Design Patterns for Cooperation Motivation
Design Patterns
The Catalogue
The research done in WP3.3 results in three different, but closely linked cross-domain solutions:
A design pattern is not a finished design. It is a template for solving special problems, which can be used in many different design projects.
Develop crossdomain reusable reference designs and design patterns for DCoS !
History of Design Patterns: • Solving problems in urban architecture (Alexander, 1977) • Software engineering system design (Beck et al., 1987) • Object oriented software design (Gamma et al., 1995) • Interface & interaction design (Borchers, 2000)
The overall design patterns catalogue has been established for cooperation, state inference and adaptation, and user interfaces as a hypertext database:
Agent Cooperation
Catalogue State Inference & Adaptation
Multimodal User Interface
Design Pattern (DP): proven solution for a recurring problem (method) Reference Design (RD): exemplary instantiation of one or more Design Patterns (techniques)
System • Initiation • Maintenance • Completion
State Adaptation
• Attention • Workload • Situation Awareness
• Arbitration • Transition • Resource Allocation
(2014) (2014)
Benefit (Gamma, 1995; Beck et al., 1996): Common design vocabulary Documentation and learning aid Team communications medium Capture essential parts Adjunct to existing methods Extracted from working design Reusable best practice
This is available online:
D3CoS Approach to Design Patterns:
Vision: Patterns for DCoS design State Inference
• HMI design: • Cooperation design:
User Interfaces • Mode • Group • Direction • Modality
Design Pattern Catalogue
Scenario Catalogue Scenario n Scenario 2 Scenario 1
Design pattern n Design pattern 2
Scenarios relevant for D3CoS
selected design patterns with design requirements
is part of
Agent Cooperation
Safety Efficiency Comfort
influence
Design pattern relevant for D3CoS
...
The following patterns have been created for the design of cooperative multi-agent systems in the phases initiation, maintenance, and completion:
influence design pattern selection criteria
Unique Scenario
influence
updates
Dynamic system design / Functions
Interaction Problem
Composition
Interface Problem
Problem
DCoS
Bottom-up generalization: document
Top-down specialization: implement
Current dynamic Configuration
is part of
TARF’s
Reference dynamic Configuration
Analysis
influence structures
⟶Problem Awareness ⟶Role Allocation ⟶Explicit Addressing ⟶Allocation to Cooperative Population ⟶Awareness of Own Task and Goal ⟶Awareness of Partners’ Task and Goal ⟶Certainty of Tasks and Goals ⟶Rewarding
updates
Unique System Static system design / Framework
Current static/ framework configuration
Problem Categories Catalogue
Modification
Reference static/ framework configuration
Problem category n Problem category 2
Problem category relevant for D3CoS
Semantic and structural interplay between design patterns and system design in D3CoS
Contact Information Dipl.-Medieninf. Markus Zimmermann Institute of Ergonomics Technische Universität München Boltzmannstraße 15 85747 Garching Tel +49 89 289-15375
[email protected] www.ergonomie.tum.de
Methods, Techniques, Tools This is a …
X Method
Technique
Method
Design Patterns
Technique
-
Tool
Catalogue of Design Patterns
X Tool
Sections: • Title: Conveys the central idea • Rank: Confidence and alternatives • Picture: Application or scheme • Context: Larger scale patterns • Problem Statement: Addressed situation • Problem Description: Design problem • Solution: Central method for solution • Diagram: “If you can’t draw it, it’s not a DP” • References: Smaller scale patterns • Literature & Authors
Explicit Addressing Problem: • Diffusion of responsibility • Selection of non-optimal agent Solution: Determine optimal and unique cooperation partner I ACT!
