#StandardsGoals for 2024: What’s new for BISAC - Tech Forum 2024
SOCIAL ADAPTATION OF ROBOTS FOR MODULATING SELF-ORGANIZATION IN ANIMAL SOCIETIES
1. SOCIAL ADAPTATION OF ROBOTS FOR MODULATING
SELF-ORGANIZATION IN ANIMAL SOCIETIES
SASO, FOCAS workshop, 8th Sept. 2014
Payam Zahadat, … ,Thomas Schmickl
Artificial Life Lab, Karl-Franzens University Graz, Austria
Email: thomas.schmickl@uni-graz.at, Twitter: @thomasschmickl
2. ASSISIBF
Animal and robot Societies
Self-organize and Integrate
by Social Interaction
BF: bees & fish
http://assisi-project.eu twitter:@AssisiEU
8. TYPES OF INTERACTION (CHANNELS)
Sensing:
• Temperature
• Proximity & Touch
• Vibration
Actuation
• Temperature
• Vibration
• Light
• Electric field
• Magnetic field
Adaptation
• Self-Organization
• GRN
• ANN
• Artificial Hormone Systems
• Evolutionary Computation
• Machine Learning
• Modelling (incl. automatized model building)
CASU
13. CASU DESIGN
(a) Top-side view
(b) Cross-section view
(c) A thermal photo showing two CASUs in which one
emits heat stimulus and the other is inactive
setups, A. Global For agents’ by the The society the and were agents. technique noise a combinatorial distance
18. A NEW PARADIGM IN EVOLUTIONARY COMPUTATION:
programmable,
evolvable
Not programmable,
but predictable
Evolutionary robotics: Natural swarm-systems:
Swarm robotics:
Novel
Paradigm:
19. A NEW PARADIGM IN EVOLUTIONARY COMPUTATION:
Novel
Paradigm:
20. THE FIRST PROOF-OF-CONCEPT MODELS
Stimulus Temp. Light Vibration
Effect attractive repellent stop-signal
Diffusion rate 0,2 0 0,01
Decay rate 0,1 1 0,9
Instantly
no yes no
reachable
Blockable by bee no yes no
Three actuators (floor patches) in
every cell generate 3 different
types of stimuli:
A: Temperature
B: Light
C: Vibration
Bees react differently to different stimuli:
- If the cell is vibrating over a given threshold à Stay
- Otherwise: Let the strongest stimulus win à Go
1D model
21. EXAMPLE FOR EVOLUTIONARY CONCEPT
formation and growth. It
applications [17] and
behaviour [18]. An
and rules. The
concentrations based
Concentrations of
of the system.
animals
CASUs in a two-dimensional
resulting patterns are
repellent stimuli
⇥ 9 grid with
cells). This fairly
how the controllers
an ecologically
j=1Σ
Time-steps (top to bottom)
arena cells (one-dimensional)
(a) unevolved
Fig. 6: arena Fitness cells (one-progress dimensional)
in evolution of AHHS controllers.
that can induce complicated animal behaviours. Since it is not
feasible to run such high numbers of iterations in practice, this
type of research calls for a simulator that can run simulations
featuring up to a hundred entities on a time scale of seconds minutes. On the other hand, the simulations should be sophis-ticated
enough that the evolved controllers can be applied a real system, possibly with only minor modifications. We are
developing such a tool, as an extension to the Enki open source
simulator [21]. Our simulator can currently model several types
of physical interactions, such as proximity, light and heat. The
source code is available on GitHub [22], licensed under the
permissive LGPL (b) evolved
license.
fitness = distmax − dist j
Nbees
formation and growth. It
applications [17] and
behaviour [18]. An
hormones and rules. The
concentrations based
Concentrations of
outputs of the system.
with animals
CASUs in a two-dimensional
resulting patterns are
repellent stimuli
⇥ 9 grid with
cells). This fairly
how the controllers
an ecologically
Time-steps (top to bottom)
arena cells (one-dimensional)
(a) unevolved
arena cells (one-dimensional)
(b) evolved