RockEU2 WP3 Cognitive Systems Coordination
Contents
- 1 Objectives
- 2 Description of Work
- 3 Deliverables
Objectives
The key goal is to accelerate the deployment of cognitive systems in industrial applications and to identify the technologies that most urgently need further development. In doing so, we address the “systems-component” dilemma faced by R&D&I in this area. Like R&D&I in operating systems and distributed systems a complete system is required for effective validation. The objective is to strike the right balance between a focus on systems engineering and specific functionality. Where possible, this work will be carried out in conjunction with the euRobotics Topic Group AICoR (Artificial Intelligence and Cognition in Robotics), as well as the topic groups Natural Interaction with Social Robots, Software Engineering, System Integration, and Systems Engineering.
- Foster effective working relationships between producers and consumers of cognitive systems technologies promoting transfer of technology and measuring the extent to which this happens.
- Identify opportunities for enhancing existing products and services with cognitive capabilities.
- Establish a reference RockEU2 Cognitive Architecture Schema
- Produce a definitive catalogue of existing algorithms and matching data structures – Numerical Recipes for Cognitive Systems – mapped to the RockEU2 Cognitive Architecture Schema.
- Identify game-changing techniques that have the potential to create a watershed in the uptake of cognitive systems technology.
- Identify attributes cognitive systems need to exhibit in robots that can engage safely, naturally and effectively with humans in the home, the workplace, and in social settings generally.
- Develop a framework for linking cognition and robot autonomy, establishing a mapping from different forms and levels of autonomy to related cognitive processes.
Description of Work
Task 3.1: Industrial Priorities
Partners: HIS, TU-WIEN, RUR, KUKA.
Duration: M1 – M18
Task Description
The goal is to identify the functional needs for cognitive robotics from the perspective of industrial developers and categorize them as essential (for competitive advantage), desirable, or ideal (i.e. desirable but optional). These needs will naturally be expressed in terms that are specific to certain industrial and business sectors. Consequently, this exercise will map these needs to underlying technical capabilities. Non-functional requirements (e.g. safety, dependability, interoperability, etc.) will be addressed where appropriate. These needs will be established by direct consultation and the formation of interactive focus groups. Where possible, metrics for measuring the degree to which the required attributes are present in a system will be developed.
An important aspect of this task is to identify game-changing techniques that have the potential to create a watershed in the uptake of cognitive systems technology
Impact and Monitoring
Knowing industry’s priority functional needs allows the community to target the matching techniques more effectively, either by deploying existing ones or developing the requisite enhancements. The impact of this is more effective of technology transfer and moving the industrial and academic community closer to the tipping point where cognitive robotics become mainstream technology in that it routinely enables better performance and new capabilities.
Monitoring requires two validation feedback loops. One is concerned with making sure that what is expressed in the list of priorities is in fact what the industrial community actually means. The other is concerned with translating these priorities into new applications or system behaviour that can be validated against the stated priority requirements. The first will be addressed in this task. The second will be effected in Task 3.3.
Role of Partners
HIS will take the lead in this task: it will organize four four-monthly focus group meetings with participants from industry, during which these priorities are framed in terms that are relevant to industry needs and useful for driving technology transfer and technology development, providing input to Tasks T3.3, T3.4, and T3.5.
Resources
The centrally administered expert fund will be utilised to elicit support from national experts. A central travel budget allocation is reserved to support third parties attending standards meetings.
Task 3.2: Catalogue of Cognitive Systems Capabilities
Partners: HIS, TU-WIEN, RUR, iTechnic.
Duration: M6 – M18
Task Description
The specific aim of this task is to produce a definitive catalogue of existing algorithms and matching data structures – Numerical Recipes for Cognitive Systems. Two categories of future capability will be exposed: one including software capabilities that have a high level of maturity (i.e. technology readiness level TRL) but require a small amount of further development and those with a lower TRL but with a very high potential payoff.
The capabilities will be organized according to a new RockEU2 taxonomy of functional capability. This taxonomy will draw on existing taxonomies in the Multi-Annual Roadmap (MAR) and the euRobotics Topic Group on Artificial Intelligence and Cognition in Robotics (AICoR). Specific attention will be paid to differentiating the requirements of cognitive robots that act autonomously with little or no interaction with people and those that are specifically focused on interaction with people, either as assistants or as equal partners.
Impact and Monitoring
Knowing what techniques are available and ready for deployment makes it much easier for application developers to integrate them into their products. This will have the impact of more widespread and more efficient use and validation of those techniques. This will leverage even greater impact by setting up a positive feedback loop from users to providers, further advancing the process of technology transfer.
Progress in this task will be monitored by creating the catalogue on a publicly accessible wiki and periodically seeking feedback from the core group of contributors.
Role of Partners
HIS will take the lead in this task: it will organize four three-monthly focus group meetings with participants from academia, during which candidate techniques are identified based on a broader canvass of the cognitive robotics community. These will then be fully documented on a dedicated wiki. This task provides essential input to Tasks T3.1, T3.3, T3.4, and T3.5.
Resources
The centrally administered expert fund will be utilised to elicit support from national experts. A central travel budget allocation is reserved to support third parties attending standards meetings.
Task 3.3: RockEU2 Cognitive Architecture Schema
Partners: HIS, TU-WIEN.
Duration: M1 – M24
Task Description
A cognitive architecture schema is not a cognitive architecture but a blueprint for the design of a cognitive architecture, setting out the component functionality, control flows, and mechanisms for specifying behaviour. It describes a cognitive architecture at a level of abstraction that is independent of the specific application niche that the architecture targets. Its purpose is to define the necessary and sufficient components and the organization required to create a complete cognitive robot system.
