Artificial immune systems and physiologies

1
Artificial immune systems and physiologies

• Mark Neal

2
The immune system

• Tightly integrated defence and monitoring
• Multiple layers
• Redundant, degenerate and pleiotropic actors
• Self organizing, very complex, huge numbers
• Layers
• Innate
• Adaptive

3
Innate layer

• Lots of variety
• Complement system
• Toll like receptors on cells
• Macrophages
• Neutrophils
• Mast cells
• Etc
• Specific responses to particular threats
• Sends signals

4
Adaptive layer

• A bit less variety, but learns
• Bcells
• Produce antibodies
• Are selected for when stimulated
• Tcells
• Activated in lymph nodes
• Go around gobbling up pathogens
• Activated by innate layer
• Cytokines (endocrine signals)
• Antigen presentation

5
So whats my vision?

• Robot control systems that treat the robot
  hardware as an artificial physiology
• Use INE controllers to do this
• Reap the benefits
• Self-organization
• Flexibility
• Adaptability
• Homeostatic properties
• Robots (and other systems) that look after
  themselves

6
How far have we got?

• ANN and AIS work fairly well
• ANN reusable as they are
• AIS need adaptation and extension
• AES are rudimentary and need development
• Can generate some useful behaviour
• Behavioural trade-offs
• Controlled oscillation

7
How Did We Get Here?

• Carefully looking at biology
• and robotics
• Taking the operating principles and teasing them
  apart
• Performing experimental analysis of the
  algorithms at work in robots
• Can INE control generate long term autonomy in
  robotic systems?