situation the operator to apply
(=process action to perform) o1 in sense of reaching a specific situation is to choose (refer to Fig 4(a)).
Based on the action space of the process and the known characteristics that are of interest describing the specific situation, such situation-specific information can be highlighted correspondingly in the HMI.
Furthermore, the projection of future options of the process and such future states that are desired can be supported by different concepts of information prioritization illustrated in Fig. 4(b-d).
Since the whole action space is modeled in advance in a suitable level of detail, the differences of the characteristics caused by the available process actions are assumed as known. The difference of the possible target situations can be seen as a virtual operator transferring one in the other or vice versa, even though this action is not available. This differentiating operator can be used to formulate the differences in the considered target situations and to detect the necessary information to feed through the HumanMachine-Interface to the user. Especially, in situations, in which process reliability and process safety can not be easily established by automated filtering of the entire characteristics to a set of essential information, suitable operators to support the human operator in the decision making and the active intervention process have to be offered by visual assistance. This approach allows different types of information filter with respect to (or for highlighting of) e.g.
•

4. Application
The production of cast parts using the No-Bake
Technique is usually related to an advanced product line-up combined by a small lot size. These facts combined with the dimension of the cast parts (> 100t) is disadvantageous regarding an automation approach and explains the traditionally established handcrafting process [16]. However, the technical ability to produce such big cast parts in a competitive quality is an unique feature of the considered casting companies.

characteristic differences between initial and tar-

110

Human operator Interface

Process controller i

Process

Process Mimic

STOP

EAs

c SRS 2012
(b) Design of the HMI visualizing the stiffness distribution

(a) Overview of the semi-automated manufacturing system

3D Laser

3D scanning

Leveling rod

Sand mixer

Mold model

Compression tool

Robot

Sand compression

Sand distribution

Multitool

Surface leveling

(c) Illustration of the developed semi-automated manufacturing demonstrator

Figure 5. Process topology of the considered semi-automated manufacturing process and its realization

4.1. Molding process

In the specific application, the model localization as well as the sand level information is traditionally captured by exploring the mold cabinet from different perspectives. In the semi-automated approach this information is captured by 3D-measurements by laser triangulation method. In order to keep the 3D-character of the mold cabinet, the data is used to render a 3Dobject that also can be efficiently canonical manipulated in the Human-Machine-Interface [17].
The representation of the stiffness information is performed using a physical force/stroke measurement, when impulse punching the sand surface on specific cast part dependent locations. The compression distribution is then applied as an interpolated textural mapping onto the visualized 3D-object. The process duration restricts the corrective rework of the human operator, and thus needs to be explicitly considered in sense of improving the situation awareness of the human operator.
In Fig. 5 an overview about the semi-automated manufacturing system (Fig. 5(a)), the developed HMI
(Fig. 5(b)), and the demonstrator (Fig. 5(c)) is shown.
Certain scenarios with skilled employees were investigated with respect to quality related working procedures such as 3D-foundry sand distribution and foundry sand compression.

After the analysis of the handcrafted process, the major segments of the molding process are identified as (i) distributing foundry sand in the mold cabinet,
(ii) uniform compression of foundry sand layers in the cabinet by impulses, (iii) repeating (i) and (ii) until the mold cabinet is completely filled, and finally (iv) leveling the mold surface. However, handcrafting according to the mentioned steps results in a suitable mold quality. The cast part quality is varying as the process has a low repeatability leading to varying manufacturing defects and individual rebuilding concerns. In literature and experiments the sand compression is identified as one major cause for these quality issues [16] meaning that the mold quality is related to an optimal, uniform compression (determined by measured sand package stiffness) during the handcrafted process.
Besides the stiffness of the foundry sand layers, the different variables of dependency are determined in order to give the human operator a complete set of relevant information as necessary to monitor, supervise, and guide the process. The major task of the information and therefore interaction visualization is to realize the model localization, to illustrate sand level/distribution or stiffness distribution, and time remaining for completing the compression. The entire set of illustrated information needs an environmental friendly coding and design also satisfying the cognitive requirements of the interacting human operator.

