abstract = "This chapter demonstrates the use of genetic
programming to automatically create programs capable of
walking. Our goal was to generate programs that enabled
a six-legged insect to walk in a simulated environment.
Furthermore, we wished to accomplish this goal using a
minimum of a priori knowledge about the task of
locomotion. In every trial of every experiment, genetic
programming evolved programs capable of efficient
walking. Moreover, the performance of the evolved
programs was at least as good and often better than
that of hand-generated programs. One feature of our
solution was an extensive use of side-effecting
functions; we show that such functions simplify the
genetic programming paradigm and enable a more powerful
machine model. We also introduce a new genetic
operator, constant perturbation, that allows the
genetic programming system to fine-tune floating-point
constants as part of the selection process.",
notes = "Simulated six legged insect. GP set-leg function with
side effects. Worked! 25%constant perturbation
.9--1.1