Task-Driven Species in Evolutionary Robotic Teams - Foundations on Natural and Artificial Computation - page 139

ASiCo basic elements and operation: there are two types of elements or
independent components in an ASiCo scenario: Individuals and Environment
components . Environment components are elements whose parameters are fixed
(i.e. rocks or walls). Individuals, on the other hand, are active elements contain-
ing parameters that change over time and which may be evolved. They include a
state, a set of actuators, some associated to actions that modify their own state
parameters (like sensing or updating internal parameters), and others that mod-
ify state parameters of other elements in the scenario, and a control system that
is going to manage the actions it is executing during operation. The rules that
define the interactions among elements and between elements and the scenario
are called the Interaction Set . Interactions usually imply cost in terms of energy
and sometimes may result in an energetic gain. Consequently, the operation of
the scenario and of all of the participating components is regulated through the
generation and use of energy . Thus, one of the main elements in ASiCo is the
energy flow strategy, which represents the rules that regulate energy variations
and transmission between the individuals and the scenario and vice versa. We
have taken inspiration from the studies of utility functions and their distribution
among individuals in order to structure the energy dynamics of the environment
to guide evolution to the task objective. Specifically, we have used the principled
evaluation function selection procedure for evolving coordinated multirobot sys-
tems developed by Agogino and Tumer [1], which establishes a formal procedure
to obtain the individual utility function from the global function.
The other main element of ASiCo is the reproductive strategy, which is the set
of rules that regulates the reproduction process. Reproduction must be defined
for each problem and is triggered by spatial interactions together with some en-
ergetic and a nity criteria. Within ASiCo, a reproduction mechanism has been
designed for fixed size populations called Embryo Based Reproduction (EBR).
EBR has been discussed in detail in [10].
Once a problem has been modeled in terms of a scenario and the individuals
and elements of this scenario have been defined, the ASiCo algorithm itself starts
to operate. The general structure of this algorithm is displayed in Fig. 1 right.
It is conceptually divided into two concurrent processes: the interactions defined
by the model of the problem and the operators for modifying the structure
of the population; this is, creating and deleting individuals during the course
of simulation. ASiCo combines in a single process the simulation of the model
for continuously evaluating each individual and the evolutionary elements that
permit modifying the structure of the population to solve the problem.
Species emergence: one of the main features of ASiCo is self-organization
of the individuals into species if required by the task. That is, ASiCo can be
seen as an Artificial Life-based system where the individuals that make up the
population constitute the solution to the problem while they are in continuous
evolution, like in natural systems. This way, when evolution starts, the individ-
uals are random and heterogeneity is maximum but, as generations pass, they
start to reproduce and die controlled by a fitness or utility function that affects
the whole system. If the solution is unreachable, the population will become