A generic design methodology for stochastic systems in dynamic environments

Presents a computational design methodology (CDM) called genetic design (GD) and demonstrates its application to the design of an abstract walking robot. The challenge in designing a system like a walking robot is that a method for generating adaptive responses is necessary to permit the automatic evaluation of the object. For example, the walking robot responds to changes in terrain by modifying its gait (gait planner). However, since the vehicle´s configuration is not known a priori, its gait planner cannot be designed a priori. Thus, the object´s adaptive responses must be designed simultaneously with the object itself. This paper shows how GD can be used to simultaneously design an object and its adaptive responses