Simulation of Adaptive Behaviour Preface of 1st Conference, Meyer & Wilson [7]:

``The object of the conference was to bring together researchers in ethology, ecology, cybernetics, artificial intelligence, robotics, and related fields so as to further our understanding of the behaviours and underlying mechanisms that allow animals and, potentially, robots to adapt and survive in uncertain environments.''
 

Adaptation
 $[$L. adaptare -- ad, to, and aptare, to fit$]$ Britannica, 1959:
Adaptation, a process of fitting, or modifying, a thing to other uses and so altering its form or original purpose. Thus in literature there may be an adaptation of a novel or drama$\ldots$ In biology, adaptation plays a prominent part as the process by which an organism or species becomes modified to suit the conditions of its life. Every change in a living organism involves adaptation; for in all cases life consists in a continuous adjustment of internal to external relations. Some adaptations are produced afresh in each generation, others are transmitted by heredity, having been probably fixed by selection.

change = adaptation for natural organisms?
 

J.H. Holland, Adaptation in Natural and Artificial Systems [5]

• Optimisation made difficult by substantial complexity and uncertainty.
• Complexity $\Rightarrow$ discovering optimum takes ages or never: must exploit best among tested options.
• Uncertainties must be reduced rapidly, so knowledge of available options increases.

``There is no collective name for such problems, but whenever the term adaptation appears it consistently singles out the problems of interest.''

• Formal distinction between adaptive system and its environment.
• Notion of an adaptive plan.

Adaptive Behaviour
 Cliff, Harvey, Husbands 1992 [2]:

``nervous systems evolved where they generated adaptive behaviours (i.e. behaviours which are likely to increase the chances that the individual animal survives to reproduce). We, in common with a growing number of other researchers, believe that the generation of adaptive behaviours should form the primary focus for research into cognitive systems, and that issues of purely transforming between representations or encodings are, at best, secondary.''

• Circular def when ``evolving adaptive behaviour''?
• What could `behaviour' be?
• Collective behaviours?
• Body as part of `behaviour' of genes?
• Extended phenotype, etc.
• `Adaptive' in mainstream engineering.
• Adaptation vs. self-adaptation.

Simulation
 Webster's New International Unabridged 1961:

simulate$[$L simulatus, part part of simulare to imitate, represent, feign, fr. similis like, similar$]$1: to give the appearance or effect of: FEIGN, IMITATE 2: to have the characteristics of : RESEMBLE, PRETEND. simulated of a feigned or imitative character: MOCK, SHAM.

simulation 1a: the act or process of simulating: IMITATION, PRETENSE b: a sham object: COUNTERFEIT. 2: willful deception: COLLUSION, MISREPRESENTATION3: one that shows a superficial resemblance: ANALOGUE.

simulator: one that simulates: a device in a laboratory that enables the operator to reproduce under test conditions phenomena likely to occur in actual performance.

• Hypothesis testing, playing, existence proofs, robustness of theory.
• Simulation vs. modelling
• Science vs. engineering

Modelling, Simulation, and ALife
 C.G. Langton, preface of Artificial Life I:

`The workshop itself grew out of my frustration with the fragmented nature of the literature on biological modelling and simulation.' Volume is dedicated to C.H. Waddington:

``It has always been clear that we were not so deeply interested in the theory of any particular biological phenomenon for its own sake, but mainly in so far as it helps to a greater comprehension of the general character of the processes that go on in living as contrasted with non-living systems.''
 

Herbert A. Simon `The Sciences of the Artificial' [8] $\bullet$ ``outer environment'' and ``inner environment''.

Understanding by Simulating$\ldots$ ``the artificial object imitates the real by turning the same face to the outer system, by adapting, relative to the same goals, to comparable ranges of external tasks.''

Techniques of Simulation computer, physical, mental. ``Generally we now call the imitation `simulation' and we try to understand the imitated system by testing the simulation in a variety of simulated, or imitated, environments.''

`A simulation is no better than the assumptions built into it.' Yes, but:

1. Difficult to discover what premises imply.
2. ``$\ldots$adaptive systems have properties that make them particularly susceptible to simulation via simplified models. $\ldots$ Resemblance in behaviour of systems without identity of the inner systems is particularly feasible if aspects in which we are interested arise out of the organization of the parts, independently of all but a few properties of the individual components.''

From `Rethinking Innateness' [3]
 ``these simulations enforce a rigor on our hypotheses$\ldots$ Implementing a theory as a computer model requires a level of precision and detail which often reveals logical flaws or inconsistencies in the theory.''

``the model's innards are accessible to analysis in a way which is not always possible with human innards. In this sense, the model functions much as animal models do in medicine$\ldots$'''
 

Brooks `Artificial Life and Real Robots' ECAL 91. [1]

• ``Previously we have been very careful to avoid using simulations for two fundamental reasons
• Without regular validation on real robots there is a great danger that much effort will go into solving problems that simply do not come up in the real world with a physical robot.
• There is a real danger (in fact, a near certainty) that programs which work well on simulated robots will completely fail on real robots$\ldots$
• Force the use of `adaptive elements' to calibrate for robot variations.
• Automatically tune simulation.
• Co-evolution of morphology and controller?

Floreano and Mondada
 Hardware Solutions for Evolutionary Robotics, 1998. [4]

``Despite the importance of keeping in mind the hard constraints of operating with physical robots$\ldots$

Why simulations: The practical reason

.

Researchers with different backgrounds, need technical skills & local support for robots. In universities software writing is considered `without costs', robots cost.

Why Simulations: The strategic reason.

Faster than real-time.

• Start in simulation, then continue on robot.
• Combine evaluations on both simulated and real robots.
• Devise a simulation that allows good transfer to real robot.

Jakobi
 Controllers evolved in simulation often don't work on real robot. Even when they do, the realism required makes the software take ages to develop and it runs slowly.

Initial `ENVELOPE OF NOISE HYPOTHESIS':

• In simulation, just add lots of noise to everything.
• `Paper over the cracks' between sim and real.
• Doesn't work.

Refine: `MINIMAL SIMULATIONS.' Model only those robot-environment interactions necessary to underpin the desired behaviour. Everything else made unreliable through careful use of randomness.
  Minimal Simulations [6]

1. Precisely define the behaviour.
2. Identify the real-world base set.
3. Build a model of the way in which the members of the base set interact with each other and react to controller output (when the robot is performing the behaviour).
4. Build a model of (enough of) the way in which the members of the base set affect controller input (when the robot is performing the behaviour).
5. Define a suitable fitness test.
6. Ensure that evolving controllers are base set exclusive.
7. Ensure that evolving controllers are base set robust.

Simulations and Reality: Old AI thorns

• Is simulated intelligence real intelligence?
• Symbol grounding, `embodiment' response to Chinese room problem.
• Get wet in simulated thunderstorm?