Dynamic biped walking is a difficult control problem. The design invol
ves that of the controller as well as the gait. A typical design proce
dure involves tedious analysis, careful planning, and testing. The pro
cedure is time consuming and the analysis is often based on some linea
rized model. Selection of control parameters and nominal trajectory de
termines the quality of control and in typical designs, some or all of
the parameters are selected intuitively. The result is often not the
best. If some special goal (such as to walk as fast as possible) is de
sirable, the design may become even harder. While the analytical appro
ach is not easy, one possible alternative is to obtain the optimal or
near-optimal design through parameter search. This study explores this
approach. The design of the biped controller and gait is formulated a
s a parameter search problem, and a genetic algorithm is applied to he
lp obtain the optimal design. Designs to achieve different goals, such
as being able to walk on an inclined surface, walk at a high speed, o
r walk with a specified step size have been evolved with the use of th
e genetic algorithm. Simulation results show that the genetic algorith
m (GA) is capable of finding good solutions. (C) 1997 John Wiley & Son
s, Inc.