Animals and their behavior have proven to be a rich source of inspiration and guidance for the design of software systems for autonomous robots. The scientific study of animal behavior has produced a wealth of models and mechanisms to explain how animals manage to successfully operate within a hostile and uncertain environment. In order to satisfy their needs and survive, living creatures must constantly decide how to apply their limited resources (e.g., their finite perceptual abilities, muscles, limited energy, etc.) to carry out numerous tasks such as feeding, sleeping, defending, mating, and so forth. Animals are able to organize their behaviors and arbitrate between them to satisfy these competing goals, they show an appropriate amount of persistence, and are suitably opportunistic [Gallistel, 1980]. Robotic creatures must confront similar challenges when going about their business of performing tasks in the real world. As with living creatures, their behavior must effectively address the issues of relevance, coherence, robustness, and persistence.
The intelligence of a robotic creature can be modeled at many different levels. The previous chapter focused on aspects of lower-level sensorimotor intelligence. This chapter focuses on modeling cognitive aspects associated with higher levels of intelligence. These include models for visual attention, motivation, decision making, and emotion. Here, a number of insights and models from the scientific study of animal behavior are introduced, particularly those that have proven to be very useful in our work in building robots that can interact with people. The sociable and expressive robot, Kismet (see Fig. 9.1), is used as a case study throughout the chapter to illustrate how ideas from scientific theories of attention, motivation, behavior, and emotion can be realized in a robotic implementation [Breazeal, 2002].
This chapter begins with models of visual attention and guided search. Then, a model of animal behavior that has been very influential in the design of robotic behavior systems that incorporates both motivational and perceptual factors is presented. How behaviors are organized into structures that allow animals (and robots) to appropriately decide which behavior to execute and for how long is discussed. These ideas are then extended to show how this framework can be used to implement basic emotive responses.
Online access to SPIE eBooks is limited to subscribing institutions.