Analog Computation For Mobile Robotics Education

In this paper, we present a mobile robot design exercise that relies on simple analog circuits to accomplish tasks that are typically carried out using microcontrollers. Students are challenged to use simple analog sensors and IC's to develop a mobile robot that is attracted to a light source while avoiding obstacles. The primary outcome of this exercise is a deeper understanding of computation for mobile robots, and a clearer view of possible alternatives to embedded processors for low-cost applications. Background Mobile robotics is a well-recognized motivational vehicle for engineering education. Not only is it an enjoyable topic for many students, but it has a broad appeal due to its wide scope, including aspects of electrical, mechanical and computer engineering. Further, the design of such systems is an excellent tool for reinforcing fundamental engineering concepts. It is important for instructors in robotics to understand, however, that robotics is not just a tool to teach other aspects of engineering. Rather, it is a robust and mature discipline with applications in a wide range of fields. The precept of this paper is that students are taking a course in mobile robot design in order to be better equipped to act as robotics designers and engineers. In this paper, we address some key issues about computation that arise in robotic systems; namely, what is the appropriate type and level of computation for a task? In our present microprocessor-intense society, students often believe that digital computer control is the only way to affect intelligent-like behavior in robotic systems. We describe in this paper a set of experiments and classroom discussions that allow students to compare traditional (microprocessor-driven) mobile robotics computation with analog-only solutions using the principles of Valentino Braitenberg (from his famous text Vehicles 1 ) and the BEAM concept of Mark Tilden 2 . In the end, our students construct a fully functional analog robot, and are able to draw valuable new insight into the fundamental nature of computation for mobile systems, as opposed to the device used for these computations. This new knowledge allows students to design mobile robot computational suites that are more appropriate for given sets of objectives, and gives them a more global perspective of robot control systems. The discussion and experiments can be integrated easily into course segments involving artificial intelligence, biomimetics, or behavior-