References:
Other Agent/ Problem Environment/ Resource Trigger
Alexander, C., Ishikawa, S., & Silverstein, M. (1977). A pattern language: Towns, buildings, construction. New York: Oxford University Press. Beck, K., & Cunningham, W. (1987). Using pattern languages for object-oriented programs. Technical Report CR-87-43, Tektronix, Inc. Presented at the OOPSLA'87 workshop on Specication and Design for Object-Oriented Programming. Beck, K., Crocker, R., Meszaros, G., Vlissides, J., Coplien, J. O., Dominick, L., & Paulisch, F. (1996). Industrial experience with design patterns. In Proceedings of the 18th international conference on Software engineering (pp. 103–114). Berlin, Germany: IEEE Computer Society. Borchers, J. O. (2000). A pattern approach to interaction design. In Proceedings of the 3rd conference on Designing interactive systems: processes, practices, methods, and techniques (pp. 369–378). New York City, New York, United States: ACM. Gamma, E. (1995). Design patterns: Elements of reusable object-oriented software. Reading, Mass: Addison-Wesley.
Cooperative Agent 1 Acknowledgement
YOU ACT!
Cooperative Agent 2
Consortium
Acknowledgments This research has been performed with support from the EU ARTEMIS JU project D3CoS (http://www.d3cos.eu) SP-8, GA No.: 269336. Any contents herein are from the authors and do not necessarily reflect the views of ARTEMIS JU. France
Germany
Design Patterns for State Inference & Adaptation Motivation State inference and system adaptation mechanisms allow an optimized task allocation and shared authority among DCoS agents: How to measure the human or machine agents‘ state?
State Adaptation Task Allocation and Resource Functions
State Inference Metrics
State Inference & Adaptation
How to adapt the system‘s state?
The following design patterns have been elaborated for the human state inference and machine state adaptation aspect:
⟶State Inference ⟶Human Attentional State Inference ⟶Human Workload State Inference ⟶Workload Visual State Inference ⟶Human Situation Awareness Inference ⟶Task Allocation and Resource Functions ⟶Conflict Solving and Replanning ⟶Resource Allocation ⟶Distributed Allocation ⟶Central Allocation
State Inference
State Adaptation
Human operator state inference relies on behavioural, physiological and cognitive metrics to predict the degradation of human/system interaction as well as the operator’s state. Those metrics were further developed towards methods: How to predict those DCoS states?
State Adaptation includes mechanisms and decision rules how to do a dynamic DCoS transition like a shift of responsibilities or resources. Those methods define basic task allocation and resource functions (TARFs), describing how to reallocate tasks, agents, or resources.
Human Workload State Inference
Distributed Resource Allocation
Problem: Keep a regular workload Solution: Integrate objective and subjective metrics • Performance measures (Errors, …) • Self-assessed measures (NASA-TLX, …) • Psychophysiological measures (HRV, …)
Problem: Resource in conflict for ad-hoc cooperation partners Solution: Transfer of distributed algorithm by Ricart & Agrawala • Agent holding resource is resource manager • All agents must request resources • Any scheduling is possible
Application Story: • Application of reference design in UAV and MARITIME experiments in all project cycles • Bottom-up update of the design pattern Workload Visual State Inference Problem: Real-time-capable set of metrics Solution: Integrate visual metrics
Mining Story: 1. Requirement: Adaptation (UAV, MAV) 2. Collection of system design problems in DP-T:
Contact Information Dipl.-Medieninf. Markus Zimmermann Institute of Ergonomics Technische Universität München Boltzmannstraße 15 85747 Garching Tel +49 89 289-15375
[email protected] www.ergonomie.tum.de
3. Scenario of the design problem in the UAV domain: cooperative UAVs observe an area of interest and transmit regularly the acquired sensor data (e.g. high res images) to the ground control station (GCS). Specific problem: Bandwidth 4. Generic interaction problem derivation: How to share a resource 5. Cross-domain problem derivation: Resource in conflict 6. Discussion of the problem cross-domain: Assign, decide, criteria, manner, … 7. Top-down theory driven solution research
Mining Story: Validation in AUT experiments
Methods, Techniques, Tools This is a …
X Method
Method
Design Patterns
Technique
-
Tool
-
Technique
Tool
Zimmermann, M., Rothkirch, I. M., & Bengler, K. (2014). Reading the Driver: Visual Workload Assessment in Highly Automated Driving Scenarios. Proceedings of the 5th International Conference on Applied Human Factors and Ergonomics (AHFE 2014), 2014.
