Jason Fleischer.  Biomimetic design of a cooperative mobile robot system for a foraging task, M.S. thesis, Colorado State University, 1999.

ABSTRACT

Traditional robot design methodologies often have problems dealing with dynamic and uncertain environments.  Behavior-based robotic architectures have gained acceptance in the last ten years as a viable approach to this problem, but there are few formal design tools in the field. Designers often rely on their own experience or examples of other similar systems to create a new robotic system.  In many cases, the design of a behavior-based robot is said to be inspired by how a natural species accomplishes a similar task. These designers are trying to capture the dynamics of the natural system's task accomplishment, and re-create a qualitatively similar set of dynamics in the artificial system. Unfortunately, this process is ad hoc, not well understood, and still relies mostly on the experience of the designer. This thesis develops a methodology for aiding the designer of behavior-based robotic systems by mimicking biology, called biomimetic design.  It examines the issues involved in designing a robotic system with biomimetic methods, and presents an abstraction of the translation between the two worlds.  This abstraction is used to better understand the process and pitfalls of re-creating the dynamics of the natural system in an artificial one, and to suggest a methodology for creating such systems. The methodology is applied to the design of a cooperative system of six small mobile robots to find and retrieve ten target objects, of unknown location, in an 8m x 8m environment. The robots are designed to mimic both the foraging division of labor and the path integration navigation methods of desert ants (genus Cataglyphis).  Testing shows that the navigation system is able to navigate the robots successfully back to home base at rates of up to ten robots every ten minutes. Results also indicate that their task performance is likely to be further improved by the addition of landmark recognition abilities, as demonstrated by experienced Cataglyphis foragers. Finally, the biomimetic design model is discussed, and directions for future work are suggested for both the implemented robotic system and the biomimetic design model.