context of the project: This small project was included in the “Animat” module of my AI master and teached by Jean-Arcady Meyer and Agnès Guillot. Its goal was to study a paper, repeat its experimentations to validate its results. Me and my friend chose to study a behavior model described in the article “Implementing Tolman’s Schematic Sowbug: Behavior-Based Robotics in the 1930’s” of Yoichiro Endo and Ronald C. Arkin (2001).
Topic of the paper: In 1939, Edward C. Tolman, a psychologist, introduced the concept of the schematic sowbug in his paper “Prediction of Vicarious Trial and Error by Means of the Schematic Sowbug” (Psychological Review). The schematic sowbug is based on Tolman’s purposive behaviorism (a extension of the behaviorism approach), and it is believed to be the first prototype in history that actually implemented a behavior-based architecture suitable for robotics. The schematic sowbug navigates the environment based on two types of vectors, orientation and progression, that are computed from the values of sensors perceiving stimuli.
The goal of the purposive behaviorism is to investigate how high-level factors, such as motivation, cognition, and purpose, were incorporated into the tight connection between stimulus and response that the prevailing behaviorist view largely ignored. Tolman derived a formula to compute the value of a behavior (B) from environmental stimuli (S), physiological drive (P), heredity (H), previous training (T), and mutuality or age(A): B = f (S, P, H, T, A)
the experimentation: As shown in the above image, two stimuli, green and red, were placed in front of and equally away from the sowbug. Initially the sowbug thinks that the green and red objects are both food objects. This experiment was set up to observe how the sowbug learns that the green stimulus is indeed a food and the red is not.