4.2. User study
The experiment focused on the initial foundry sand distribution and the way, how and where the individual

111

helpful approaches in their working environment. The lack of look and feel was confusing at some instances during the evaluation due to missing feedback of the
HMI. However, the visual feedback was sufficient to finish the process. Concluding from the employees feedback, none of the test subjects were rejecting assistance functionalities as filling path offer or automatically compression energy determination. Even though only relevant information were displayed accordingly to the proposed information filtering methods and prioritization of process relevant characteristics (mold sand stiffness), the test subjects were distracted from the moving parts within the manufacturing cell that sometimes lead to faulty insertion by accident. Concluding from the experiments, there is the need of analyzing and implementing of attention attraction methods [18] and to concentrate even more on usercentered design approaches [19] of the HMI.

skilled employee has to insert the foundry sand into the mold. The assistance system was designed so that there was a given ”optimal” filling path offered to the skilled employees when they were approaching the corresponding situation. All six test subjects in the age of 20-50 had an one hour offline training in using the touch panel HMI (refer to Fig. 5(b)) prior to the experiment with the demonstrator. The test subjects started all the same by firstly scanning the flask content. Then the assistance systems offered to fill in foundry sand by displaying the filling route.
However, since within the environment measuring failures cannot be excluded to happen, the test subjects were asked to check the correctness and proceed with the next process step (filling in foundry sand). The test subjects were asked to start the process as they usually do and in order to prevent mold issues. The given filling path was totally accepted by two test subjects.
One test subject dismissed filling at all and the three remaining test subjects corrected the path remarkable.
Results of selected test subjects are illustrated in Fig. 6.
The filling speed profile (corresponding to the insert volume of foundry sand) is automatically assigned.
This procedure was repeated on several layers with alternating filling compression. This brief examination of course is not representative, however the results are sill valuable from a qualitative point of view.

5. Summary and future work
Within this contribution the Situation-OperatorModel was used to create an integrated model of a
Human-Automation-System. The proposed event discrete approach (SOM) allows the realization of user assistance of complex technical processes based on information filtering methods. The proposed approach relates available information to formulated tasks and reduces model-based information to subsets of interest.
The reduced information sets are used for situational assistance by improving the user’s situation awareness and allow situational guidance of technical processes.
The SOM information prioritization methodology was successfully applied to an engineering (mold manufacturing) example and tested with skilled employees in order to show performance of the method. Performed experiments show reasonable results in the sense of process guiding and achieving task achievement by pointing out differences in the individual process understanding of test subjects. However, the experimental results were only evaluated qualitatively but show some important aspects related to upcoming extensions. Hence, there is the need of designing experiments leading to resilient and measurable performance indicators of guided and unguided test subjects and the resulting mold quality.

Systems default

Orientation

Filling path

Filling path marker Figure 6. Experimental results with skilled employees (red), Systems default (yellow)

4.3. Discussion
Qualitatively, the incremental assistance by the predefined action space supported the user to focus on the next process action and corresponding process supervision. The situational assistance system supported the users efficiently through the molding process, even though there was only a short training session. Each test subject evaluated the semi-automation mold manufacturing cell and its functionalities as suitable and

Acknowledgment
This research was funded by the “Bundeswirtschaftsministerium f¨ r u Wirtschaft und Forschung”,
Germany
entitled
”Entwicklung
eines flexiblen Mensch-Maschine-Systems f¨ r ein u 112

teilautomatisiertes Einf¨ llen, Verteilen und Verdichten u von kaltharzgebundenen Formstoffen in Gießereien”
(08/09-12/11).

[10]

