ABSTRACT
This chapter discusses Integration of Representation Into Goal-Driven Behavior-Based Robots. The robot was tested over a period of 2 months, in over 40 trials, in a cluttered office environment. The data were gathered by attaching a marker to the base of the robot and recording its path on the floor covered with 1-square-foot tiles. The robot's control system consists of three competencies, integrated into a homogeneous, behavior-based representation basic navigation: obstacle avoidance and boundary tracing, landmark detection and map-related computation map construction, map update and path planning. The chapter describes how the mapping algorithm distinguishes between the two possibilities, and how the open-space behavior is triggered. The qualitative, procedural nature of the landmark detection algorithm adds robustness to the system by not relying on sensor precision or position control. Due to sensor noise, false negatives occurred in approximately one third of the trials when the robot, while following a boundary, failed to recognize it as a landmark.
