The equilibrium-action cycle as a mechanism for design-evolution integration in autonomous behavior design

Evolution plays an increasingly important role in enabling autonomous robots to survive in a partly uncertain and changing environment. Yet the growing success of domestic and field robotics shows the importance of a design framework. The design methodology presented here explicitly addresses design and evolution integration. Mimicking the biological nervous system´s network of neurons, a comparable low-level ubiquitous element is developed, called the imbalance element. This element´s model stems from comparing diverse biological behavior phenomena from which results a mechanism called the equilibrium-action cycle. In this cycle, action is driven by imbalanced elements and is specifically aimed at restoring the stimuli necessary for elements to maintain equilibrium. The structure of the imbalance element network is first designed, leaving a set of parameters which can either be set analytically or evolved. Overall, integration of parameters are more predictable due to a clear mapping between a behavior and a parameter, as made possible by the design methodology´s focus on the equilibrium-action cycle. A simple line follower´s nervous system is designed to serve as example of the suggested design and evolution integration. Nonuniform Gaussian evolution is performed online on a real robot.