Extracting principles from biology for application to running robots

A traditional approach to designing hardware or control sys tems is to begin with a task to be accomplished, proceed to brainstorm possible solutions, and select from the options based on their relative merits. This approach discourages settling for preconceived, suboptimal solutions. In cases where millions of years of evolution appear to point toward a particular answer, we may be tempted to abandon this methodology and adopt the biological solution. It is precisely in these situations, however, that we must be most cautious of accepting the apparent answer. Evolution tends to improve survival, but not necessarily optimize performance of any one desired task. Therefore, we should attempt to extract only the principles relevant to the task of interest and measure them against other design candidates. I propose a framework for extracting biological principles u sing optimization and adapting them for application to robot s.