The goal is to consolidate existing knowledge to create a cognitive architecture schema that can be used both to provide a context for constituent cognitive system algorithms and the functionality they encapsulate and to identify those aspects that require the greatest attention in the short-term to develop effective cognitive robots. As such, it provides an effective way of organizing the information produced in Task T3.2. In doing this, certain difficult questions will be addressed concerning, for example, whether or not it is feasible to design a general reference schema, what is required to develop a re-configurable cognitive architecture, what are the software engineering implications, and whether any existing models can be used as a starting point (e.g. the CoSy Architecture Schema and Toolkit CAST [Hawes and Wyatt 2008] or the Cognitive Robot Abstract Machine CRAM [Beetz et al 2010]).
Impact and Monitoring
Effective deployment and reuse of cognitive capabilities depends on a degree of interoperability between components and the existence of a reference framework in which these components can be integrated. By establishing such a framework in the guise of a reference cognitive architecture schema, this task will promote reuse and deployment of cognitive capabilities in robot systems.
The plan is to create this schema collaboratively with a small core group of system architects meeting every six months. The schema will be documented on a public wiki so that progress will be clearly visible and open to scrutiny; comments and contributions by all interested parties will be solicited.
Role of Partners
HIS will take the lead in this task, exploiting established expertise in the domain of cognitive architectures. TUW and BRSU will assist.
Resources
The centrally administered expert fund will be utilised to elicit support from national experts. A central travel budget allocation is reserved to support third parties attending standards meetings.
Task 3.4: Cognition-Autonomy Framework
Partners: HIS, TU-WIEN.
Duration: M1 – M18
Task Description
The aim is to develop an organising framework for linking cognition and robot autonomy so that one can map from the many different forms of autonomy to related cognitive processes. This in turn will establish how various cognitive systems processes contribute to the autonomy of a robot. An important aspect of this framework will be to address the different degrees of autonomy that a robot can exhibit and how this can change as the task the cognitive robot is doing evolves and as the needs of the people it is interacting with alter.
The motivation for this task is the possibility that the terminology used when discussing autonomous systems may be a more natural way for users and robot developers to express their needs, compared with the terminology of cognitive systems.
The language we use when discussing autonomy tends to focus on what the system does and why, whereas discussions of cognitive systems tend to revolve around how they do these things. This may not be helpful to roboticists who are concerned with using cognition to achieve a desired behaviour or action and are less concerned with the mechanisms by which this is achieved. The goal is to explore the utility of casting requirements in terms of autonomy and then mapping these to requirements in terms of cognitive systems.
Impact and Monitoring
If successful, this task will result in a new way of expressing the requirements of industrial robotics for cognitive capabilities, one that will map to the underlying computational model required to deliver it. The impact will be significant in that it will make it easier for industrial roboticists to identify the needs of their systems and make technology transfer more efficient, effective, and less prone to failure.
As with the previous task, the goal is to establish this framework collaboratively with a small core group of system architects meeting every six months. To be efficient, these meetings will be held at the same time as the architecture schema workshops. Similarly, the framework will be documented on a public wiki; comments and contributions by all interested parties will be solicited.
Role of Partners
HIS will take the lead on this task, with the assistance of TUW and BRSU.
Resources
The centrally administered expert fund will be utilised to elicit support from national experts. A central travel budget allocation is reserved to support third parties attending standards meetings.. 
Task 3.5: Software Engineering Factors in Cognitive Robotics
Partners: BRSU, HIS, TU-WIEN.
Duration: M9 – M21
Task Description
The goal of this task is to develop a set of guidelines dealing with software engineering factors that are specific to cognitive robotics, including interoperability in robot development platforms such as ROS, OROCOS, and YARP. This task will draw heavily on previous work on Best Practice in Robotics BRICS [Bruyninckx et al. 2013] and will set out the essential, desirable, and ideal attributes of well- engineered cognitive robotics software. Agent-based approaches, rule-based systems, component- based software engineering, and model-driven engineering will be addressed. The goal is not to develop new approaches but to present existing ones in a way that lowers the barrier to entry and take-up. This task complements work in T3.3 on the RockEU reference cognitive architecture schema.
Impact and Monitoring
This task aims, though the adoption of effective software engineering guidelines, at producing guidelines that will significantly increase the reuse of software that encapsulates cognitive capabilities. The goal is to make it easier to adhere to these guidelines and the impact of doing so will be the adoption of better practices and the creation of a body of interoperable components capable of being integrated in a single system.
The guidelines, together with example code and documentation, will be hosted of public wiki so that progress will be visible and open to scrutiny; feedback from all interested parties will be solicited.
Role of Partners
BRSU will lead this task, with the assistance of HIS and TUW.
Resources
The centrally administered expert fund will be utilised to elicit support from national experts. A central travel budget allocation is reserved to support third parties attending standards meetings.
Deliverables
Each of the tasks above will produce a deliverable. The title of the deliverable reflects the title of the task. The description of the deliverable can be inferred directly from the task description.
- D3.1 Industrial Priorities for Cognitive Robotics
- D3.2 Catalogue of Cognitive Systems Capabilities
- D3.3 RockEU2 Cognitive Architecture Schema
- D3.4 Cognition-Autonomy Framework
- D3.5 Software Engineering Factors in Cognitive Robotics