Consortium
Acknowledgments This research has been performed with support from the EU ARTEMIS JU project D3CoS (http://www.d3cos.eu) SP-8, GA No.: 269336. Any contents herein are from the authors and do not necessarily reflect the views of ARTEMIS JU. France
Germany
Design Patterns for Cooperative User Interfaces Motivation Cooperative interaction via multimodal user interfaces targets the reduction of complexity and the communication of agents’ intentions: Cooperative Information
How to communicate cooperative aspects in dynamic systems
Information Modality
Cooperative UI
Cooperative Information
Information Modality
We transferred design patterns for cooperation to user interface design patterns for communicating cooperative information; such as Explicit Addressing to Directed Information or Allocation to Cooperative Population to Group Information.
We researched several design patterns concerning different information modalities in the demonstrator experiments: Multimodality Problem: Human operator has limited attentional resources; specific information modalities may not be available. Solution: Multimodal fusion • Distribute information over interfaces • Link modality to action and attention • (Re)capture attention multimodal • Escalate modalities • Relieve overused channels
Action Suggestion UI How to choose the appropriate information modality in cooperative situations?
The following design patterns have been elaborated for cooperative user interfaces:
⟶Cooperative User Interfaces ⟶Information Modality ⟶Ambient Information ⟶Augmented Reality ⟶Multimodality ⟶Mode Information ⟶Group Information ⟶Directed Information ⟶Mutual Control UI ⟶Action Suggestion UI
Problem: Information about state of traffic and available tasks and resources needed for performing a task. Solution: Suggest actions which can be accepted or declined (common course) • Suggest • Resource, e.g. lane • Task, e.g. lateral control • Interact: Arbitrate, execute, communicate • Conclude
Mining Story: Top-down research, issues tested in AUT. Ambient Information
Mining & Application Story: AUT experiments across different modalities for refinement of DP
Problem: Visual attention; focus overused. Solution: Use peripheral vision
Experience:
Augmented Reality Zimmermann, M., Bauer, S., Lütteken, N., Rothkirch, I. M., & Bengler, K. (2014). Acting Together by Shared Control: Evaluating a Multimodal Interaction Concept for Cooperative Driving. The 2014 International Conference on Collaboration Technologies and Systems (CTS 2014), 2014.
Problem: Gain attention during cooperative situation Solution: Use augmented reality
Mode Information Problem: Mode of cooperation and machine intention need to be communicated by the interaction without mode error. Solution: Communicate mode information • What is the current mode (actions, space) • Why is the system in that mode (explan.) • What will the system do next (time) • Who is the cooperation partner
Contact Information Dipl.-Medieninf. Markus Zimmermann Institute of Ergonomics Technische Universität München Boltzmannstraße 15 85747 Garching Tel +49 89 289-15375
[email protected] www.ergonomie.tum.de
X Method
Method
Design Patterns
Technique
-
Tool
-
Technique
Mining & Application Story: AUT
Group Information Problem: How to communicate group membership Solution: • Connect graphically • Link iconographically • Relate by shape / colour
Mining & Application Story:
Mining & Application Story: • Specialization of DP-COOP Allocation to Cooperative Population • Bottom-up AUT and UAV derivation
Methods, Techniques, Tools This is a …
Mining & Application Story: MARITIME
Tool
Zimmermann, M., & Bengler, K. (2013). A Multimodal Interaction Concept for Cooperate Driving. In 2013 IEEE Intelligent Vehicles Symposium (IV) (pp. 1285–1290).
Consortium
Acknowledgments This research has been performed with support from the EU ARTEMIS JU project D3CoS (http://www.d3cos.eu) SP-8, GA No.: 269336. Any contents herein are from the authors and do not necessarily reflect the views of ARTEMIS JU. France
Germany