Industrial Applications of Self-organizing Systems

Task description

• Discuss the need for self-organizing systems in industrial applications
• Give examples of success stories/potentials
• What is missing to "make things work"?

Results

Industrial View

• Collaboration has to improve the Business Case
• Return on investment
  • Someone has to pay for it
  • Add value
• Who is in control up to in the end?

Basis for collaboration

• Bring together industry and academia
  • Two different languages
  • Partnerships
• Early transfer to the industry
  • NOTE: easier to justify than in the industry
• Present a benefit to the industry: way to make money
  • New concepts difficult to accept; conservativeness
  • Common set of working assumptions
  • The power of a business plan
  • Do not switch research agenda

What is missing?

• Research is there; sometimes too advanced
• Assurance of method properties is a necessary prerequisite for application of SOS methods in industrial products/settings
• Clear understanding on the application of self-organizing principles to the specific application
  • internet-based applications: interaction between human self-organization and overlay network self-organization
  • vehicular communications: micro-macro dependencies
  • future wireless networks: self-organization of heterogeneous networks
  • smart grid: self-organization up to a point, but needs to be able to react to severe oscillations
• Communication between industrial and academic needs to be improved
  • Research Days
• Open questions:
  • Who is control up to in the end?
  • Timeliness and external constraints (regulation)
• Assurances that self-organising behaviour works
  • Formal proofs, formal assurances, safety
  • We cannot give this at the micro level, so provide macro-level behaviour assurances
• Need for Education - Training
  • Change the way of thinking
  • Eg. Stability may not be ensured, but the system works
  • Provide the engineers with a black box that “works” at the macro-level
• Industry should give it a try
• Learning by doing (improve the way we show it works)
• Design Patterns (describing problems)
• To make things work, we need to start applications from scratch with self-org techniques to achieve the same functionality
  • (eg. Traffic control, computers)

Success stories

• LTE?!
• Power distribution
• Internet-based industrial applications
• Vehicular networking
• Honey Bees derived algorithm.
  • Used for aggregating robots with low computational power to find lightspot on the table
  • Experiments and simulations
  • The bio-inspired code is "superior" to traditional code as it allows adaptation (e.g. to obstacle, many light sources) without further modifications.
  • Problem: not possible to guarantee that it will work
  • Engineers do not like this
• Internet: not perfect but we accept this.

Self-Organisation may not apply everywhere

• The car is not complex but complicated, and we don't necessarily need self-organisation for the car
• There are applications where self-organisation does not apply

Where to use self-organisation?

• Lower-level parts are built traditionally following traditional engineering principles
• Higher-level system follows self-organisation principles
  • Ex. Individual displays, computer, keyboard work, together they follow a self-org approach
  • Car traffic control
• For systems where we can accept impreciseness (eg. Wikipedia), we gain more – acceptance of uncertainty
• When there is uncertainty, unforeseen situations