References
[1] D. Shin, R. A. Wysk, and L. Rothrock, “A formal control-theoretic model of a human automation interactive manufacturing system control,” International Journal of Production Research, vol. 44, no. 20, pp. 4273 –4295, 2006.
[2] S. Combefis, D. Giannakopoulou, C. Pecheur, and
M. Feary, “A formal framework for design and analysis of human-machine interaction,” in IEEE
International Conference on Systems, Man, and
Cybernetics (SMC), 2011, pp. 1801 –1808.
[3] D. S¨ ffker, “Interaction of intelligent and auo tonomous systems - part I: qualitative structuring of interaction,” Mathematical and Computer modelling of dynamical systems, vol. 14, no. 4, pp. 303 –318, 2008.
[4] A. Shepherd, “HTA as a framework for task analysis,” in Task Analysis, J. Annett and N. A.
Stanton, Eds. London: Taylor and Francis, 2000, pp. 9 –24.
[5] T. Sheridan, “Adaptive automation, level of automation, allocation authority, supervisory control, and adaptive control: Distinctions and modes of adaptation,” IEEE Transactions on Systems,
Man, and Cybernetics, Part A: Systems and Humans, vol. 41, no. 4, pp. 662 –667, july 2011.
[6] N. Nwiabu, I. Allison, P. Holt, P. Lowit, and
B. Oyeneyin, “User interface design for situationaware decision support systems,” in IEEE International Multi-Disciplinary Conference on Cognitive Methods in Situation Awareness and Decision Support (CogSIMA), 2012, pp. 332 –339.
[7] R. Parasuraman, T. Sheridan, and C. Wickens, “A model for types and levels of human interaction with automation,” IEEE Transactions on Systems,
Man, and Cybernetics, Part A: Systems and Humans, vol. 30, no. 3, pp. 286 –297, may 2000.
[8] T. Helldin and G. Falkman, “Human-centered automation for improving situation awareness in the fighter aircraft domain,” in IEEE International Multi-Disciplinary Conference on Cognitive Methods in Situation Awareness and Decision
Support (CogSIMA), 2012, pp. 191 –197.
[9] A. Fereidunian, C. Lucas, H. Lesani, M. Lehtonen, and M. Nordman, “Challenges in implementation of human-automation interaction models,” in Mediterranean Conference on Control

[11]

[12]

[13]

[14]

[15]

[16]

[17]

[18]

[19]

113

Automation, 2007. MED ’07., june 2007, pp. 1
–6.
R. Breton, E. Bosse, R. Rousseau, and S. Tremblay, “Framework for the analysis of information relevance (fair),” in IEEE International MultiDisciplinary Conference on Cognitive Methods in Situation Awareness and Decision Support
(CogSIMA), 2012, pp. 210 –213.
N. Nwiabu, I. Allison, P. Holt, P. Lowit, and
B. Oyeneyin, “Case-based situation awareness,” in IEEE International Multi-Disciplinary Conference on Cognitive Methods in Situation Awareness and Decision Support (CogSIMA), 2012, pp.
22 –29.
J. Tan, F. Duan, Y. Zhang, and T. Arai, “Task decomposition of cell production assembly operation for man-machine collaboration by HTA,” in IEEE International Conference on Automation and Logistics, 2008. ICAL 2008., sept. 2008, pp.
1066 –1071.
S. P. Robertson, “Generating object-oriented design representations via scenario queries,” in
Scenario-based design, J. M. Carroll, Ed. New
York, NY, USA: John Wiley & Sons, Inc., 1995, pp. 279 –308.
B. Kirwan, A Guide To Task Analysis: The Task
Analysis Working Group.
Taylor & Francis,
1992.
M. R. Endsley, “Toward a theory of situation awareness in dynamic systems: Situation awareness,” Human factors, vol. 37, no. 1, pp. 32 –64,
1995.
D. S¨ ffker, X. Fu, M. Langer, and D. Gamrad, o “Bedienerunterst¨ tzung und -¨ berwachung bei u u komplexen interaktiven Prozessen,” in Innovative
Interaktionstechnologien f¨ r Mensch-Maschineu
Schnittstellen.
Berlin: Deutsche Gesellschaft f¨ r Luft und Raumfahrt (DGLR), 52. Fachausu schusstagung, 7. - 8. Oktober, 2010.
D. A. Bowman, E. Kruijff, J. J. LaViola, and
I. Poupyrev, 3D User Interfaces: Theory and
Practice. Redwood City, CA, USA: Addison
Wesley Longman Publishing Co., Inc., 2004.
C. Wu, Y. Lin, and W. J. Zhang, “Human attention modeling in a human-machine interface based on the incorporation of contextual features in a bayesian network,” in Proc. IEEE Int Systems, Man and Cybernetics Conf, vol. 1, 2005, pp. 760 –766.
D. Z¨ hlke, Useware-Engineering f¨ r technische u u
Systeme. Springer, 2004.